Physiological responses to mixing in large scale bioreactors

Journal of Biotechnology 85 (2001) 175–185

Physiological responses to mixing in large scale bioreactors

S.-O. Enfors a,*, M. Jahic a, A. Rozkov a, B. Xu a, M. Hecker b, B. Jurgen b,E. Kruger b, T. Schweder b, G. Hamer c, D. O’Beirne c, N. Noisommit-Rizzi d,

M. Reuss d, L. Boone e, C. Hewitt e, C. McFarlane e, A. Nienow e, T. Kovacs f,C. Tragardh f, L. Fuchs g, J. Revstedt g, P.C. Friberg h, B. Hjertager i,

G. Blomsten j, H. Skogman j, S. Hjort k, F. Hoeks l, H.-Y. Lin m,P. Neubauer m, R. van der Lans n, K. Luyben n, P. Vrabel n, A, . Manelius o

a Department of Biotechnology, Royal Institute of Technology, S-10044 Stockholm, Swedenb Department of Microbiology and Molecular Biology, Ernst-Moritz-Arndt Uni6ersitat, Friedrich-Ludwig-Jahn-Strasse 15a,

D-17487 Greifswald, Germanyc Department of Chemical Engineering, Uni6ersity College Dublin, Belfield, Dublin 4, Ireland

d Department of Biochemical Engineering, Uni6ersity of Stuttgart, Allmandring 31, D-70569 Stuttgart, Germanye Department of Chemical Engineering, Uni6ersity of Birmingham, Birmingham B15 2TT, UK

f Department of Food Engineering, Uni6ersity of Lund, Box 124, S-22100 Lund, Swedeng Department of Heat and Power Engineering, Uni6ersity of Lund, P.O. Box 118, S-221 00 Lund, Sweden

h TelTek, N-3914 Porsgrunn, Norwayi Department of Chemistry and Applied Engineering Science, Aalborg Uni6ersity-Esbjerg, Niels Bohrs Vej 8,

DK-6700 Esbjerg, Denmarkj BioGaia Fermentation AB, P.O. Box 965, S-22009 Lund, Sweden

k Scaba AB, Enhags6agen 12, PO Box 2018, S-18302 Taby, Swedenl Lonza Ltd, Valais Works, CH-3930 Visp, Switzerland

m Department of Biochemistry and Biotechnology, Martin-Luther-Uni6ersitat, Weinbergweg 16a, D-06120 Halle, Germanyn Department of Biochemical Engineering, Delft Uni6ersity of Technology, Julianalaan 67, NL-2628 BC Delft, The Netherlands

o Pharmacia & Upjohn AB, Mariefreds6agen 37, S-645 41 Strangnas, Sweden

Received 10 February 2000; received in revised form 21 June 2000; accepted 29 June 2000

Abstract

Escherichia coli fed-batch cultivations at 22 m3 scale were compared to corresponding laboratory scale processesand cultivations using a scale-down reactor furnished with a high-glucose concentration zone to mimic the conditionsin a feed zone of the large bioreactor. Formate accumulated in the large reactor, indicating the existence of oxygenlimitation zones. It is suggested that the reduced biomass yield at large scale partly is due to repeated production/re-assimilation of acetate from overflow metabolism and mixed acid fermentation products due to local moving zoneswith oxygen limitation. The conditions that generated mixed-acid fermentation in the scale-down reactor also induced

www.elsevier.com/locate/jbiotec

* Corresponding author. Fax: +46-8-7231890.E-mail address: [email protected] (S.-O. Enfors).

0168-1656/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.

PII: S 0168 -1656 (00 )00365 -5

S.-O. Enfors et al. / Journal of Biotechnology 85 (2001) 175–185176

a number of stress responses, monitored by analysis of mRNA of selected stress induced genes. The stress responseswere relaxed when the cells returned to the substrate limited and oxygen sufficient compartment of the reactor.Corresponding analysis in the large reactor showed that the concentration of mRNA of four stress induced genes waslowest at the sampling port most distant from the feed zone. It is assumed that repeated induction/relaxation of stressresponses in a large bioreactor may contribute to altered physiological properties of the cells grown in large-scalebioreactor. Flow cytometric analysis revealed reduced damage with respect to cytoplasmic membrane potential andintegrity in cells grown in the dynamic environments of the large scale reactor and the scale-down reactor. © 2001Elsevier Science B.V. All rights reserved.

Keywords: Scale-up; Escherichia coli ; Mixed-acid fermentation; Stress responses; Concentration gradients; CFD

1. Introduction

It is often observed that the biomass yield onthe C/energy source is decreased by the scale-upof aerobic processes. In a baker’s yeast process,the biomass yield on molasses increased by about7% when the process was scaled down from 120m3 to 10 l using an identical strain, medium andprocess control. At the same time, the gassingpower of the yeast in sweet dough, but not in astandard dough, was decreased by the scale-down,indicating multiple effects on cell physiology(George et al., 1998). In an Escherichia coli basedrecombinant protein process the biomass yield onglucose and the maximum cell density reached inthe process dropped by about 20% when scalingup from 3 l to 9 m3 (Bylund et al., 1998). Further-more, the pattern of acetate formation changed.However, no mechanism behind these phenomenahas been presented backed up with experimentaldata. In a methanol limited chemostat the pulsedaddition of the methanol feed, but not of the restof the medium, created a large drop in biomassyield of Pseudomonas on methanol and an in-creased yield of carbon dioxide on methanol, indi-cating that the fluctuating methanol concentrationcaused increased combustion of the methanol forenergy metabolism, in other words an increasedmaintenance metabolism (Brooks and Meers,1973).

Also in large scale fed-batch processes it can beexpected that the concentration of the limitingsubstrate fluctuates due to the difficulties in mix-ing the feed solution fast enough. This feed solu-tion is usually applied in as concentrated form aspossible to minimise the dilution effects. The feedconcentration is about 500 g l−1 in the case of

glucose. The limiting substrate concentration in afed-batch reactor depends on the sugar uptakekinetics, but when the cells grow at half maximumspecific rate, the concentration is equal to thesaturation constant of the Monod model, whichfor glucose is about 5 mg l−1 for E. coli andabout 180 mg l−1 for Saccharomyces cere6isiae.This creates a very dynamic situation in a largefed-batch bioreactor: When the individual cellrepeatedly passes close to the feed point it must beexposed, for a few seconds, to a much highersugar concentration than the mean concentrationin the reactor. If the residence time of the cell inthis feed zone is long enough it may result inrapid utilisation of the sugar and concomitantlocal oxygen limitation when the cell density ishigh enough.

Furthermore, many micro-organisms respondby overflow metabolism when exposed to glucoseconcentration above a critical level. Thus, S. cere-6isiae produces ethanol (Sonnleitner and Kappeli,1986) and E. coli produces acetate (Xu et al.,1999b) at glucose concentrations above about 30mg l−1. Under conditions when oxygen-limitationis created, anaerobic reactions and concomitantstress responses may result (Smith and Neidhardt,1983). However, such responses, which are well-known from microbiological research, havemostly been studied with a much longer time scalethan the residence times of individual cells inhypothetical oxygen-limitation zones in a largebioreactor.

During the period 1996–1999, a joint Europeanproject was focused on the interaction betweenthe fluid dynamics and physiological responses ina bioreactor. Reported here are data that supportthe hypothesis that moving oxygen-limiting zones

S.-O. Enfors et al. / Journal of Biotechnology 85 (2001) 175–185 177

in large fed-batch bioreactors induce bothmetabolic and stress responses. Additionally it isproposed that such conditions may account forthe reduced biomass yield on scale-up and alteredphysiological properties of cells grown underlarge-scale conditions.

2. Materials and methods

2.1. Organism

E. coli W3110 was used in all experiments.

2.2. Medium

A glucose-ammonia-mineral salts medium wasused (Xu et al., 1999a).

2.3. Feeding

The process was started with about 500 mg l−1

glucose and an exponential feed with a 454 g l−1

glucose solution was started within 1 h to forcethe cells to grow with a specific growth rate ofabout 0.3 h−1, which limited the acetate forma-tion from over-flow metabolism. After 14 h at acell density of about 11 g l−1 the feed was keptconstant until the end when about 35–40 g l−1

cell dry weight was obtained.

2.4. Reactors

The laboratory scale reference cultivations wereperformed in conventional stirred tank bioreac-tors. The large scale process was performed in a30 m3 (total volume) reactor fitted with fourRushton turbines and with the feeding close to thetop impeller or fitted with four Scaba impellersand feeding at the top or between the middleimpellers. The culture volume was kept between21 and 22 m3 by intermittent withdrawals. Thethird type of reactor was a scale-down reactorcomposed of a standard stirred tank reactor(STR) fitted with a static mixer loop with a liquidvolume of 10% (aerobic loop) or 15% (non-aer-ated loop) of the total reactor liquid volume. Thestatic mixer created plug-flow behaviour in the

loop in spite of the vigorous mixing (George etal., 1993). The culture was circulated with a resi-dence time of about 54 s in the plug-flow loop(PFR). This time should be seen in perspective ofthe mixing time of the 22 m3 reactor, which variesin the range of 1–2 min. All glucose was fed atthe inlet of the PFR resulting in an inlet glucoseconcentration about 1 g l−1. Four samplingpoints along the PFR permitted analysis of thekinetics of the metabolic and stress responses.

2.5. Analyses

The analytical procedures are described else-where: mixed acid fermentation products (Xu etal., 1999b); messenger RNA (Schweder et al.,1999); flow cytometry (Hewitt et al., 1999).

2.6. CFD simulations

CFD simulations were made as described previ-ously (Revstedt et al., 1998).

3. Concentration gradients

Simulations using computational fluid dynam-ics (CFD) according to the Large Eddy principleshowed that considerable glucose gradients couldbe expected when a 500 g l−1 glucose solutionwas added at the top stirrer in the 22 m3 bioreac-tor (Fig. 1). The figure shows that the compart-ment around the top stirrer has a much higherglucose concentration than the bulk of the reac-tor. The Large Eddy Simulation also permits cal-culation of the concentration as function of time.Fig. 2A shows that according to this calculation,considerable fluctuations in glucose concentrationcould be expected in the vicinity of the feed point.Such rapid fluctuations were indeed experimen-tally observed, as shown in Fig. 2B. In this case, acontinuous flow of biosuspension was sampledinto a rotating sample collector with pre-weighedtubes containing cold perchloric acid for quench-ing of the metabolism (Larsson and Tornkvist,1996). In this way several series of glucose analy-ses were made, each including 380 glucose analy-ses with a mean interval of 470 ms.


According to these analyses the cells were fre-quently exposed to peak glucose concentrationsseveral times higher than the mean glucose con-centration. Even at the mean glucose concentra-tion in Fig. 2B, acetate accumulation is expectedfrom over-flow metabolism of E. coli. However, atthe time in the process when these analyses weremade (at 20 h in Fig. 3), the acetate concentrationwas low and did not increase, indicating eitherthat the exposure of the cell to the high glucoseconcentration was not long enough to result inacetate formation or that produced acetate was

Fig. 2. A: Large Eddy Simulation of glucose concentrationfluctuations at one point close to the feed point in the 22 m3

bioreactor (no reaction included). B: Analysis of glucose con-centration fluctuations close to the vessel wall 45 (pos 1) and135 (pos 2) degrees after the feed point at the level of the feedpoint in a 22 m3 culture, as indicated in Fig. 1. C: Correspond-ing analyses in two similar processes with the difference thatfour Scaba impellers were used and the feeding points were atthe top impeller or between the mid impellers. Sampling wasmade close to the vessel wall 0.5 m above the feed points. Celldensity was about 18 g l−1 at the time of sampling in allexperiments. Mean sample interval was 470 ms, which meansthat there are about 21 analyses between each tick on the timescale.

Fig. 1. Large Eddy Simulation of instantaneous glucose con-centration in a 22 m3 bioreactor fed with a 500 g l−1 solutionat a rate of 180 l h−1, that was used in the fed-batch process.Four Rushton turbines and the location of the feed point areindicated. The simulation did not include the microbial con-sumption of the glucose and can therefore only be used toillustrate the mixing efficiency (Methodology according toRevstedt et al., 1998).


re-assimilated in other compartments of thereactor.

Space-dependent concentration gradients, withdeclining glucose concentration with increasingdistance from the feed point has earlier beendemonstrated in large-scale processes (Larsson etal., 1996; Bylund et al., 1998). In the presentinvestigation, the mean glucose concentrationsover the whole fed-batch phase were 27, 34 and 57mg l−1 at the bottom, mid and top samplingports, respectively, when the Rushton turbineswere used and the mixing time was 1–2 min.Thus, a cell circulating in this bioreactor wassubjected to relatively long term exposure to highglucose concentrations when passing the feedpoint compartment, due to space dependent gra-dients, but also to much shorter and more fre-quent concentration peaks when exposed to thefluctuations shown in Fig. 2B–C. In the two

experiments with Scaba impellers using top andmidpoint feed, however, no significant differencein the mean concentration was observed at thethree sampling levels but the glucose concentra-tions were much more evenly distributed at allthree sampling levels. This is also reflected by thelarger peeks observed during the fast samplingshown in Fig. 2C.

The question is then: does the cell respond tothese rapid changes in glucose concentration? Ifso, what kind of reactions is induced? A first hintto the answer was obtained by the analysis ofmixed-acid fermentation products in the 22 m3

cultures. Fig. 3 shows the concentrations of ace-tate and formate. Only traces of lactic acid andsuccinate were observed while the ethanol concen-tration was constant at about 33 mg l−1 duringthe constant feed phase. When the cell densityreached about 30 g l−1 formate started to accu-mulate rapidly while the other mixed-acidmetabolism products did not accumulate. Theinitial, transient acetate accumulation can be ex-plained by over-flow metabolism. Correspondingacetate was produced also in the laboratory scaleprocess (Xu et al., 1999a). The second acetatepeak was probably caused by mixed-acid fermen-tation, since it was accompanied by a formatepeak. However, the rapid accumulation of for-mate starting at about 30 h was not accompaniedby any corresponding accumulation of othermixed-acid fermentation products. Thus, somebut not all the anaerobic E. coli products ap-peared in spite of the fact that the mean dissolvedoxygen tension, analysed at the bottom and themiddle sampling ports, were 53 and 25% air sat.,respectively. Mixed-acid metabolism, however,produces one acetate per formate through thesplit of pyruvate by pyruvate formate lyase. For-mate can be further split to hydrogen and carbondioxide by formate hydrogen lyase. Therefore theratio acetate to formate should be equal to orlarger than one. The obvious possibility that ace-tate and also other anaerobic products wereformed in oxygen limitation compartments andthen re-assimilated in oxygen sufficient compart-ments was investigated with the scale-downreactor.

Fig. 3. Data from a 22 m3 E. coli culture with 4 Rushtonimpellers and feeding at the top impeller. Formate first accu-mulated slowly but then rapidly when the cell density reachedabout 30 g l−1. No corresponding accumulation of othermixed acid products was observed. DOT was analysed at thebottom (1.1 m) and middle (4.1 m) ports.


Fig. 4. Principle of the scale-down reactor for analysis ofphysiological responses in a zone of low glucose input (upperpanel, with feeding in the STR) or a zone of high glucoseconcentration (lower panel, with feeding at the inlet of thePFR). The plots show schematically the concentration fluctua-tions with respect to glucose and DOT (in the high-glucosezone) experienced by an individual cell circulating between theSTR and the PFR.

time in the PFR was 56 s and the mean residencetime in the STR was 9 min. By sampling at portsalong the PFR, analyses were made 14, 28, 32 and56 s after entrance into the PFR. In this way thedynamics of mixed acid fermentation in a high-glucose zone was obtained.

The glucose was fed at the inlet of the PFR andno extra oxygen was added to the PFR, besidesthe amount entrapped with the incoming biosus-pension. The inlet glucose concentration was thenabout 2 g l−1 during the constant feed phase, butit declined gradually along the PFR. No ethanolor succinic acid formation was observed in thePFR but acetate, lactate and formate appeared asshown in Fig. 5. In the oxygen sufficient STR,however, mainly formate accumulated, indicatingthat acetate and lactate were readily re-assimilatedthere.

Based on data of the specific product formationrate in the oxygen limited PFR (Fig. 5) andaccumulation in the STR (Fig. 6A), the totalproduction of these products could be calculated(Fig. 6B). It is obvious that the enzyme systemsinvolved in the mixed-acid fermentation in E. coli,though inhibited and in several cases repressed byoxygen, are present at a high enough concentra-tion in the aerobically grown cell to producemeasurable amounts of product within secondswhen exposed to anaerobic conditions. Further-more, in the oxygen sufficient compartment, lac-tate and acetate are more easily re-assimilatedthan formate and this results in formate being themain product in the total scale-down reactor sys-tem. However, also most of the produced formatewas re-assimilated. The relative amounts of re-as-similation in this experiment were greater than99% for lactate, 97% for acetate and 90% forformate. This may explain why formate suddenlyappeared without a corresponding increase in ac-cumulation of other mixed acid products in the22 m3 culture when the cell density became high.

5. Stress responses

The fact that the large bioreactor containedcompartments that created conditions for anaero-bic responses made it plausible that the anaerobi-

4. Dynamics of the metabolic responses

A scale-down reactor was used to investigatethe dynamics of metabolic and stress responseswhile the cells passed through a high-glucose con-centration zone. The principle of this scale-downreactor is shown in Fig. 4. It is composed of aplug flow reactor (PFR) fitted with a static mixerto permit high oxygen transfer rate (KLa=600–1000 h−1), connected as a loop to a stirred tankreactor (STR) (George et al., 1993). The residence


cally induced stress response also may be induced.The short exposure time made it, however,difficult to analyse the responses at the proteinlevel. Instead, transcriptional analysis was used

Fig. 6. Accumulation (top panel) and production (bottompanel) of mixed acid fermentation products in the scale-downfermenter with glucose feeding at the PFR inlet. Data onaccumulation was obtained from chemical analyses while theproduction data were derived from data on production in thePFR. Data from Xu et al. (1999a). Note the different concen-tration scales.

Fig. 5. Mixed-acid fermentation in the PFR-loop of the scale-down reactor. The glucose feed was applied at the inlet of thePFR, creating oxygen limitation there. After about 40 s thelactate production rate increased rapidly, and this was accom-panied by an increase in the glucose consumption (Xu et al.,1999a), possibly due to the fact that lactate formation regener-ates NAD+ , which is not the case with acetate and formateproduction.

(Schweder et al., 1999). A number of stress in-duced genes were selected, including not only theoxygen starvation induced pyruvate formate lyase(pfl) and fumarate reductase ( frd), but also anosmosensitive proline transporter (proU), a chap-erone (dnaK) and the protease subunit clpB. Theconcentration of mRNA from these genes wasanalysed along the PFR in the scale-down reactoroperated with feeding in the PFR inlet. In thisway the cells were repeatedly exposed to pulses ofhigh glucose concentration and an increasing de-gree of oxygen limitation with increasing cell den-sity during the process.

An example of the response of proU duringsuch an experiment is shown in Fig. 7. Two main


conclusions were drawn from the experiments(Schweder et al., 1999): Firstly, the concentrationof mRNA of all genes (pfl, frd, proU and clpB)increased along the PFR. In some cases and espe-cially at high cell concentration this response wasobtained already after 14 s in the PFR (the firstsampling port). Secondly, these stress responseswere not sustained, since the mRNA concentra-tion did not increase in the STR during the exper-iment in spite of a continuous production ofmRNA in the loop. This means that the meanresidence time of 10 min in the glucose limitedand oxygen sufficient STR was enough to relaxthe stress response. This is in agreement with theshort half life of mRNA, that is only a fewminutes.

Analyses of mRNA from the stress inducedgenes frd, uspA, dnaK and pfl were also made insamples drawn from three levels (1.1, 4.1 and 6.5m from the bottom) in the 22 m3 culture with fourRushton turbines and feeding at the top impeller(Schweder et al., 1999). Samples were drawn fromthese ports at four time intervals during the pro-cess. Since the relative concentrations of some ofthese mRNAs changed with process time, the datawere normalised to 100% at the bottom port ateach time. Fig. 8 shows that the mRNA concen-tration in all cases was lowest at the bottom port,

Fig. 8. Analysis of mRNA concentrations of four stress sensi-tive genes at three levels of the 22 m3 fed-batch culture of E.coli. The concentrations from four sampling events coveringcell densities from 6 to 35 g l−1 cell dry weight were nor-malised to 100% at the bottom port. Data from Schweder etal. (1999).

which was also the port most remote from thefeed zone at the top impeller. Based on the scale-down reactor data it is tempting to assume thatthe passage of the cells through the upper reactorcompartment where glucose concentration washighest induced stress responses that are switchedoff when the cells moved on from thisenvironment.

6. Are there beneficial effects of large scaleconditions?

A common feature of high cell density fed-batch cultures of E. coli is that the viable count(c.f.u.) rapidly drops when the specific growth rateapproaches about 0.05 h−1. The mechanism be-hind this phenomenon is not fully understood, butit seems as if at least part of the drop in c.f.u. iscaused by segregation to non-dividing cells ratherthan cell death (Andersson et al., 1996). Such adrop in viable count was not observed in the three22 m3 cultures, though it was evident in thelaboratory scale control experiments.

Using multi-parameter flow cytometry it is pos-sible to classify the physiological state of individ-

Fig. 7. Analysis of stress responses in the scale-down reactor.The open bars represent analyses of mRNA from proU alongthe PFR and the figures under the bar are the time elapsedfrom entrance into the PFR (in seconds). The dashed barsrepresent analyses in the STR compartment, and the figuresbelow show the process time (in hours). The first series ofanalyses in the PFR was made at 12.5 h and the last series at18 h process time.


Fig. 9. Flow cytometry characterisation of membrane function of E. coli cells during a laboratory fed-batch culture. Staining byBOX (bis-oxanol) indicates depolarised cell membrane and staining by PI (propidium iodide) indicates permeabilised membrane.Left hand figure shows staining pattern in the beginning of the process and the right hand figure shows staining at the end of theprocess. Data from Hewitt et al. (1999).

ual bacterial cells beyond culturability, based onthe presence or absence of an intact polarisedcytoplasmic membrane. The presence of both anintact polarised cytoplasmic membrane and activetransport systems are essential for a fully func-tional healthy cell. Propidium iodide (PI) binds toDNA but cannot cross an intact cytoplasmicmembrane, whilst BOX is cationic, lipophilic andaccumulates intracellularly provided the cytoplas-mic membrane is depolarised. During a labora-tory scale well mixed fed-batch process suchtechniques revealed a progressive change in cellphysiological state, based on the simultaneousuptake of both PI and BOX. Towards the end ofthe fermentation three main sub-populations ofcells can be distinguished, corresponding tohealthy cells, no staining, cells with no membranepotential, stained with BOX and cells with perme-abilised membranes (i.e. dead cells), stained withboth PI and BOX. Whereas at the start of thefermentation there are only the healthy cells, andno staining can be identified (Hewitt et al., 1999;Fig. 9).

When the same technique was used at the 22 m3

scale no progressive change in individual cell

physiological state was observed. All the cellswere healthy with intact polarised cytoplasmicmembranes, whilst about 15% of the cells in thecorresponding well mixed laboratory scale processtook up both PI and BOX indicating that theyhad depolarised, permeabilised cytoplasmic mem-branes and were in fact dead. When similar mea-surements were made in the scale-down reactor it

Fig. 10. Viability analysis using flow cytometry in a homoge-neous laboratory scale process, two scale-down reactor pro-cesses and in a 22 m3 process. Data adopted from Hewitt et al.(1999).


showed that increasing degree of heterogeneityactually protected the cells from death (Fig. 10).

7. Concluding remarks

This study reveals that a large-scale fed-batchbioreactor provides a very dynamic environmentfor the cells circulating in the reactor. If feeding ismade at one point, the cells are exposed to twotypes of glucose gradients: firstly by exposure tothe rapidly fluctuating concentrations in the feedzone, and secondly by circulating around the reac-tor where the mean concentration declines withincreasing distance from the feed zone. At leastsomewhere, probably in the compartments repre-sented by the glucose peaks in the feed zone, therapidly moving high glucose zones create corre-sponding oxygen limitation zones, since formicacid was produced in the large-scale reactor. For-mic acid cannot be produced as the only anaero-bic product, and therefore substantial amounts ofother mixed-acid products should have been pro-duced in oxygen insufficient zones and re-assimi-lated in oxygen sufficient zones. This may be onemechanism behind the loss in biomass yield ob-served at scale-up of E. coli processes. Also with-out oxygen limitation zones, a correspondingproduction and re-assimilation of acetate fromoverflow metabolism could be expected and per-haps also contribute to the biomass yield loss.

Another mechanism that could contribute tothe reduced yield is the observed induction fol-lowed by relaxation of stress responses when thecells repeatedly pass through zones with highglucose and corresponding low oxygen concentra-tion. This means that the cells continuously try toadapt to a situation that lasts for only a fewseconds. It is not difficult to imagine that this mayhave consequences for the physiology of the cell.

If loss of biomass yield is a problem in aprocess, these effects can probably be reduced byfeeding at several points in the reactor and byusing lower concentration of the feed. However,we observed also other effects of insufficient mix-ing that indicate that the dynamic environment inlarge bioreactors may be beneficial. The reducedcell death, as analysed by flow cytometry, may

very well balance the lost cell yield at scale-up. Amajor conclusion therefore is that the conditionsin a large-scale fed-batch bioreactor induce multi-ple physiological responses in the cells that arenot observed in homogeneous lab-scale reactors.

Acknowledgements

This paper is based on results from a projectfinanced by the Biotechnology Programme of Eu-ropean Commission, contract no BIO4-CT95-0028.

References

Andersson, L., Yang, S., Neubauer, P., Enfors, S.-O., 1996.Impact of plasmid presence and induction on cellularresponses in fed-batch cultures of Escherichia coli. J. Bio-technol. 46, 255–263.

Brooks, J.D., Meers, J.L., 1973. The effect of discontinuousmethanol addition on the growth of a carbon-limited cul-ture of Pseudomonas. J. Gen. Microbiol. 77, 513–519.

Bylund, F., Collet, E., Enfors, S.-O., Larsson, G., 1998. Sub-strate gradient formation in the large-scale bioreactor low-ers cell yield and increases by-product formation. Bioproc.Eng. 18, 171–180.

George, S., Larsson, G., Enfors, S.-O., 1993. A scale-downtwo-compartment reactor with controlled substrate oscilla-tions: Metabolic response of Saccharomyces cere6isiae.Bioproc. Eng. 9, 249–257.

George, S., Larsson, G., Olsson, K., Enfors, S.-O., 1998.Comparison of the Baker’s yeast process performance inlaboratory and production scale. Bioproc. Eng. 18, 135–142.

Hewitt, C.J., Nebe-Von Caron, G., Nienow, A.W., McFar-lane, C.M., 1999. The use of multi-parameter flow cytome-try to compare the physiological response of Escherichiacoli W3110 to glucose limitation during batch, fed-batchand continuous culture cultivations. J. Biotechnol. 75,251–264.

Larsson, G., Tornkvist, M., 1996. Rapid sampling, cell inacti-vation and evaluation of low extracellular glucose concen-trations during fed-batch cultivation. J. Biotechnol. 49,69–82.

Larsson, G., Tornkvist, M., Stahl-Wernersson, E., Tragardh,C., Noorman, H., Enfors, S.-O., 1996. Substrate gradientsin fed-batch bioreactors: origin and consequences. Bioproc.Eng. 14, 281–289.

Revstedt, J., Fuchs, L., Tragardh, C., 1998. Large eddy simu-lations of the turbulent flow in a stirred reactor. Chem.Eng. Sci. 53, 4041–4053.


Schweder, T., Kruger, E., Xu, B., Jurgen, B., Blomsten, G.,Enfors, S.-O., Hecker, M., 1999. Monitoring of genesthat respond to process-related stress in large-scale bio-processes. Biotechnol. Bioeng. 65, 151–159.

Smith, M.W., Neidhardt, F.C., 1983. Proteins induced byanaerobiosis in Escherichia coli. J. Bacteriol. 154, 336–343.

Sonnleitner, B., Kappeli, O., 1986. Growth of Saccha-romyces cere6isiae is controlled by its limited respiratory

capacity: formulation and verification of a hypothesis.Biotechnol. Bioeng. 28, 927–937.

Xu, B., Jahic, M., Blomsten, G., Enfors, S.-O., 1999a. Glu-cose overflow metabolism and mixed-acid fermentation inaerobic large-scale fed-batch processes with Escherichiacoli. Appl. Microbiol. Biotechnol. 51, 564–571.

Xu, B., Jahic, M., Enfors, S.-O., 1999b. Modelling ofoverflow metabolism in batch and fed-batch cultures ofEscherichia coli. Biotechnol. Progress 15, 81–90.

.

Physiological responses to mixing in large scale bioreactors

Documents

Transcript of Physiological responses to mixing in large scale bioreactors