gill et al 2008a

This article was published in an Elsevier journal. The attached copyis furnished to the author for non-commercial research and

education use, including for instruction at the authors institution,sharing with colleagues and providing to institution administration.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further informationregarding Elseviers archiving and manuscript policies are

encouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Biochemical Engineering Journal 39 (2008) 164176

Design and characterisation of a miniature stirred bioreactorsystem for parallel microbial fermentations

N.K. Gill a, M. Appleton b, F. Baganz a, G.J. Lye a,a The Advanced Centre for Biochemical Engineering, Department of Biochemical Engineering, University College London,

Torrington Place, London, WC1E 7JE, UKb Bioxplore, 50 Moxon Street, Barnet, Hertfordshire, EN5 5TS, UK

Received 2 August 2007; accepted 3 September 2007

Abstract

The establishment of a high productivity microbial fermentation process requires the experimental investigation of many interacting variables.In order to speed up this procedure a novel miniature stirred bioreactor system is described which enables parallel operation of 416 independentlycontrolled fermentations. Each miniature bioreactor is of standard geometry (100 mL maximum working volume) and is fitted with a magneticallydriven six-blade miniature turbine impeller (di = 20 mm, di/dT = 1/3) operating in the range 1002000 rpm. Aeration is achieved via a sintered spargerat flow rates in the range of 02 vvm. Continuous on-line monitoring of each bioreactor is possible using miniature pH, dissolved oxygen andtemperature probes, while PC-based software enables independent bioreactor control and real-time visualisation of parameters monitored on-line. Inaddition, a new optical density probe is described that enables on-line estimation of biomass growth kinetics without the need for repeated samplingof individual bioreactors. Initial characterisation of the bioreactor involved quantification of the volumetric oxygen mass transfer coefficient asa function of agitation and aeration rates. The maximum kLa value obtained was 0.11 s1. The reproducibility of E. coli TOP10 pQR239 andB. subtilis ATCC6633 fermentations was shown in four parallel fermentations of each organism. For E. coli (1000 rpm, 1 vvm) the maximumspecific growth rate, max, was 0.68 0.01 h1 and the final biomass concentration obtained, Xfinal, was 3.8 0.05 g L1. Similarly for B. subtilis(1500 rpm, 1 vmm) max was 0.45 0.01 h1 and Xfinal was 9.0 0.06 g L1. Biomass growth kinetics increased with increases in agitation andaeration rates and the oxygen enrichment for control of DOT levels enabled max and Xfinal as high as 0.93 h1 and 8.1 g L1 respectively to beachieved. Preliminary, scale-up studies with E. coli in the miniature bioreactor (100 mL working volume) and a laboratory scale 2 L bioreactor(1.5 L working volume) were performed at matched kLa values. Very similar growth kinetics were observed at both scales giving max values of0.94 and 0.97 h1, and Xfinal values of 5.3 and 5.5 g L1 respectively. The miniature bioreactor system described here thus provides a useful toolfor the parallel evaluation and optimisation of microbial fermentation processes. 2007 Elsevier B.V. All rights reserved.

Keywords: Miniature bioreactor; Parallel operation; Fermentation; On-line monitoring

1. Introduction

The design and optimisation of industrial fermentationprocesses requires the experimental investigation of many inter-acting biological and physical variables. Advances in metabolicengineering and protein evolution techniques now enable therapid creation of large libraries of recombinant microorganisms[1]. These are normally evaluated in parallel microwell culturesand a small number of promising strains identified based on sim-ple initial screens for product yield or activity [2]. For the chosen

Corresponding author. Tel.: +44 20 7679 7942.E-mail address: [email protected] (G.J. Lye).

strains product synthesis can be further enhanced by study of cul-ture medium composition, nutrient feeding regimes and physicalvariables such as temperature, pH and dissolved oxygen levels.Here large numbers of parallel shake flask or stirred-bioreactorexperiments must be performed because of the number of exper-imental variables requiring investigation per strain. Finally, oncea production strain is identified, further experimental investiga-tion over a range of scales is necessary to establish the operatingboundaries and robustness of the process for validation purposes[2]. At this stage virtually all processes would be performed instirred bioreactors because of the dominance of this design inthe chemical and biopharmaceutical sectors.

Small scale bioreactor systems, that enable parallel and auto-mated operation of several fermentations simultaneously, have

1369-703X/$ see front matter 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.bej.2007.09.001


N.K. Gill et al. / Biochemical Engineering Journal 39 (2008) 164176 165

Nomenclature

CL concentration of dissolved oxygen in fermenta-tion broth (kg m3)

Cp normalised oxygen concentrationC* saturated dissolved oxygen concentration

(kg m3)dB width of baffle (mm)di diameter of impeller (mm)dT diameter of vessel (mm)DOT dissolved oxygen tension (%)hi distance between centre line of impeller and base

of vesselkLa volumetric oxygen mass transfer coefficient (s1)mO2 oxygen required for cell maintenance

(mol gDCW1 h1)OTRmax maximum oxygen transfer rate (mmol L1 h1)OURmax maximum oxygen uptake rate (mmol L1 h1)P power input (W)t time (s)tm mass transfer time, 1/kLa (s)V volume (L)vvm volumetric air flow per volume of broth per minuteXfinal final biomass concentration (g L1)YX/O2 yield of biomass on oxygen (g mol1)

Greek lettersmax maximum specific growth rate (h1)p probe response time (s)S superficial gas velocity (m s1)

the potential to increase the rate at which the necessary exper-iments are performed thus reducing fermentation developmenttimes and costs [3]. In recent years various designs of paral-lel miniature bioreactor systems have been reported includingstirred tank bioreactors, bubble columns and shake flasks [47].The key design features of many of these will be discussed indetail later (Section 3.5). While each bioreactor design aims tosatisfy all the requirements for rapid fermentation process devel-opment there is normally a trade-off between throughput and theinformation content of the data obtained from each experiment[2]. This is most clear when comparing related developmentsin microwell fermentations [610] with the small scale stirredbioreactor designs considered here. Irrespective of the approachtaken for high throughput fermentation process developmenthowever, it will be important that advances in throughput arematched by the creation of microscale downstream process-ing operations. In this regard the generation of quantitativeprocess data from a number of microwell-based downstreamunit operations has recently been reported [3,11,12] as has theautomated operation of linked process sequences [13,14]. Theautomated linkage of upstream and downstream operations atsuch scales now offers the potential for the integrated optimisa-tion of the entire process from fermentation through to purifiedproduct.

In this work we report on the design, instrumentation andcharacterisation of a novel miniature stirred bioreactor system(maximum working volume 100 mL) that can support the par-allel operation of 416 independently controlled bioreactors.Each bioreactor is of standard geometry being agitated by asingle six-blade miniature turbine impeller. Continuous on-linemonitoring and control of pH, dissolved oxygen and tempera-ture is achieved using miniature steam sterilisable probes whilea novel optical density (OD) probe allows on-line estimationof biomass growth kinetics. Bioreactor oxygen mass transfercharacteristics have been studied as a function of agitation andaeration rates giving a volumetric oxygen mass transfer coef-ficient, kLa, of up to 0.11 s1. Parallel E. coli TOP10 pQR239and B. subtilis ATCC6633 fermentations have been shown tobe highly reproducible. In the case of E. coli TOP10 it has alsobeen shown how biomass growth kinetics and yields vary asa function of agitation and aeration conditions. Finally initialresults using constant kLa as a basis for scale-up have shownthat results obtained in the miniature bioreactor can accuratelybe reproduced in a conventional laboratory scale stirred biore-actor.

2. Materials and methods

2.1. Chemicals and microorganisms

The chemicals used in this work were obtained from BDH(Dorset, UK) unless otherwise stated and were of the highestpurity available. RO water was used for all experiments. E. coliTOP10 pQR239, which expresses cyclohexanone monooxyge-nase (CHMO) under the control of an l-arabinose promotor[15], and B. subtilis ATCC6633 [16] were used in the fermenta-tion studies. Cells were maintained as 40% (v/v) glycerol stocksolutions at 80 C.

2.2. Miniature bioreactor design and instrumentation

2.2.1. Design of individual bioreactorsThe miniature bioreactor was designed to be geometrically

similar to conventional laboratory scale stirred fermenters. Eachbioreactor consisted of a borosilicate glass vessel (100 mLmaximum working volume), allowing visual inspection of thecontents of the bioreactor, and was sealed with a stainless steelhead plate, as shown in Fig. 1.

Mixing of the bioreactor was achieved by a magneti-cally driven, six-blade miniature turbine impeller (di = 20 mm,di/dT = 1/3) fabricated from PEEK (poly-ether ether ketone).There was a distance of 20 mm between the centre line of theRushton turbine and the base of the vessel (di/hi = 1). Eightcylindrical magnets (3 mm in diameter and 2 mm long) wereuniformly distributed and embedded into a small PEEK disc10 mm below the miniature turbine as shown in Fig. 1(A).The whole impeller structure was mounted onto a hollow,non-rotating stainless steel stirrer shaft that screwed into theheadplate of the bioreactor. This design eliminated the need forrotating mechanical seals and reduced the area occupied on theheadplate. The PEEK impeller assembly, which was magneti-


166 N.K. Gill et al. / Biochemical Engineering Journal 39 (2008) 164176

Fig. 1. Mechanical drawing showing key design features and dimensions of asingle miniature bioreactor: (A) cross section through bioreactor; (B) plan viewof head plate. Further details given in Section 2.2.1.

cally driven from below, had a stirring range of 1002000 rpm.Each bioreactor was also equipped with four equally spacedremovable baffles of width 6 mm (dB/dT = 1/10). Aeration ofthe vessel was via a narrow sparger located directly beneaththe miniature turbine. The sparger was fitted with a 15mstainless steel sinter to create finer gas bubbles and promotemore efficient oxygen transfer. The air flow rate was manu-ally regulated by a standard laboratory rotameter in the range0200 mL min1 (Fisher Scientific, UK) and it was possible touse atmospheric air alone or oxygen-enriched air via a gas blend-ing system (HEL Ltd., UK). The gas flow rate to each reactorwas frequently checked in order to ensure constant gas flowrate.

As shown in Fig. 1(B), the headplate accommodated a total offour probes for continuous on-line monitoring of pH, dissolvedoxygen tension (DOT), optical density (OD) and temperature asdescribed in Section 2.2.2. As a result of the limited space avail-able on the headplate a narrow opening in the centre alloweda thermocouple for temperature sensing to be located withinthe thin walled hollow stirrer shaft. All liquid additions andsample removals were via a single multi-port on the headplatewhich had five openings closed by self-sealing septa (HEL Ltd.,UK). Liquid additions were made via sterile hypodermic nee-dles, securely mounted by luer lock fittings. The exhaust gas portwas fitted with a stainless steel, water cooled, condenser whichprevented liquid loss from the medium by evaporation. The inletand exhaust gas lines were filtered through 0.2m filters (FisherScientific, UK).

Fig. 2. Schematic diagram of in situ optical density probe: (A) dimensions ofprobe; (B) arrangement of light source and detectors. Further details given inSection 2.2.2.

2.2.2. On-line instrumentation of individual bioreactorsContinuous on-line monitoring of each bioreactor was

facilitated using a miniature pH probe (Hamilton BonaduzAG, Switzerland), polargraphic oxygen electrode (HamiltonBonaduz AG, Switzerland), a narrow K-type thermocouple(HEL Ltd., UK) and a novel optical probe (HEL Ltd., UK)for optical density measurement. All probes were steamsterilisable.

The optical probe, as shown in Fig. 2(A), had a PEEK bodyand was sized to fit each bioreactor headplate. The end of theprobe, Fig. 2(B), had a white light source and two detectors thatwere each located in sealed glass tubes. The light source (LS) andone of the detectors (D1) were positioned directly opposite eachother approximately 10 mm apart. D1 thus gave a direct mea-surement of the optical density of the broth. A second detector(D2) was placed at right angles to the direction of transmit-ted light approximately 5 mm away from the beam. The outputof this second detector measured the amount of light scatteredby the broth. The light source had a sufficiently narrow angleof emission such that any light reaching the right-angle detec-tor (D2) by direct transmission was considered negligible. Inthis work only the data from the optical density signal (D1) isreported. A calibration curve of on-line and off-line OD valueswas generated for each type of fermentation completed in theminiature bioreactors.

2.2.3. Design and control of parallel bioreactor systemsTo support the individual bioreactors a modular base unit

was designed which securely located four miniature bioreac-tors and their respective peristaltic pumps (RS ComponentsLtd., UK), rotameters, electrical disc heaters and magnetic driveassemblies. Cables from the various probes plugged into thebase unit which housed all the associated electronics creat-ing a compact unit with a footprint of 480 by 362 mm. Up



to four of these modular base units can be used in a singlesystem.

A custom piece of PC-based software was written whichallowed independent monitoring and control of up to 16 minia-ture bioreactors. This was based on an architecture originallyused for control of automated chemical reactors that has beenreported to be easily programmed and simple to use [17]. Exper-iments with multiple operating conditions and set points couldbe conducted by generating an experimental plan that dictatedwhich piece of equipment should be running and what it wasrequired to do. Each experimental plan can consist of severalsteps with each step being carried out automatically in sequence.The end to each step is specified by a terminating condition,which is any parameter defined by the user, e.g., time, tem-perature, pH, OD, DOT, etc. The software also enabled keyparameters such as pH, DOT, temperature and optical densityto be displayed in real time on interactive graphs for each biore-actor.

2.3. Characterisation of bioreactor oxygen transfer rates

The oxygen transfer capability of the miniature bioreac-tor was assessed using the dynamic gassing out technique[18]. Before each experiment 10 g L1 NaCl was dissolved infresh de-ionised water and the dissolved oxygen probe cali-brated between 0% and 100% saturation by sparging nitrogenand air, respectively. All experiments were carried out at aconstant temperature of 37 C at predetermined stirrer speeds(10002000 rpm) and aeration rates (12 vvm) using eitheratmospheric air or air enriched with oxygen. Assuming the liq-uid in the bioreactor was well mixed, the volumetric oxygenmass transfer coefficient, kLa, was determined from the mea-sured dissolved oxygentime profiles accounting for the proberesponse time according to Eq. (1) [19]:

Cp = 1tm p

[tm exp

(ttm

) m exp

(tp

)](1)

whereCp is the normalised dissolved oxygen concentration mea-sured by the probe at time t, tm equals 1/kLa and p is the proberesponse time (18 s at 37 C). All gassing out experiments wereperformed in triplicate with the maximum coefficient of variancefor kLa determination being 6.1%.

2.4. Parallel E. coli and B. subtilis fermentations

Two different microorganisms were used in this work. Inthe case of E. coli TOP10 pQR239 the growth media con-sisted of 10 g L1 each of tryptone, yeast extract, glycerol,NaCl (SigmaAldrich, Poole, UK) and 50 mg L1 ampicillin(SigmaAldrich, Poole, UK). Each miniature bioreactor wassterilised as a complete unit (121 C for 20 min) with all mediacomponents (apart from ampicillin) and 0.2 mL L1 of addedantifoam (polypropylene glycol 2000). After cooling the temper-ature was maintained at 37 C (0.2) via an electrical disc heaterpositioned beneath the glass vessel. Filter sterilised (0.2m)ampicillin was added immediately prior to inoculation. Each

bioreactor was inoculated with 2 mL (2%, v/v inoculum) of brothfrom a shake flask culture (100 mL in a 1 L shaken flask) grownfor 14 h on the same medium at 37 C and 200 rpm on a hor-izontal shaken platform (New Brunswick, USA). The pH wascontrolled at 7 (0.1) by the metered addition of 3 M NaOH and3 M H3PO4.

In the case of B. subtilis ATCC6633 chemically definedgrowth media was used. The media components wereprepared and sterilised (121 C for 20 min) in five sepa-rate groups [16]. One litre of biomedia consisted of: (a)100 mL of a 200 g L1 solution of d-glucose, (b) 895 mLof a solution of 11.2 g L1 (NH4)2SO4, 15.2 g L1 KH2PO4,12.1 g L1 K2HPO4, 4.0 g L1 Na2HPO4, and 1.1 g L1antifoam (polypropylene glycol 2000), (c) 2 mL of a 246 g L1solution of MgSO47H2O, (d) 1 mL of a 147 g L1 solution ofCaCl22H2O and (e) 2 mL of an acidified (pH 1) solution of tracemetals (40 g L1 FeSO47H2O, 5 g L1 MnSO4H2O, 2 g L1CoCl26H2O, 1 g L1 ZnSO47H2O, 1 g L1 MoO4Na22H2O,0.5 g L1 CuCl22H2O, 2 g L1 H3BO3). Each miniature biore-actor was sterilised as a complete unit (121 C for 20 min)containing group (b) media components. After cooling the tem-perature was maintained at 32 C and the remaining groups ofmedia components were added aseptically immediately prior toinoculation. Each bioreactor was inoculated with 10 mL (10%,v/v inoculum) of an actively growing culture (100 mL in a 1 Lshaken flask) with an optical density of approximately 2. Theshake flask culture was grown on the same medium at 32 Cand 300 rpm on a horizontal shaken platform. The pH was con-trolled at 6.8 ( 0.1) by the metered addition of 3 M NaOH and3 M H3PO4.

For both E. coli TOP10 and B. subtilis ATCC6633 fermen-tations, agitation rates in the miniature bioreactors were variedbetween 1000 and 2000 rpm and aeration rates varied between1 and 2 vvm (certain experiments with E. coli involved aerationwith oxygen-enriched air maintaining the DOT at a set pointof 30% or 50%). Fermentations were performed in parallel on afour pot system either under identical agitation and aeration con-ditions (to show reproducibility) or under different agitation andaeration conditions (to show parallel evaluation of fermentationconditions).

2.5. Analytical techniques

In addition to the on-line optical density measurements (Sec-tion 2.2.2), cell growth was also monitored by taking brothsamples of up to 1 mL at regular intervals and measuring theOD600 of appropriately diluted samples (1 in 2 to 1 in 20 dilu-tions were used) off-line (Ultraspec 4000 spectrophotometer,Pharmacia Biotech, USA). All biomass concentrations reportedhere are dry cell weight (DCW) concentrations and were fromappropriate experimentally determined calibration curves foreach microorganism and medium. The DCWs used to createthe calibration curve were determined in triplicate from knownvolumes of fermentation broth. After centrifugation at 1300 rpmfor 15 min (Eppendorf AG, Germany), cell pellets were washedonce with 10 g L1 NaCl solution and dried at 100 C for 24 hin pre-weighed and dried 2 mL Eppendorf tubes.



3. Results and discussion

3.1. Miniature bioreactor design and operation

The basis of the miniature bioreactor design was that it shouldbe as geometrically similar to conventional laboratory scalebioreactors as possible and be agitated by a standard minia-ture turbine impeller. This was dictated by the desire to obtainquantitative and scaleable data from each miniature bioreactorand efficient oxygen transfer during microbial fermentations. Inpractice these constraints limited the volume of each bioreactorto a minimum of 100 mL, to ensure that the miniature impellerwas completely submerged in liquid and maximum agitationrates of up to 2000 rpm could be achieved. In order to simplifythe assembly of each unit, and to minimise any risk of contam-ination, the need for a rotating mechanical seal on the impellerdrive shaft was overcome by opting for a magnetically drivenstirrer. This was only possible by compromising on the design ofthe impeller and the incorporation of a flat disk at the end of theimpeller drive shaft as shown in Fig. 1(A). This was necessaryto house sufficient magnets to enable agitation at rates of up to2000 rpm without decoupling the impeller from the magneticdrive. Even with these modifications visual observation showeduniform distribution of gas bubbles throughout the entire liq-uid volume of each vessel provided that the agitation rate was1000 rpm or above.

To facilitate parallel operation, four miniature bioreactorscould be assembled and sterilised simultaneously in a typicallaboratory autoclave. A hydrostatic pressure test indicted thateach vessel could typically withstand an applied pressure of8 bar before rupturing. An initial medium sterilisation and holdtest, in which DOT, pH and OD were monitored over 4 daysshowed no signs of contamination. This was confirmed by theabsence of any colonies from medium samples withdrawn peri-odically and grown on nutrient agar plates at 37 C for 24 h. Tofurther facilitate parallel and unattended operation of multipleunits each bioreactor was instrumented with pH, temperatureand dissolved oxygen probes as well as a novel on-line opticaldensity probe (as described in Section 2.2.2). The small area ofeach bioreactor head plate meant that it was necessary to sourcethe smallest available probes. In order to then have sufficientports for liquid additions, sampling and the possibility of con-tinuous bioreactor operation all liquid handling operations tookplace via a specially designed multi-port. Finally, it was pos-sible to independently monitor and control up to 16 miniaturebioreactors using custom-written, PC-based software.

3.2. Characterisation of bioreactor oxygen transfercapability

Since the majority of industrial fermentations use aerobicmicroorganisms the oxygen transfer capability of the miniaturebioreactor is of great interest. Parameters such as kLa are alsouseful for comparing different bioreactor designs and provide auseful criterion for scale-up. Fig. 3 shows a series of dynamicgassing out experiments performed at different stirrer speedsand aeration rates for aeration with atmospheric air. The rate of

Fig. 3. Influence of bioreactor operating conditions on oxygen uptake kineticsduring dynamic gassing out experiments. From left to right: () 2000 rpm,2 vvm; (- ) 2000 rpm, 1 vvm; (- - -) 1500 rpm, 1 vvm; () 1000 rpm,2 vvm; () 1000 rpm, 1.5 vvm; () 1000 rpm, 1 vvm. Experiments performedat 37 C in 10 g L1 NaCl solution as described in Section 2.3.

oxygen uptake is seen to increase with both increasing agitationand aeration rates. For each condition the corresponding kLavalues were calculated using Eq. (1) and are plotted in Fig. 4 asa function of stirrer speed. The maximum kLa was determinedas 0.11 s1.

The results in Fig. 4 show that kLa increases linearly withstirrer speed over the range investigated. Stirrer speed is alsoseen to have a more significant effect on kLa than increasesin aeration particularly at aeration rates above 1.5 vvm where

Fig. 4. Bioreactor oxygen mass transfer coefficient (kLa) as a function of stirrerspeed and aeration rate with atmospheric air: () 1 vvm; () 1.5 vvm; () 2 vvm.Values of kLa calculated from the dynamic gassing out data shown in Fig. 3 usingEq. (1). Solid lines fitted by linear regression.



the improvement in oxygen transfer is minimal. Classically theimpeller power input and aeration rate have been correlated inthe literature using the well known vant Riet correlation:

kLa = K(

PgV

)

S (2)

where Pg/V is the impeller gassed power per unit volume, S isthe superficial gas velocity, K is a constant and and are expo-nents in the range of 0.4 < < 1 and 0 < < 0.7, respectively [18].For kLa measurements in ionic solutions, the reported values ofthe constants K, and were 2.0 103, 0.7 and 0.2, respec-tively. Although the vant Riet correlation was determined formuch larger vessels than used here, fitted with standard Rushtonturbine impellers, the stronger dependency of kLa on Pg/V (andhence stirrer speed) than on uS shown in Fig. 4 initially suggeststhat results for the miniature bioreactor are consistent with theestablished theory. At present the Power number of the miniaturebioreactor is not known and so absolute values of Pg/V cannotbe estimated.

3.3. Parallel fermentations and in-situ monitoring

3.3.1. Reproducibility of parallel E. coli fermentationsA key requirement of any parallel bioreactor system is that

cultivations in separate bioreactors should be highly repro-

ducible if performed under identical operating conditions. Fig. 5shows biomass growth kinetics and dissolved oxygen tension(DOT) profiles for typical parallel E. coli fermentations in thefour-pot miniature bioreactor system. These were carried out ata fixed impeller speed and aeration rate of 1000 rpm and 1 vvm,respectively which gave a relatively low kLa value of 0.04 s1(as measured in 10 g L1 NaCl). Considering the first fermen-tation shown in Fig. 5 (bioreactor B1), the entire fermentationlasted a total of 540 min with the end of the exponential cellgrowth phase occurring around 300 min. This coincided withthe point at which oxygen mass transfer limitation occurred andthe measured DOT reached zero. The particular strain of E. coliused here is known to have a high specific oxygen demand so thefact that the measured DOT reached zero at this kLa value is notsurprising [15]. The period of exponential growth was followedby an almost linear increase in biomass concentration for a fur-ther 160 min at which point the culture entered stationary phaseand ceased to grow. The biomass growth kinetics and dissolvedoxygen profiles for the other three bioreactors were very similarapart from the DOT profile in bioreactor B4 which reached zeroDOT somewhat earlier.

Table 1 shows the key kinetic parameters derived from theindividual fermentation profiles shown in Fig. 5. The maxi-mum specific growth rates were very similar, giving an averagevalue of 0.68 0.01 h1 as were the final biomass concentra-tions achieved where the average value was 3.8 0.05 g L1.

Fig. 5. Parallel batch fermentation kinetics of E. coli TOP10 pQR239 grown under identical conditions: () off-line biomass concentration; (-) DOT. Experimentsperformed at 1000 rpm and 1 vvm as described in Section 2.4. B1B4 refer to bioreactors one to four, respectively.



Table 1Reproducibility of batch E. coli TOP10 pQR239 fermentation kinetics from four parallel fermentationsParameter B1 B2 B3 B4 BAV

max (h1) 0.68 0.66 0.68 0.69 0.68 0.01Xfinal (g L1) 3.7 3.8 3.8 3.8 3.8 0.05tDOT = 0 (min) 290 300 300 243 283 27.2OURmax (mmol L1 h1) 18.1 18.4 18.4 18.4 18.3 0.15Kinetic parameters derived from the fermentation profiles shown in Fig. 5. B1B4 refer to bioreactors one to four respectively, while BAV indicates the mean valuesof the kinetic parameters (error indicated represents one standard deviation).

The time at which the measured DOT reached zero was also com-parable in the majority of cases. Overall these results indicateexcellent reproducibility of the four-pot system.

At the point where the measured DOT first reaches zero, itcan be assumed that the maximum oxygen uptake rate (OURmax)and the maximum oxygen transfer rate (OTRmax) are equal. TheOURmax can thus be estimated from:

OURmax = XYX/O2

+ mO2X (3)

where is the specific growth rate at the time point when theDOT first reaches zero, X is the corresponding biomass concen-tration, YX/O2 is the yield of biomass on oxygen, which wastaken to be 1.92 g g1 [20] and m is the oxygen required for cellmaintenance which was taken to be 0.003 mol O2 g1 h1 [20].

Similarly, OTR can be estimated from:

OTR = kLa(C CL) (4)where C* is the saturated dissolved oxygen concentration, esti-mated to be 6.9 mg L1 [21] andCL is the actual concentration ofdissolved oxygen in the fermentation broth. At the point wherethe measured DOT first reaches zero, CL is also zero and so Eq.(4) reduces to:OTRmax = kLaC (5)

From Eq. (3) the average OURmax for the fermentationsshown in Fig. 5 was calculated to be 18.3 0.15 mmol L1 h1.In order to estimate the corresponding OTRmax value from Eq.(5), bioreactor kLa values were measured by further gassing outexperiments using spent biomedia containing the antifoam PPG

Fig. 6. Parallel batch fermentation kinetics of B. subtilis ATCC6633 grown under identical conditions: () off-line biomass concentration; (-) DOT. Experimentsperformed at 1500 rpm and 1 vvm as described in Section 2.4. B1B4 refer to bioreactors one to four, respectively.



used during fermentation experiments. The presence of PPGis known to significantly reduce kLa values [22] and the mea-sured value of 0.03 s1 was not surprisingly 25% lower thanthe value measured in electrolyte solution. Using this lower kLavalue the calculated value of OTRmax was thus estimated to be18.2 mmol L1 h1 which is in excellent agreement with thecalculated OURmax values.

3.3.2. Reproducibility of parallel B. subtilis fermentationsIn addition to showing the reproducibility of parallel fermen-

tations of E. coli, a facultative anaerobe, the flexibility of theminiature bioreactor system was shown using parallel fermenta-tions of B. subtilis, a strict aerobe. Fig. 6 shows typical biomassgrowth kinetics and DOT profiles for four parallel B. subtilisfermentations in the miniature bioreactors. These fermentationswere carried out at a fixed impeller speed and aeration rate of1500 rpm and 1 vvm, respectively corresponding to a kLa valueof 0.06 s1. Considering the first fermentation shown in Fig. 6(bioreactor B1) it can be seen that the entire fermentation lasted660 min with the exponential phase of cell growth lasting untilapproximately 540 min. At this point the measured DOT againreached zero at which point the culture rapidly entered stationaryphase and ceased to grow. As found with E. coli, very similarfermentation profiles for B. subtilis were obtained in all fourminiature bioreactors.

Table 2 shows the key kinetic parameters derived from theindividual fermentation profiles shown in Fig. 6. The maximumspecific growth rates were again very similar, giving an averagevalue of 0.45 0.01 h1 while the average final biomass concen-tration achieved was 9.0 0.06 g L1. The average time takenfor the DOT level to reduce to zero was 565 33.7 min. The levelof variation seen for the B. subtilis fermentations is comparableto that determined for the earlier E. coli work and again indicatesexcellent reproducibility of the four-pot bioreactor system.

In addition to using the gassing out method to determine kLavalues (as described in Section 3.2) in the case of a strict aerobelike B. subtilis an oxygen mass balance can also be used, whereunder steady state conditions oxygen uptake and consumptionrates must balance [21], thus from Eqs. (3) and (4):dCLdt

= kLa(C CL) OUR (6)

For this particular medium and strain of B. subtilis CLand YX/O2 have been reported to be 6.8 mg L1 and 1.6 g g1

Table 2Reproducibility of batch B. subtilis ATCC6633 fermentation kinetics from fourparallel fermentations

Parameter B1 B2 B3 B4 BAV

max (h1) 0.46 0.45 0.45 0.44 0.45 0.01Xfinal (g L1) 9.0 9.0 9.0 9.1 9.0 0.06tDOT = 0 (min) 529 572 550 608 565 33.7kLa (s1) 0.061 0.062 0.061 0.059 0.061 0.001Kinetic parameters derived from the fermentation profiles shown in Fig. 6.B1B4 refer to bioreactors one to four respectively, while BAV indicates themean values of the kinetic parameters (error indicated represents one standarddeviation).

respectively [16] and the contribution from cell maintenanceis negligible. Over a short time period dCL/dt is constant andthe kLa was calculated. This method gave a kLa value of0.062 0.001 s1 which is in excellent agreement with thatmeasured using the gassing out technique under the same oper-ating conditions (Table 2).

3.3.3. Correlation of on-line and off-line optical densitymeasurements

As described in Section 2.2.2, each miniature bioreactor wasequipped with a novel on-line OD probe (Fig. 2). This wasdesigned to facilitate continuous on-line monitoring of biomassgrowth kinetics which could be important given the small vol-ume of each bioreactor and the potential supervision of up to

Fig. 7. On-line measurement of optical density in E. coli TOP10 pQR239 fer-mentations. (A) Comparison of biomass growth kinetics from an individualfermentation: () off-line optical density; () on-line optical density. (B) Par-ity plot of on-line and off-line biomass concentration data from nine identicalfermentations. Experiments performed at 1000 rpm and 1 vvm as described inSection 2.4.



Fig. 8. Influence of agitation and aeration rates and oxygen enrichment on batch fermentation kinetics of E. coli TOP10 pQR239: () off-line biomass concentration;(-) DOT. Experimental conditions: (A) 2000 rpm, 1 vvm; (B) 1000 rpm, 2 vvm; (C) 2000 rpm, 1 vvm and DOT controlled at 30%; (D) 2000 rpm, 1 vvm and DOTcontrolled at 50%. Fermentations performed as described in Section 2.4.

16 bioreactor units in parallel by a single operator. As shownin Fig. 7(A) the growth profile generated during a single E. colifermentation by the on-line OD probe was virtually identical tothat from off-line OD measurements, apart from during the finalstages of the culture. The OD reading of the probe was initiallycalibrated using the sterilised culture medium. Calculated val-ues of the maximum specific growth rate from the on-line andoff-line data were 0.67 h1 and 0.68 h1, respectively indicatingthat the on-line data can be used for quantitative analysis of cellgrowth kinetics. Fig. 7(B) shows a parity plot of on-line and off-line biomass concentration data collected from nine identicalE. coli fermentations. There is seen to be excellent correlationbetween the two types of biomass concentration data indicatingthe robustness of the probe after repeated cycles of sterilisationand fermentations.

3.4. Inuence of agitation and aeration conditions on E.coli fermentations

The initial E. coli and B. subtilis fermentations reported inSection 3.3 were carried out at relatively low kLa values and sooxygen limitations were observed in both cases. To explore theoxygen uptake requirements of the E. coli strain further, a seriesof fermentations were carried at higher agitation rates of 1500

and 2000 rpm (aeration rate remained constant at 1 vvm) andhigher aeration rates of 1.5 and 2 vvm (agitation rate remainedconstant at 1000 rpm). Aeration was with either atmospheric airor oxygen-enriched air utilising the gas-blending capability ofthe miniature bioreactor system.

Fig. 8 shows examples of biomass concentration and DOTprofiles under the different agitation and aeration conditions. Thederived kinetic parameters from the entire series of experimentsare summarised in Table 3. For aeration with atmospheric airthere is a significant increase in both the maximum growth rate(0.750.94 h1) and final biomass concentration (5.15.6 g L1)in all cases when compared to the values obtained with thelowest agitation and aeration conditions used previously in Sec-tion 3.3.1 (0.68 h1 and 3.8 g L1). These increases in cellgrowth rate are broadly in line with the measured increasesin kLa values. The largest value of OURmax was calculated tobe 46.8 mmol L1 h1. In all cases, however, oxygen transferlimitations were observed to remain with the measured DOTreaching zero at some point during the course of the indi-vidual fermentations. In order to overcome this problem thebioreactors could be aerated with oxygen-enriched air suchthat the DOT was maintained at a constant set point value.For agitation at 2000 rpm and aeration at 1 vvm in this case,graphs C and D in Fig. 8 show that the DOT could be



Table 3Variation of bioreactor oxygen mass transfer coefficient and E. coli TOP10 pQR239 fermentation kinetics as a function of agitation and aeration conditions includingthe use of oxygen enrichment to control DOT levels

Parameter Agitation and aeration conditions

1000 rpm1 vmm

1500 rpm1 vmm

2000 rpm1 vmm (A)

1000 rpm1.5 vmm

1000 rpm2 vmm (B)

2000 rpm 1 vmm (C) 2000 rpm 1 vmm (D)

Gas blending DOT = 30% DOT = 50%kLa (s1) 0.04 0.06 0.08 0.06 0.06 max (h1) 0.68 0.83 0.94 0.75 0.79 0.86 0.93Xfinal (g L1) 3.8 5.1 5.3 5.6 5.6 7.6 8.1OURmax (mmol L1 h1) 18.4 35.7 46.8 26.3 27.5 ND ND(A)(D) correspond to the fermentation profiles shown in Fig. 8. Oxygen mass transfer coefficient and values derived from Fig. 3. N/D = not determined.

maintained and controlled at a minimum level of 30 or 50%respectively. In the absence of oxygen limitations the maxi-mum growth rate of the culture increased further to 0.93 h1and the highest biomass concentration obtained was 8.1 g L1.While gas blending can perhaps be avoided during the initialstages of fermentation process development it will be neces-sary to implement it during the later stages when high celldensity cultures must be attained in the miniature bioreac-tors.

3.5. Comparison with other miniature bioreactors

Several miniature bioreactor systems for rapid bioprocessdevelopment have been reported in recent years. A number ofdesigns have been investigated but the recent trend has beentoward mechanically stirred bioreactors as these are most widelyused in industry for development and large scale operation. Themajority of the bioreactors have been designed with paralleloperation in mind and there is a consensus that 10s of experi-ments be performed in parallel for the systems to be of value.In this respect the bioreactor block [23] currently gives thehighest degree of parallelisation, 48, but must be operated onthe deck of a dedicated laboratory automation platform. Thesmaller units (110 mL) tend to be single use disposable itemswhile the larger ones (100200 mL) like that described here canbe repeatedly steam sterilised.

In terms of agitation and aeration systems those bioreac-tors featuring mechanical agitation tend to have the highest kLavalues. Stirrer speeds necessarily increase dramatically as theworking volume of the bioreactors drop below about 10 mL. Inmany cases non-standard gas-sparging impellers [23] and gasblending [24,25, this work] have been implemented to main-tain sufficient oxygen transfer rates. The highest kLa reported todate for any of the miniature bioreactors has been for the biore-actor block, achieving 0.4 s1 [23]. kLa values of this orderare necessary to support the microbial cell densities that couldbe achieved in approximately half of the designs that are capa-ble of fed-batch or continuous operation. Betts et al. [25] havereported kLa values in the order of 0.1 s1 for a 10 mL stirredminiature bioreactor (the detailed design for this vessel has beenpreviously described by Lamping et al. [26]). The commerciallyavailable Cellstation (Fluorometrix, USA) [27,28] allows up to12 stirred bioreactors to be operated in a single run with work-

ing volumes of 35 mL. However, the oxygen transfer rates aresignificantly lower compared to other systems described in thissection (0.01 s1).

In addition to stirred vessels, small scale bubble columns havealso be designed in recent years, high-lighting the importanceand the need for a variety of bioreactor systems to address dif-ferent process applications. Currently a 200 mL bubble columnis available from Infors, Switzerland and is reported to achievekLa values of up to 0.16 s1 [29,30], with the capacity to operateup to 16 vessels in parallel. Doig et al. [16] have characteriseda system with up to 48 miniature bubble columns operating inparallel with working volumes of 2 mL, producing kLa values ofaround 0.06 s1. Although these systems lack the mixing capa-bility of a stirrer, sufficient aeration and some degree of mixing isachieved by direct sparging. These researchers have also demon-strated good correlation with a laboratory scale bubble column(100 mL) in terms of oxygen transfer and volumetric power con-sumption [31]. Cell cultivations were also scaled up based onconstant kLa to bench scale stirrer bioreactor, and the resultswere comparable.

The miniature bioreactor systems described herein arequickly replacing the traditionally popular shake flask for par-allel, high throughput operation, given the level of sophisticatedtechnology providing large quantities of data. However, someresearchers have attempted to enhance the design of the shakeflask to facilitate some degree of online monitoring, theseinclude the RAMOS (HITECZang GmbH, Germany) [32] andthe Fedbatch-pro (DasGip,Germany) [33] systems.

Table 4Comparison of E. coli TOP10 pQR239 fermentations carried out in miniatureand conventional laboratory bioreactors using constant kLa as a basis for scale-up(based on aeration with atmospheric air)Parameter Miniature bioreactor Laboratory bioreactor

Working volume (mL) 100 1500Stirrer speed (rpm) 2000 1000Aeration rate (vvm) 1.00 0.67kLa (s1) 0.08 0.08Inoculum (%working volume) 2.0 6.7max (h1) 0.94 0.97Xfinal (g L1) 5.3 5.5Kinetic parameters derived from Fig. 9.



In order to asses the reliability of miniature high throughputsystems as a tool for scale-down studies, it is critical to demon-strate their ability to mimic the conditions and productivity thatwould be expected at larger scales. A majority of the minia-ture systems that have been reported over the years have notdone this, however, some scale comparison studies have beencarried out comparing the growth and productivity of Bacil-lus subtillis RB50 and the productivity of riboflavin [34] in thebioreactor block [23] and a 7 L bioreactor. Betts et al. [25]have carried out more rigorous scale-up studies using a definedscale-up criterion based on constant power per unit volume giventhe geometric similarity between their vessel and a conventional

Fig. 9. Batch fermentation kinetics of E. coli TOP10 pQR239 fermentationscarried out at miniature (100 mL) and conventional laboratory bioreactor (1.5 L)scales: (A) off-line biomass concentration; (B) DOT. Experiments performed atmatched kLa values as described in Section 2.4. Laboratory bioreactor data fromDoig et al. [15].

bioreactor. Comparing the results of the miniature bioreactor andthe 7 L vessel concluded reasonable agreement in terms of max-imum specific growth rates for E. coli DH5 producing plasmidDNA.

Virtually all the designs feature the standard on-line probesexpected for effective bioreactor monitoring and control thougha number feature novel probes for on-line monitoring of cultureoptical density [23,27,28, this work]. The striking trend is theswitch from standard probe technologies to fluorescent/opticalprobes once the bioreactor volume drops below about 10 mL.All the designs for which parallel operation has been demon-strated feature dedicated PC-based software necessary for thesupervision of multiple units.

3.6. Scale-up from miniature to laboratory scalebioreactors

While the miniature stirred bioreactor described here iscapable of automated parallel operation, it is important toasses the scale-up potential of the results obtained. In thissection the relationship between miniature bioreactor resultsand those obtained using the same E. coli strain in a typicallaboratory scale 2 L bioreactor (1.5 L working volume ves-sel fitted with two top driven Rushton turbine impellers) willbe considered [15]. As a basis for predictive scale translationexperiments were initially performed at matched kLa valuesof 0.08 s1 based on direct measurements of oxygen transferrates at the two scales. Table 4 shows the corresponding agi-tation and aeration conditions. For E. coli cultures at the twoscales Fig. 9(A) shows that there was good agreement betweenbiomass growth kinetics while Fig. 9(B) shows the correspond-ing agreement between measured DOT profiles. Experimentsat both scales were performed without gas-blending so theexpected oxygen limitations toward the end of the exponentialgrowth phase are again seen. Given the 15-fold scale dif-ference there is excellent agreement between the calculatedmaximum specific growth rates and final biomass concen-trations achieved (Table 4). Similar agreement between thetwo scales was determined for matched fermentations at kLavalues over the range 0.060.11 s1 (data not shown). Thisgives confidence that results obtained from parallel experi-ments in the miniature bioreactors can be rapidly translated tothe conventional scales used for fermentation process develop-ment.

4. Conclusions

A parallel miniature bioreactor system (100 mL workingvolume) designed to be geometrically similar to conventionallaboratory scale stirred bioreactors has been constructed andevaluated in this study. The oxygen transfer characteristics ofthe miniature bioreactor were first evaluated in terms of theoxygen mass transfer coefficient, kLa, as a function of agi-tation and aeration rates. Values as high as 0.11 s1 couldbe achieved comparable to those found in typical laboratoryscale bioreactors. For identical fermentations performed in thefour-pot bioreactor system, excellent reproducibility between



parallel E. coli and B. subtilis fermentations was shown interms of the average calculated maximum specific growth rates(0.68 0.01 h1 and 0.45 0.01 h1 for E. coli and B. sub-tilis, respectively). Biomass growth rates and yields for E. colicould be improved, and oxygen transfer limitations overcome,by increases in agitation and aeration rates and by the implemen-tation of gas blending. Finally, kinetic parameters determinedfrom miniature bioreactor experiments were shown to be inaccord with those from a conventional 2 L stirred bioreac-tor (1.5 L working volume) when experiments were performedat matched kLa values. The ability to obtain quantitative andscaleable data from up to 16 miniature bioreactors in parallelmakes the system described here a useful tool for the paral-lel optimisation of microbial fermentation processes. Currentexperiments are aimed at a better characterisation of the engi-neering environment within the miniature bioreactors with aview to improved designs and predictive scale-up to larger biore-actors.

Acknowledgements

The authors would like to thank the UK Joint Infras-tructure Fund (JIF), the Science Research Investment Fund(SRIF) and the Gatsby Charitable Foundation for fundsto establish the UCL Centre for Micro Biochemical Engi-neering. Financial support from the UK Engineering andPhysical Sciences Research Council (EPSRC) and HELLtd., in the form of an Engineering Doctorate (EngD) stu-dentship for Naveraj Gill, is also acknowledged. The authorsare grateful to Martin Peacock for his contribution to thiswork.

References

[1] E.G. Hibbert, F. Baganz, H.C. Hailes, J.M. Ward, G.J. Lye, J.M. Woodley,P.A. Dalby, Directed evolution of biocatalytic processes, Biomol. Eng. 22(2005) 1119.

[2] S.D. Doig, F. Baganz, G.J. Lye, High throughput screening and processoptimisation, in: C. Ratledge, B. Kristiansen (Eds.), Basic Biotechnology,third ed., Cambridge University Press, UK, 2006.

[3] G.J. Lye, P. Ayazi-Shamlou, F. Baganz, P.A. Dalby, J.M. Wood-ley, Accelerated design of bioconversion processes using automatedmicroscale processing techniques, Trends Biotechnol. 21 (2003) 2937.

[4] S. Kumar, C. Wittman, E. Heinzle, Minibioreactors Biotechnol. Lett. 26(2004) 110.

[5] D. Weuster-Botz, Parallel reactor systems for bioprocess development,Adv. Biochem. Eng Biotechnol. 92 (2005) 125143.

[6] P. Fernandes, J.M.S. Cabral, Review: microlitre/millilitre shaken biore-actors in fermentative and biotransformation processes, Biocatal.Biotransform. 24 (2006) 237252.

[7] J.I. Betts, F. Baganz, Review: miniature bioreactors: current practices andfuture opportunities, Microb. Cell Factories 5 (2006) 21.

[8] W.A. Dutez, L. Ruedi, R. Hermann, K. OConner, J. Buchs, B. Witholt,Methods for intense aeration, growth, storeage and replication of bacte-rial strains in microtiter plates, Appl. Env. Microbiol. 66 (2000) 26412646.

[9] I. Elmahdi, F. Baganz, K. Dixon, T. Harrop, D. Sugden, G.J. Lye, pH controlin microwell fermentations of S. erythraea CA340: influence on biomassgrowth kinetics and erythromycin biosynthesis, Biochem. Eng. J. 16 (2003)299310.

[10] G.T. John, I. Klimant, C. Wittmann, E. Heinzle, Integrated optical sens-ing of dissolved oxygen in microtitre plates: a novel tool for microbialcultivation, Biotechnol. Bioeng. 81 (2003) 829836.

[11] K. Rege, M. Pepsin, B. Falcon, L. Steele, M. Heng, High-throughput pro-cess development for recombinant protein purification, Biotechnol. Bioeng.93 (2006) 618630.

[12] N.B. Jackson, J.M. Liddell, G.J. Lye, An automated microscale techniquefor the quantitative and parallel analysis of microfiltration operations, J.Membrane Sci. 276 (2006) 3141.

[13] C. Ferreira-Torres, M. Micheletti, G.J. Lye, Microscale process evaluationof recombinant biocatalyst libraries: application to Baeyer-Villiger mono-oxygenase catalysed lactone synthesis, Bioprocess Biosystems Eng. 28(2005) 8393.

[14] M. Micheletti, T. Barrett, S.D. Doig, F. Baganz, M.S. Levy, J.M.Woodley, G.J. Lye, Fluid mixing in shaken bioreactors: impli-cations for scale-up predictions from microlitre scale microbialand mammalian cell cultures, Chem. Eng. Sci. 61 (2006) 29392949.

[15] S.D. Doig, L.M. OSullivan, S. Patel, J.M. Ward, J.M. Woodley, Largescale production of cyclohexanone monooxygenase from Escherichia coliTOP10 pQR239, Enzyme Microb. Technol. 28 (2001) 265274.

[16] S.D. Doig, A. Diep, F. Baganz, Characterisation of a novel miniature bubblecolumn bioreactor for high throughput cell cultivation, Biochem. Eng. J.23 (2004) 97105.

[17] M.E. Van Loo, P.E. Lengowski, Automated workstations for parallel syn-thesis, Org. Process Res. Dev. 6 (2002) 833840.

[18] K. vant Riet, Review of measuring methods and results in non-viscousgas liquid mass transfer in stirred vessels, Ind. Eng. Chem. Process DesignDev. 18 (1979) 357364.

[19] I.J. Dunn, A.J. Einsele, Oxygen transfer coefficients by the dynamicmethod, J. Appl. Chem. Biotechnol. 25 (1975) 707720.

[20] B. Atkinson, F. Mavituna, Biochemical Engineering and BiotechnologyHandbook, second ed., Stockton press, 1999.

[21] Bailey, Ollis, Biochemical Engineering Fundamentals, McGraw-Hill Inter-national, Singapore, 1986.

[22] A. Morao, C.I. Maia, M.M.R. Fonseca, J.M.T. Vasconcelos, S.S. Alves,Effect of antifoam addition on gas-liquid mass transfer in stirred fermenters,Bioprocess Eng. 20 (1999) 165172.

[23] R. Puskeiler, K. Kaufmann, D. Weuster-Botz, Development, paralleli-sation, and automation of a gas inducing millilitre-scale bioreactor forhigh-throughput bioprocess design (HTBD), Biotechnol. Bioeng. 89 (2005)512523.

[24] R. Puskeiler, A. Kusterer, G.T. John, D. Weuster-Botz, Minia-ture bioreactors for automated high-throughput bioprocess design(HTBD): Reproducibility of parallel fed-batch cultivations withEscherichia coli, Biotechnol. Appl. Biochem. 42 (2005) 227235.

[25] J.I. Betts, S.D. Doig, F. Baganz, Charaterisation and application of a minia-ture 10 mL stirred-tank bioreactor, showing scale-down equivalence witha conventional 7 L reactor, Biotechnol. Prog. 22 (2006) 681688.

[26] S.R. Lamping, H. Zhang, B. Allen, P. Ayazi-Shamlou, Design of a prototypeminiature bioreactor for high throughput automated bioprocessing, Chem.Eng. Sci. 58 (2003) 747758.

[27] Y. Kostov, P. Harms, L. Randers-Eichhorn, G. Rao, Low-cost microbiore-actor for high-throughput bioprocessing, Biotechnol. Bioeng. 72 (2001)346352.

[28] P. Harms, Y. Kostov, J.A. French, M. Soliman, M. Anjanappa, A. Ram, G.Rao, Design and performance of a 24-station high throughput microbiore-actor, Biotechnol. Bioeng 93 (2006) 613.

[29] S. Dilsen, W. Paul, D. Herforth, A. Sandgathe, J. Altenbach-Rehm, R.Freudl, C. Wandrey, D. Weuster-Botz, Evaluation of parallel operatedsmall-scale bubble columns for microbial process development usingStaphylococcus carnosus, J. Biotechnol. 88 (2001) 7784.

[30] D. Weuster-Botz, J. Altenbach-Rehm, A. Hawrylenko, Process-engineering characterisation of small-scale bubble columns for microbialprocess development, Bioprocess Biosystems Eng. 24 (2001) 311.

[31] S.D. Doig, K. Ortiz-Ochoa, J.M. Ward, F. Baganz, Character-ization of oxygen transfer in miniature and lab-scale bubble



column bioreactors and comparison of microbial growth perfor-mance based on constant kLa, Biotechnol. Prog. 21 (2005) 11751182.

[32] T. Anderlei, J. Buchs, Device for sterile online measurement of the oxygentransfer rate in shaking flasks, Biochem. Eng. J. 7 (2001) 157162.

[33] D. Weuster-Botz, J. Altenbach-Rehm, M. Arnold, Parallel substrate feedingand pH control in shaking flasks, Biochem. Eng. J. 7 (2001) 163170.

[34] B. Knorr, H. Schlieker, H.P. Hohmann, D. Weuster-Botz, Scale-down andparallel operation of the riboflavin production process with Bacillus sub-tilis, Biochem. Eng. J. 33 (2007) 263274.

gill et al 2008a

Documents