the baffled microtiter plate: increased oxygen transfer and improved online monitoring in small...

ARTICLE

The Baffled Microtiter Plate: Increased OxygenTransfer and Improved Online Monitoring inSmall Scale Fermentations

Matthias Funke,1 Sylvia Diederichs,1 Frank Kensy,2 Carsten Muller,2 Jochen Buchs1

1AVT.Biochemical Engineering, RWTH Aachen University, Worringerweg 1,

52074 Aachen, Germany; telephone: þ49-241-80-23569;

fax: þ49-241-80-22570; e-mail: [email protected] GmbH, Forckenbeckstraße 6, 52074 Aachen, Germany

Received 11 March 2009; accepted 24 March 2009

Published online 1 April 2009 in Wiley InterScience (www.interscience.wiley.com). D
OI 10.1002/bit.22341
ABSTRACT: Most experiments in screening and processdevelopment are performed in shaken bioreactors. Today,microtiter plates are the preferred vessels for small-scalemicrobial cultivations in high throughput, even though theyhave never been optimized for this purpose. To interpret theexperimental results correctly and to obtain a base for ameaningful scale-up, sufficient oxygen supply to the cultureliquid is crucial. For shaken bioreactors this problem cangenerally be addressed by the introduction of baffles. There-fore, the focus of this study is to investigate how baffling andthe well geometry affect the maximum oxygen transfercapacity (OTRmax) in microtiter plates. On a 48-well platescale, 30 different cross-section geometries of a well werestudied. It could be shown that the introduction of bafflesinto the common circular cylinder of a microtiter plate welldoubles the maximum oxygen transfer capacity, resulting invalues above 100 mmol/L/h (kLa> 600 1/h). To also guar-antee a high volume for microbial cultivation, it is importantto maximize the filling volume, applicable during orbitalshaking. Additionally, the liquid height at the well bottomwas examined, which is a decisive parameter for online-monitoring systems such as the BioLector. This technologyperforms fiber-optical measurements through the well bot-tom, therefore requires a constant liquid height at all shakingfrequencies. Ultimately, a six-petal flower-shaped well geo-metry was shown to be the optimal solution taking intoaccount all aforementioned criteria. With its favorable cul-ture conditions and the possibility for unrestricted onlinemonitoring, this novel microtiter plate is an efficient tool togain meaningful results for interpreting and scaling-upexperiments in clone screening and bioprocess development.

Biotechnol. Bioeng. 2009;103: 1118–1128.

� 2009 Wiley Periodicals, Inc.

KEYWORDS: microtiter plate; shaken bioreactor; maxi-mum oxygen transfer capacity; baffle; screening; BioLector

Correspondence to: J. Buchs

Contract grant sponsor: German Federal Ministry of Education and Research (BMBF)

Contract grant number: 0313811A

1118 Biotechnology and Bioengineering, Vol. 103, No. 6, August 15, 2009

Introduction

The majority of industrial-scale bioprocesses take place instirred tank reactors. A pool of literature is available thatdescribes geometric properties of such stirred vessels, thedesign of the agitators or the influence of baffles. Just likestirred tanks play the major role in industrial-scale bio-processes, so do shaken bioreactors in research and processdevelopment. High throughput in user-friendly and inexpen-sive small scale-vessels is therefore necessary. Shake flasks—the main tool for microbial cultivation since the early daysof microbiology—and microtiter plates—nowadays thepredominant vessel for screening and high-throughputexperiments—fulfill these criteria and are used in almostany microbial laboratory today. Even though these devicesare commonly used, researchers often do not know or donot care about the culture conditions inside these devices.This lack of knowledge or consideration results in verydiverse and sometimes unsuited culture conditions inshaken bioreactors (Buchs, 2001). Thus, within the pastthree decades, intense efforts have been made to characterizefluid movement, gas transfer, energy input and mixing insmall-scale shaken bioreactors (reviewed in Buchs, 2001;Duetz, 2007; Fernandes and Cabral, 2006).

This characterization work has mainly been performed inshaken bioreactors with round cross-section (Duetz, 2007).These bioreactors often cannot provide a sufficient oxygensupply to the microorganisms. This parameter, however,is crucial not only for industrial production, but also formeaningful screening and process development (Freyeret al., 2004). To study the effect of organism properties,medium composition or cultivation strategy on growth andproduction, oxygen unlimited cultivations are absolutelynecessary. Otherwise wrong information about the variablesunder study might be obtained (McDaniel et al., 1965;Peter et al., 2004; Zimmermann et al., 2006). Ultimately, thismight lead to the selection of suboptimal strains, media or

� 2009 Wiley Periodicals, Inc.

culture conditions, which cannot be compensated in laterprocess development steps.

Besides increasing the diameter or frequency of shaking ordecreasing the filling volume, one possibility to achieve highmaximum oxygen transfer capacities (OTRmax) is to modifythe usual round geometry of the shaken vessel. As shown bymany groups, changing the round shape and/or introducingbaffles into shake flasks result in a significant increase inmaximum oxygen transfer capacity (Gaden, 1962; Guptaand Rao, 2003; McDaniel et al., 1965; Tunac, 1989). TheOTRmax can be increased up to 5- to 10-fold even at lowershaking frequencies. However, there are also disadvantagesby using baffled shake flasks, for example, that thereproducibility of microbial growth might be poor (Delgadoet al., 1989; McDaniel et al., 1965). Small differences indepth and positioning of the baffles lead to significantdifferences in oxygen supply, growth and product formationof parallel cultivations. Furthermore, in baffled shake flasksthe so called ‘‘out-of-phase’’ phenomenon might occur, inwhich the centrifugal force is not sufficient to move theliquid regularly (Buchs et al., 2001). Consequently, most ofthe liquid does not circulate in phase with the orbital motionof the shaker. This results in a lower mixing time, energyinput and OTRmax. One reason for this out-of-phasephenomenon can indeed be the excessive baffling by toomany or too large baffles or baffles with too sharp edges.Additionally, the liquid flow in baffled shake flasks is notwell-defined and cannot be modeled mechanistically.Moreover, the shaking frequency in this flasks has to bereduced to avoid splashing of the liquid. Were droplets toreach the plug of the flasks, gas transfer limitations orcontaminations might occur. Thus, in many cases the useof baffled shake flasks could not be recommended(Buchs, 2001; Henzler and Schedel, 1991).

Whereas some alternatives for the well bottom designhave been marketed, the only established alternative to around cross section geometry of a microtiter plate well is thesquare shape. This type of a modified well geometry has beeninvestigated mainly by Duetz and Witholt (2001, 2004) andDuetz et al. (2000). In these studies, the effects found forsquare wells are comparable to those in baffled shake flasks.Even though the OTRmax in square wells is more thandoubled compared to round wells, the aforementionedproblems described for shake flasks—splashing and out-of-phase phenomena—limit also the utilization of square deepwell plates as cultivation vessels.

However, as described above, the geometry of a shakingvessel significantly influences the OTRmax. Focused on thisimportant parameter of microbial cultivations, this articleprovides a systematic investigation of the influence of abaffled well geometry on the maximum oxygen transfercapacity in microtiter plates. Here, baffling is achieved byintroducing rectangular or rounded wall structures pointingtowards the well center, as it is established in stirred tankreactors. Moreover, edges or convex bulges pointingoutwards from the well center are also called ‘‘baffles’’ inthis work, since they redirect the flow of the rotating liquid

just like the classical baffles. As described above, theintroduction of baffles may cause several problems such asspilling and splashing during orbital shaking. Therefore,it was important to identify the geometry which avoidswetting of the well sealing, applying at the same time amaximum filling volume. Additionally, the amount ofliquid at the well bottom during orbital shaking has to bemaximized in order to facilitate correct optical analysis ofthe culture liquid through the transparent well bottom.The optimization of all three parameters is essential forthe application of the newly developed microtiter plate foronline-monitored small scale cultivations. Their suitabilityfor this purpose has been proven, applying the fiberoptical microfermentation system BioLector (Samorskiet al., 2005).

Materials and Methods

Design of the Baffled Well Geometries

The 30 different geometries of a microtiter plate (MTP) wellcross-section implemented in this work are shown inFigure 1. Three different sets of interrelated well geometriesrepresent the gradual transition from the most pronouncedbaffling (the square and the pentagon, respectively) to theleast pronounced geometry (the round) in three differentways, that is, by increasing the number of edges, roundingthe edges originating from a square, rounding the edgesoriginating from a pentagon (Fig. 1A). In addition to this,three groups of random well geometries were constructed byintroducing different types of baffles and/or rounding thecorners of an edged geometry (Fig. 1B). Only for the star-shaped geometries, the orbital shaking direction is restrictedto be counter-clockwise, since these geometries are notsymmetrical and the liquid is intended to be redirectedby the lower ‘‘uphill slope’’ of these baffles. To allow acomparison between all well designs, the cross section areaswere kept constant by constructing or rescaling them inAutoCAD (Release 14.01; Autodesc, Inc., San Rafael, CA) toa fixed value of 112 mm2. This value was adopted from theconventional round 48-well MTP. Out of a 20 mm highacrylic glass plate (polymethyl methacrylat, PMMA), MTPprototypes with outer dimensions of 128 mm� 85 mmwereproduced by laser cutting. To seal the bottom, a 2 mmPMMA plate was glued to the laser-cut plates.

Measurement System

All parameters were measured with an adapted BioLectormeasurement device (Fig. 2). The BioLector technique wasfirst described by Samorski et al. (2005) and is meanwhilecommercialized by m2p-labs (m2p-labs GmbH, Aachen,Germany). The utilized online-monitoring system consistsof a modified orbital shaker (based on Lab-Shaker LS-W,Kuhner AG, Basel, Switzerland), a x–y linear motion

Funke et al.: The Baffled Microtiter Plate 1119

Biotechnology and Bioengineering

Figure 1. Design principles of MTP well geometries: (A) Sets of interrelated well geometries (increasing number of edges, rounding of edges originating from square or

pentagon). Grey circles indicate the rounding of edges. The diameter of the rounded edge, the side length of the polygon or the side length of the underlying square or pentagon is

given in mm. These dimensions were selected, to construct geometries with an equivalent cross section area of 112 mm2. B: Random well geometries originate from square or circle

by introducing different styles of baffles. The dimensions of the baffles and the dimensions of the underlying polygon are given in mm. After construction with the given dimensions,

all geometries were rescaled in AutoCAD (Release 14.01; Autodesc, Inc.) to an equivalent cross section area of 112 mm2.

module (Bosch Rexroth AG, Lohr am Main, Germany),a custom-made filter fluorescence spectrometer (PreSensGmbH, Regensburg, Germany) and a computer. Theimplemented orbital shaker was modified to realize ashaking diameter of 3 mm and shaking frequencies up to1,000 rpm. An interruption of oxygen supply and mixingduring the measurement procedures is avoided due tocontinuous shaking of theMTP. To also reduce evaporation,a hood is placed above the MTP on the shaker tray,flushed continuously with humidified air (not shown inFig. 2).


For determining the biomass concentration, the back-scattered light in an angle of 1808 is measured (Zavrel et al.,2009). The light source of the custom made spectrometeris a light emitting diode. To exclude most of disturbingfluorescence signals of the biological sample, an optical cut-off filter was used which allows only light with a wavelengthabove 515 nm to enter the photomultiplier tube (PMT) ofthe spectrometer. Its output signal is amplified by a lock-in-amplifier (SR 830; Scientific Instruments GmbH, Gilching,Deutschland). A LabVIEW program (National Instruments,Austin, TX) acquires and finally saves the data from the

Figure 2. Modified BioLector measurement system. The optical fiber is fixed

on the x–y linear motion module at an angle of approximately 308, to avoid the

interference by light directly reflected from the well bottom.

lock-in-amplifier. Moreover, this program controls alsothe linear motion module in order to achieve a recurrentmeasurement in each individual well.

Characterization of the Maximum Oxygen TransferCapacity With the Sulfite System

To characterize the maximum oxygen transfer capacity(OTRmax) of the different well geometries, the sulfiteoxidation was applied as described in detail from Hermannet al. (2003, 2001) and Kensy et al. (2005). In contrast to theabove-mentioned literature, the system was slightly mod-ified by replacing the pH-indicator. The test solution iscomposed of 0.012 M Na2HPO4/NaH2PO4 phosphatepuffer, 0.5 M Na2SO3 (all: >98% purity, Roth, Karlsruhe,Germany), 10�7 M CoSO4 and 2� 10�8 M hydroxypyr-enetrisulfonic acid (HPTS) as fluorescent pH-indicator(both: Fluka, Buchs, Switzerland). In the original recipe bro-mthymol blue was used as a color-changing pH-indicator.The pH value is adjusted to pH 8 with sulfurous acid prior tothe experiments.

The time is measured, until the oxidation of the sulfite tosulfate is completed and therefore, a constant low pH valueis reached. The change of the pH value is quantitativelymeasured by the BioLector system. The fluorescence ofHPTS excited at 420 nm and measured above 515 nmdecreases with decreasing pH. Using the time until sulfiteoxidation is completed (tOx), the initial concentration ofsulfite (csulfite) and the stoichiometric coefficient for oxygenvO2

¼ 0:5, the oxygen transfer rate (OTR) to the solutioncan be calculated:

OTR ¼ kLaðcO2 � cLÞ ¼csulfitenO2

tOx(1)

With an oxygen concentration in the bulk liquid (cL) nearzero, the concentration difference reaches its maximum

and a maximum oxygen transfer capacity (OTRmax) can bedefined:

OTRmax ¼ kLac�O2

¼ kLaLO2pG (2)

The oxygen concentration at the bulk liquid interface ðc�O2Þ

can be defined as the product of oxygen solubility in thetest solution (LO2

¼ 8:35� 10�4 mol=L=bar at 258C) andoxygen partial pressure in the gas phase (pG¼ 0.2095 bar).Thus, the specific mass transfer coefficient (kLa) can becalculated from the OTRmax according to Equation (2).

Since cL is not necessarily zero during the sulfiteoxidation, it has to be calculated in order to obtain thevalue OTRmax, by transforming the mass balance for oxygento cL¼OTR/k1. With the first-order reaction constant ofthe sulfite oxidation k1¼ 2,385 1/h (Hermann et al., 2003),the OTRmax can be calculated from the measured OTRaccording to the aforementioned literature:

OTRmax ¼ kLac�O2

¼ OTRc�O2

c�O2� ðOTR=k1Þ

(3)

Liquid Height at the Well Bottom

For measuring the liquid height at the well center, thechange in the fluorescence signal of a 1.25� 10�6 Mfluorescein (sodium salt, Fluka) solution in 0.2 M sodiumphosphate buffer (pH 7) (Roth) with increasing shakingfrequency was measured. To measure the vertical liquidheight, the optical fiber of the BioLector has to be arrangedperpendicularly underneath the well (not at 308 as shown inFig. 1). The calibration was conducted by filling differentvolumes of the fluorescein solution inside the wells andmeasuring the fluorescence signal at a shaking frequency of200 rpm, at which no liquid movement occurs. Since thebase area for each geometry is constant, the liquid height canbe calculated from the filling volume. Thus, a geometry-independent correlation of fluorescence signal and liquidheight was achieved.

Maximum Possible Filling Volume

The maximum liquid volume, which could be filled in thewell until the rotating bulk liquid or drops reach the sealingof the shaken MTP, was examined. The maximal liquidvolume was determined by fixing a sheet of blotting paperon top of a MTP prototype. At a shaking frequency of1,000 rpm, the filling volume was increased in incrementsof 50 mL until the blotting paper became moistened.

Cultivation Experiments

The cultivation experiments were performed with Escher-ichia coli K12 (DSM 498) in buffered Terrific Broth (TB)



medium. The medium consists of: glycerol (5 g/L), tryptone(12 g/L), yeast extract (24 g/L), K2HPO4 (12.5 g/L), KH2PO4

(2.3 g/L) (all: Roth). The plates were sealed with a breathablesealing tape (AB-0718; ABgene, Epsom, United Kingdom).The bacteria were cultivated in the MTP prototypes and inround 48-well MTP (Art. No. MTP-R48-BOH) as well as inthe Flowerplate (Art. No. MTP-48-BOH) (both m2p-labsGmbH). All cultivations were carried out with the BioLectormeasurement system described above in a temperaturecontrolled room at 378C.

As a biomass indicator, the scattered light signal ismonitored by the BioLector system. Calibration of thissignal can be obtained by referencing against conventionalOD600 nm measurements in a photospectrometer. Since thedetermination of absolute biomass concentrations is notthe focus of this study, this calibration procedure has notbeen conducted and the scattered light values are given inarbitrary units. Moreover, the scattered light signal dependsalso on the liquid height at the well bottom. The light beamreaches the sample through the well bottom. If the liquidheight at this measurement point decreases, the measure-ment signal is reduced (see also discussion of Figs. 6 and 8).To quantify in detail the actual biomass concentrationindependently from the culture liquid height, the scatteredlight signal would have to be elaborately calibrated for allwell geometries. But to avoid this, the scattered light valuesplotted in Figure 5 are normalized between 0% and 100%and in Figure 8 to a joint start value of 100%.

Figure 3. Maximum oxygen transfer capacity (OTRmax) and specific mass

transfer coefficient (kLa) obtained from the sulfite system at increasing shaking

frequencies and different filling volumes. Set of interrelated geometries with increas-

ing number of edges. Measurements performed with a modified BioLector measure-

ment system: shaking diameter 3 mm, temperature 258C. Bars represent the standard

deviation around mean of at least three independent experiments.

Results and Discussion

Maximum Oxygen Transfer Capacity andWell Geometry

All 30 microtiter plate (MTP) well geometries were investi-gated regarding their maximum oxygen transfer capacity(OTRmax) at filling volumes from 200 to 600 mL and shakingfrequencies from 500 to 1,000 rpm. It has already beenshown, that a specific critical shaking frequency has to beexceeded so that the centrifugal force becomes larger thanthe interfacial surface tension (Hermann et al., 2003; Kensyet al., 2005). Therefore, shaking frequencies lower than500 rpm were not applied, since below this limit there is nosignificant fluid movement and thus no significant increasein oxygen transfer (data not shown).

The change of the OTRmax by varying the filling volume isshown in Figure 3 for the set of interrelated well geometrieswith increasing number of edges. Independent of thewell geometry, higher mass transfer is obtained at lowerfilling volumes. However, the principal correlation betweenOTRmax and vessel geometry does not change significantly.It is obvious, that a change in the filling volume influencesthe absolute height of the OTRmax value, but not the relationbetween the well geometries. Therefore, only the data for500 mL are shown for all other geometries (Fig. 4).


Figure 4. Maximum oxygen transfer capacity (OTRmax) and specific mass transfer coefficient (kLa) obtained from the sulfite system at increasing shaking frequencies and

filling volume of 500 mL. A: Set of interrelated well geometries with rounding of edges (square). B: Set of interrelated well geometries with rounding of edges (pentagon). C: Random

well geometries. Measurements performed with a modified BioLector measurement system: filling volume 500 mL, shaking diameter 3 mm, temperature 258C. Bars represent the

standard deviation around mean of at least three independent experiments.

The well geometry clearly determines the absolute heightof the OTRmax. To generalize, it is possible to state thatthe introduction of baffles leads to higher oxygen transfercompared to the round well geometry. This rule is absolutelyvalid for lower shaking frequencies (�800 rpm) at any fillingvolume. For all sets of interrelated well geometries presentedin Figures 3 and 4A,B a steady increase in the OTR canbe observed when the geometry changes from the lesspronounced shape (round) to the most baffled shape(square and pentagon, respectively). Also for the randomwell geometries (Fig. 4) an increase up to a doubled value ofOTRmax, compared to the round well geometry, is visible atmid-shaking frequencies. With increasing shaking frequency(>800 rpm), some of the tested baffled geometries showexceptions from this rule, that is, a steady increasein maximum oxygen transfer capacity with baffling is nolonger observed. The most baffled geometry of the sets ofinterrelated well geometries does not show the maximumoxygen transfer value anymore. For the set with roundededges in pentagon, the OTRmax only slightly peaks at thepentagon with edge diameter of 2 mm (Fig. 4B). Such a peakis clearly detectable at the pentagon within the geometryset with increasing number of edges (Fig. 3) and is mostpronounced at the set with rounding of edges in square,

namely at the square with edge diameter of 4 mm (Fig. 4A).The introduction of more pronounced baffles leads for allsets of interrelated well geometries to a decrease in theOTRmax values when higher shaking frequencies are applied.With some of the randomly designed baffles (Fig. 4C),that is, the edged baffles and the star geometries, even alower oxygen transfer compared to the round geometryis observed at high shaking frequencies. The decreasingOTRmax values with increased baffling are most probablycaused by of out-of-phase phenomena. As described byBuchs et al. (2001), a large number and/or size of baffles isone factor that might lead to out-of-phase liquid behavior.This behavior is characterized by the breakdown of therotating liquid bulk. Consequently, most of the liquidremains at the base of the shaken vessel. Only a minor partof the liquid is still rotating, which leads to reducedmass transfer. From the presented results it could at leastqualitatively be stated that the introduction of bafflessignificantly hinders the rotation of the culture liquid. This,in turn, adversely affects the mass transfer. The distinctive-ness of this effect depends on the size and the shape ofthe baffles. More detailed research on the influence of bafflesize, number and shape on this out-of-phase phenomenonshould be performed.



The results obtained from the chemical model systemhave been validated for all 30 geometries using cultivationsof E. coli K12 in complex TB-medium (Fig. 5). Thisbiological test system was chosen, since it requires a highamount of oxygen for unlimited growth (above 110 mmol/L/h, data not shown). The scattered light signal, representingthe biomass growth, is normalized between 0% and 100%and plotted over cultivation time. Figure 5 shows thatintroducing baffles generally leads to faster biomass growth,whereby for the sets of interrelated well geometries (Fig. 5A–C) the growth curves are arranged strictly in the order of thebaffling intensity. Their slopes, representing their particularbiomass growth rate, increases with more pronouncedbaffling since these baffles cause a higher OTR into theculture liquid. Also in case of the randomwell designs shownin Figure 5 D–F, the order of the growth curves reflectsthe height of the OTRmax values measured by the sulfitesystem. As previously discussed, the well geometry

Figure 5. Cultivation of Escherichia coli K12 in TB-medium in microtiter plate proto

measured values of the scattered light signal are normalized and plotted between 0% and

volume 500 mL, shaking frequency 700 rpm, shaking diameter 3 mm, temperature 378C.


absolutely determines the OTRmax values and thus thesupply of the substrate oxygen to the microorganisms.Consequently, a strong correlation between the well geo-metry (respectively the OTRmax) and the biomass growthcan be observed, even though the chemical and thebiological systems are completely different.

Influence of Well Geometry on the Liquid Heightat the Well Bottom

In the BioLector system, the MTP is continuously shaken.Since the light beam for the optical measurement is directedthrough the well bottom, its signal intensity depends on theamount of liquid at the well bottom. To determine thisinfluence, the liquid height at the center of each wellgeometry was measured at shaking frequencies of up to1,000 rpm (Fig. 6).

types. A–C: Sets of interrelated well geometries. D–F: Random well geometries. The

100%. Cultivations monitored with a modified BioLector measurement system: filling

The liquid height in each well stays constant until theshaking frequency reaches the critical value of 400 rpm. Upto this value, the centrifugal force is not sufficient to movethe liquid. Above this value the liquid height in the unbaffledand slightly baffled wells drastically drops (Fig. 6A–C). Fromthe sets of interrelated well geometries only the square and

Figure 6. Filling height at the center of the wells with increasing shaking

frequency. A–C: Sets of interrelated well geometries. D: Random well designs.

Measurements were performed with a modified BioLector measurement system:

shaking diameter 3 mm, filling volume 500 mL, well height 20 mm.

the square with slightly rounded edges (diameter 2 mm)show relatively stable liquid coverage of the well bottom upto shaking frequencies of 1,000 rpm. With the exception ofthe 4-edged flower with edge diameter of 8 mm, all randomwell designs also show a relatively stable liquid height(Fig. 6D). In the well with round geometry or with slightlypronounced baffles, the liquid rotates as a homogeneousbulk around the wall of a roundMTP well. At higher shakingfrequency the culture liquid drains off the bottom andclimbs up the well wall. Only if the introduced baffles arehighly pronounced, the liquid rotation is disturbedeffectively. As a result, a higher amount of liquid stays atthe well bottom even at higher shaking frequencies. Thissufficient bottom coverage with culture liquid is crucial toguarantee a stable measurement signal with the BioLectorsystem at any shaking frequency.

Influence of Well Geometry on the MaximumPossible Filling Volume

To achieve OTRmax values sufficient for highly oxygen-consuming cultures, the shaking frequency has to be high.At these frequencies, the rotating liquid or detached dropsmight reach the plate sealing and lead to moistening andclogging. The maximum liquid volume that does not reachthe well sealing was determined at a shaking frequency of1,000 rpm (Fig. 7).

Figure 7. Maximum filling volume of the different investigated well geometries

at a shaking frequency of 1,000 rpm. Measurements performed on the orbital shaker of

a modified BioLector measurement system: shaking diameter 3 mm, well height 20 mm.



The maximum filling volume of 1,300 mL could only beapplied in well geometries closest to the round shape. Thisvolume represents 58% of the total well volume (2,250 mL)of the MTP prototypes. At higher filling volumes, therotating liquid creeps up to the top of the well. Theintroduction of highly pronounced baffles significantlyreduces the applicable filling volume to 750–800 mL (33% ofwell volume). In the sharp edges of baffled wells the liquidobviously rises above the bulk liquid due to capillary forces.Additionally, increased turbulence induced by the bafflesmight lead to a detachment of drops and spraying of theculture liquid. As a result, all three sets of interrelated wellgeometries show a steady decrease in the maximum fillingvolume with more pronounced baffling. This results insmaller culture volumes and thus smaller sample volumesfor offline analyses. With relatively high maximum fillingvolumes of 1,100–1,200 mL, the flower and the 6-edged stargeometries represent an exception. The shape of these bafflesis obviously advantageous. In particular the slight uphillslope of these baffles and the absence of sharp edges whichwould cause the capillary effect, are most probably thereason for the high possible filling volume.

Ranking for the Best Well Geometry

All 30 selected MTP well geometries have been investigatedregarding maximum oxygen transfer capacity, liquid heightat the well center and maximum possible filling volumeduring orbital shaking. Only by optimizing all these threeparameters together can the best MTP be designed foreffectively cultivating and monitoring microorganisms.

Figure 8. Cultivation of Escherichia coli K12 in TB-medium in (A) 48-well microtiter pl

m2p-labs GmbH). The scattered light signal of all curves is normalized to a joint initial value o

frequency 1,000 rpm, shaking diameter 3 mm, temperature 378C, filling volume as indicate


Altering the cross-section geometry by introducing bafflesleads to higher maximum oxygen transfer capacities. Thus,the highest values are obtained using the square and thepentagon geometry and especially their derivatives withslightly rounded edges (diameter 2 and 4 mm, respectively).From the random well designs, the rounded baffles as well asthe flower-shaped geometries are best. However, if bafflingbecomes too pronounced, the advantage of higher oxygentransfer vanishes. The square, the edged baffles and the stargeometries show worse maximum oxygen transfer capacitiesat least at higher shaking frequencies. In contrast to theaforementioned ranking according to the OTRmax, highlypronounced baffling is preferred to achieve a sufficientliquid coverage of the well bottom. From the sets of inter-related well geometries only the strongly baffled square andthe square with a rounded edge of 2 mm in diameter canguarantee a stable optical measurement through the wellbottom. Additionally, most of the random well designs showthis advantage, too. However, in many of these geometriesonly a relatively low filling volume of 800 mL can be used. Inpractice this will result in significantly reduced culture andsample volumes, which limits offline analyses. By taking intoaccount all three parameters, only the flower geometriesresulted in enhanced growth combined with better onlinemonitoring. Ultimately, the 6-petal flower shape with anedge diameter of 5mmwas chosen as best and is marketed ina commercial product (Flowerplate; m2p-labs GmbH).

Figure 8 depicts the direct comparison of microbialcultivations in the Flowerplate and a conventional, round-well MTP. The aforementioned parameters OTRmax andmaximum possible filling volume are discussed by means ofE. coli K12 cultivations in TB-medium at filling volumesof up to 1,200 mL. In Figure 8, the scattered light signal

ate with round wells and (B) wells with 6-edged flower (Ø5) shape (Flowerplate) (both

f 100%. Cultivations monitored with a modified BioLector measurement system: shaking

d, well height 40 mm.

is plotted as an indicator for biomass concentration. Tocorrect small differences between the individual wells, allgrowth curves are normalized to a joint initial value of100%, by dividing every value of the curve by the initialvalue of the specific curve.

The biggest advantage of the flower-shaped well geometryis the possibility to perform oxygen-unlimited, at least,less oxygen-limited microbial cultivations. The scatteredlight curves of the Flowerplate cultivations show a nearlyoxygen-unlimited exponential growth at filling volumes upto 600 mL. Compared to this, the cultures in the round wellsare all oxygen-limited, showing a linear growth with lowergrowth rates. This oxygen limitation leads not only tosignificantly reduced microbial growth rates but also tolonger cultivation times. Moreover, it is known that thislimitation results in changes of the metabolism of themicroorganism and the formation of significant amounts ofundesired by-products. The comparison of the 800 mLcultivation in the round well and the 1,200 mL cultivation inthe Flowerplate show nearly similar growth kinetics andfinal biomass concentrations. It can be concluded, that theculture volume at a shaking frequency of 1,000 rpm in theFlowerplate can be approximately 400 mL higher, until anoxygen limitation level comparable to that of the round wellis reached. Consequently, cultivation in the flower-shapedwell provides bigger culture and sample volumes. Thus, it isalso less affected by culture liquid evaporation.

In Figure 8B, an increase in the final biomass concentra-tion with decreasing filling volume can be observed. This canbe explained by the higher biomass yield per amount ofconsumed substrate. This higher yield in cultivations withlow filling volume is caused by higher OTRs, which supporta highly productive aerobic growth. The expected differencein the values of the final biomass concentrations is obviouslymeasured correctly in the baffled MTP. In contrast to thebiomass monitoring in the Flowerplate, in the round well arelatively low scattered light signal is measured up to fillingvolumes of 600 mL (Fig. 8A). The signals obtained for theselow filling volumes are smaller than the signal for the 800 mLcultivation. As described above, this is a contradiction to thebiomass values theoretically expected. The observed reducedscattered light signal at low filling volumes does not reflectthe correct biomass concentrations in the conventionalMTP. However, the measurement signal is caused by areduced liquid height at the well center (see also Fig. 6).In the MTP with round wells, the culture liquid creeps upthe wall of the well and drains off the well bottom. Thusthe liquid height at the well center is drastically reduced.In particular, at lower filling volumes, the liquid heightbecomes too small and the measurement signal finallydecreases. This incorrect low biomass signal at low fillingvolumes may lead to misinterpretation of experimentalresults. To compensate this, an elaborate biomass calibra-tion would be necessary in the round well MTP for everyfilling volume. To summarize this, a second big advantageof the Flowerplate is that it is much better suited formeaningful online monitoring. As shown before during

fluorescence measurement (Fig. 6), this fact is caused by arelatively high and stable liquid height at the well center evenat small filling volumes.

Conclusion

In this study the well geometry of a microtiter plate (MTP)was optimized to provide sufficient oxygen supply to aerobiccultivations and guarantee a meaningful online monitoringof microbial growth. By introducing baffles, the maximumoxygen transfer capacity could be doubled to more than100 mmol/L/h (kLa> 600 1/h) compared to round 48-wellMTP. The liquid height at the well center, which isimportant for online-monitoring systems such as theBioLector and the maximum possible filling volume ofthe baffled geometries were also investigated. A compromisewas found between: (I) Strong baffling—required for astable liquid height at the well center during orbital shaking;(II) Moderate baffling—required for high maximumoxygen transfer capacity at any filling volume and anyshaking frequency; (III) Low baffling—required to applyhigh filling volumes and avoid spilling and splashing ofculture liquid. By optimizing all three parameters, the6-petaled flower with 5 mm edge diameter was found to bethe optimum. The advantages of this well geometry couldbe proven for online-monitored cultivations of E. coli inTB-medium.

In small-scale experiments, medium and fermentationparameters have to mimic large-scale production processes.This will increase the possibility that the improvementsfound in screening and process development can also beattained in large-scale production (Parekh et al., 2000).One of the most important parameters is the OTR. Inconventional MTPs, OTRmax values reported so far do notgenerally exceed 50 mmol/L/h (Duetz, 2007). Values of theOTR in stirred tank reactors can reach up to 500 mmol/L/h,but a mean for many standard batch fermentations isapproximately 100 mmol/L/h (van’t Riet, 1983, 1979;Yawalkar et al., 2002). Hence, there is a substantialdifference between the OTRmax in small-scale and indus-trial-scale. However, by considering the OTRmax values,attained in the baffledMTP, the gap narrows and both scalesapproximate each other at least in terms of oxygen transfer.To summarize, the baffled Flowerplate represents a uniqueand efficient tool for clone screening and bioprocessdevelopment. The oxygen supply to the culture liquid,which used to be the most crucial process parameter foraerobic microbial cultivations, now has a less decisiveimpact on small-scale cultivations. The performance ofmeaningful experiments and the generation of morerelevant data for scale up to stirred tank reactors are nowpossible.

Besides this improvement in maximum oxygen transfercapacity, the filling volume and the liquid coverage of thewell bottom have been maximized using the new flower-shaped well design. Hence, more precise online monitoringis possible at high shaking frequencies. Moreover, relatively



high volumes of culture liquid can be applied that allowshigh sample volumes for offline analyses and reducesnegative influence of culture liquid evaporation.

The work was supported by the German Federal Ministry of Educa-

tion and Research (BMBF), under grant No. 0313811A. The authors

would like to thank Robert Huber for the intense and fruitful

discussions and Mary-Joan Blumich for her kind and patient correc-

tion of the manuscript.

References

Buchs J. 2001. Introduction to advantages and problems of shaken cultures.

Biochem Eng J 7(2):91–98.

Buchs J, Lotter S, Milbradt C. 2001. Out-of-phase operating conditions, a

hitherto unknown phenomenon in shaking bioreactors. Biochem Eng J

7(2):135–141.

Delgado G, Topete M, Galindo E. 1989. Interaction of cultural conditions

and end-product distribution in Bacillus subtilis grown in shake flasks.

Appl Microbiol Biotechnol 31:288–292.

Duetz WA. 2007. Microtiter plates as mini-bioreactors: Miniaturization of

fermentation methods. Trends Microbiol 15(10):469–475.

Duetz WA, Witholt B. 2001. Effectiveness of orbital shaking for the aeration

of suspended bacterial cultures in square-deepwell microtiter plates.

Biochem Eng J 7(2):113–115.

Duetz WA, Witholt B. 2004. Oxygen transfer by orbital shaking of square

vessels and deepwell microtiter plates of various dimensions. Biochem

Eng J 17:181–185.

Duetz WA, Ruedi L, Hermann R, O’Connor K, Buchs J, Witholt B. 2000.

Methods for intense aeration, growth, storage, and replication of

bacterial strains in microtiter plates. Appl Environ Microbiol 66(6):

2641–2646.

Fernandes P, Cabral JMS. 2006. Microlitre/millilitre shaken bioreactors in

fermentative and biotransformation processes—A review. Biocatalysis

Biotransformation 24(4):237–252.

Freyer SA, Konig M, Kunkel A. 2004. Validating shaking flasks as repre-

sentative screening systems. Biochem Eng J 17(3):169–173.

Gaden EL. 1962. Improved shaken flask performance. Biotechnol Bioeng

4(1):99–103.


Gupta A, Rao G. 2003. A study of oxygen transfer in shake flasks using a

non-invasive oxygen sensor. Biotechnol Bioeng 84(3):351–358.

Henzler H-J, Schedel M. 1991. Suitability of shaking flask for oxygen supply

to microbial cultures. Bioprocess Eng 7:123–131.

Hermann R, Walther N, Maier U, Buchs J. 2001. Optical method for the

determination of the oxygen-transfer capacity of small bioreactors

based on sulfite oxidation. Biotechnol Bioeng 74(5):355–363.

Hermann R, Lehmann M, Buchs J. 2003. Characterization of gas-liquid

mass transfer phenomena inmicrotiter plates. Biotechnol Bioeng 81(2):

178–186.

Kensy F, Zimmermann HF, Knabben I, Anderlei T, Trauthwein H, Din-

gerdissen U, Buchs J. 2005. Oxygen transfer phenomena in 48-well

microtiter plates: Determination by optical monitoring of sulfite

oxidation and verification by real-time measurement during microbial

growth. Biotechnol Bioeng 89(6):698–708.

McDaniel LE, Bailey EG, Zimmerli A. 1965. Effect of oxygen supply rates on

growth of Escherichia coli. I. Studies in unbaffled and baffled shake

flasks. Appl Microbiol 13:109–114.

Parekh S, Vinci VA, Strobel RJ. 2000. Improvement of microbial strains and

fermentation processes. Appl Microbiol Biotechnol 54(3):287–301.

Peter CP, Lotter S, Maier U, Buchs J. 2004. Impact of out-of-phase

conditions on screening results in shaking flask experiments. Biochem

Eng J 17(3):205–215.

Samorski M, Muller-Newen G, Buchs J. 2005. Quasi-continuous combined

scattered light and fluorescence measurements: A novel measurement

technique for shaken microtiter plates. Biotechnol Bioeng 92(1):61–68.

Tunac JB. 1989. High-aeration capacity shake-flask system. J Ferment

Bioeng 68(2):157–159.

van’t Riet K. 1979. Review of measuring methods and results in nonviscous

gas-liquid mass-transfer in stirred vessels. Ind Eng Chem Process Des

Dev 18(3):357–364.

van’t Riet K. 1983. Mass transfer in fermentation. Trends Biotechnol 1(4):

113–119.

Yawalkar AA, Heesink ABM, Versteeg GF, Pangarkar VG. 2002. Gas-liquid

mass transfer coefficient in stirred tank reactors. Can J Chem Eng 80(5):

840–848.

Zavrel M, Bross D, Funke M, Buchs J, Spiess AC. 2009. High-throughput

screening for ionic liquids dissolving (ligno-)cellulose. Bioresour Tech-

nol 100(9):2580–2587.

Zimmermann HF, Anderlei T, Buchs J, Binder M. 2006. Oxygen limitation

is a pitfall during screening for industrial strains. Appl Microbiol

Biotechnol 72(6):1157–1160.

the baffled microtiter plate: increased oxygen transfer and improved online monitoring in small...

Documents