A novel milliliter-scale chemostat system for parallel cultivation of microorganisms in stirred-tank bioreactors

Accepted Manuscript Title: A novel milliliter-scale chemostat system for parallel cultivation of microorganisms in stirred-tank bioreactors Author: Andreas Schmideder Timm Steffen Severin Johannes Heinrich Cremer Dirk Weuster-Botz PII: DOI: Reference:

S0168-1656(15)30028-6 http://dx.doi.org/doi:10.1016/j.jbiotec.2015.06.402 BIOTEC 7147

To appear in:

Journal of Biotechnology

Received date: Revised date: Accepted date:

27-3-2015 10-6-2015 16-6-2015

Please cite this article as: Schmideder, Andreas, Severin, Timm Steffen, Cremer, Johannes Heinrich, Weuster-Botz, Dirk, A novel milliliter-scale chemostat system for parallel cultivation of microorganisms in stirred-tank bioreactors.Journal of Biotechnology http://dx.doi.org/10.1016/j.jbiotec.2015.06.402 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

A novel milliliter-scale chemostat system for parallel cultivation ofmicroorganisms in stirred-tank bioreactors Andreas Schmideder, Timm Steffen Severin, Johannes Heinrich Cremer, Dirk Weuster-Botz* Institute ofBiochemical Engineering, Technische Universität München, Boltzmannstr. 15, 85748 Garching

E-Mail-Addresses: [email protected] [email protected] [email protected] [email protected] *Corresponding footnote: Postal address: Institute of Biochemical Engineering, TechnischeUniversitätMünchen, Boltzmannstr. 15, 85748 Garching, Germany. Phone: + 49 (089) 28915712; Fax: + 49 (089) 28915714; E-mail: [email protected]

1

Running title: milliliter-scale chemostat system

2

Abstract

ApH-controlled parallel stirred-tank bioreactor system was modified for parallel continuous cultivation on a 10 milliliter-scale by connecting multichannel peristaltic pumpsfor

feeding

and

medium

removalwith

micro-pipes

(250

µm

inner

diameter).Parallel chemostat processes with E. colias an example showed high reproducibility with regard to culture volume and flow rates as well as dry cell weight, dissolved oxygen concentration and pH control at steady states (n = 8, coefficient of variation< 5 %). Reliable estimation of kinetic growth parametersof E. coliwas easily achieved within one parallel experiment by preselecting ten different steady states. Scalability of milliliter-scale steady state results was demonstrated bychemostat studies with a stirred-tank bioreactor on a liter-scale. Thus, parallel and continuously operated stirred-tank bioreactors on a milliliter-scale facilitate timesaving and cost reducing steady state studies with microorganisms. The applied continuous bioreactor system overcomes the drawbacks of existing miniaturized bioreactors, like poor mass transfer and insufficient process control. Key Words Miniaturized stirred-tank bioreactors, chemostat, Escherichia coli, growth kinetics, scale-up

3

Introduction

The applicationofchemostats (continuously operated ideal stirred-tank bioreactors) enablephysiological studies of cells at defined and controlled reaction conditions at steady states since the 1950s (Monod, 1950; Nocick& Szilard, 1950). The growth rate of cells can easily be controlled by the residence time of the medium in a steady state. Nowadays, in the post-genomic era, a fundamental knowledge of microbial genomes and technologies for studies of the intracellular protein, mRNA, metabolite profiles and intracellular fluxes are available.Hence, the “omic” technologies provide the opportunity to characterize physiology of microorganisms at a molecular level. To gain a maximum of reliable data, the growth of cells under a defined, constant and highly controllable set of physico-chemical reaction conditions is essential. Therefore, the chemostat is the ideal experimental system for such investigations (e.g. Hoskisson and Hobbs,2005). Parameters like medium composition, pH, temperature and oxygen supply can be controlled. Furthermore, cells are kept at a steady state with a constant growth rate and metabolic activity.That provides the possibility of detailed analysis of microbial metabolism under single substrate limitation (Hoskisson and Hobbs, 2005). The major drawbacks of chemostatcultivations are the time-consuming experimental setup and the high substrate consumption, which can be a problem if expensive substratesare used, e.g. isotopic labelled substrates for

13C

metabolic flux analysis

(Niklas et al., 2010). A general strategy to avoid these disadvantages is the miniaturization and parallelization of bioreactor systems (Akgun et al., 2008; WeusterBotz, 2005). However, only few miniaturized chemostat bioreactor systems have been

4

developed in recent years. Nanchen et al. (2006) reported the design of parallel bioreactors for continuous cultivation of Escherichia coli(E. coli) in Hungate tubes with a working volume of 10mL. Water saturated air was sucked into the bioreactors by applying negative pressure, and small stirrer bars inside the culture vessels allowed sufficient mixing and oxygen transfer. The system was equippedwith online measurement of dissolved oxygen concentration (DO) and online analysis of exhaust gas (Klein et al., 2013). Seletzky et al. (2007) introduced special shake flasks for continuous cultivation of Corynebacteriumglutamicum.Furthermore, various systems for continuous operation on amicroscale (<1mL) have been reviewed (Zhang et al., 2006; Lee et al., 2011). Most of the developed miniaturized approaches show low oxygen transfer rates. Thatlimits maximal biomass concentrations and feeding rates (Kirk and Szita, 2013). Further drawbacks are insufficientprocess control andthe restricted probe volume for further analysis of cells and fermentation broth. In this study, a well-characterized parallel bioreactorsystem with up to 48 parallel stirred-tank single-use bioreactors and a working volume of 8–14mL will be modified for continuous operation. Sufficient oxygen transfer (kLaof up to 0.4s-1) is achieved using magnetic driven gas-inducing stirrers (Puskeiler et al. 2005; Weuster-Botz et al. 2005). The system is equipped with fluorometric sensors for individual online measurement of DO and pH as well as heat exchangers for temperature control (Kusterer et al. 2008). The parallel bioreactorsystem is controlled by a liquid-handling system, which facilitates pH control and different feed profiles for fed-batch operation by intermittent addition of titration agents or carbon sources. Various batch and fedbatch process applications with different microorganisms have already been shown in the past (Knorr et al., 2007; Vester et al., 2009; Höfelet al.,2010; Gebhardt et al.,

5

2011; Hortsch et al., 2011; Höfel et al., 2012; Schmidt and Weuster-Botz, 2012;Bendig and Weuster-Botz, 2013; Faust et al., 2014). The novel milliliter-scale chemostat system will be applied exemplarily for the identification of kinetic growth parameters of an E. colistrain based on the steady state data of one parallel run on a milliliter-scale.AnE. coliBL21(DE3) was chosen harboring a pET28a(+) plasmid with a foreign gencoding for PAmCherry (27 kDA) with kanamycin as selection marker. Scalability and steady state operating points will be evaluated with a standard bench-scale stirred-tank bioreactor on a liter-scale.

Materials and Methods

Bacterial strain An E. coli BL21(DE3) strain harboring a pET28a(+) plasmid with the gene for PAmCherry (Clontech, Saint-Germain-en-Laye, France) was used for continuous cultivations. Media, seed culture and inoculation Seed cultures weregrown either in defined mediumcontaining 2gL-1glucose or in low salt LB medium (5gL-1 yeast extract, 10gL-1 peptone, 5gL-1NaCl). All fermentations on amilliliter- and liter-scale were carried out in defined mediumaccording to Riesenberg et al. (1991)with following variations: No thiamin was needed for the strain in this study and another antifoam agent (Antifoam 204, Sigma Aldrich,Taufkirchen,

Germany)

was

used

with

a

concentration

of

0.1%

(v/v).Glucoseserved as batch substrateat a concentration of20gL-1. Ammonia (12.5%

6

(v/v))was used to adjust the pH to 7.0.The defined medium was also used as feeding solution with 25gL-1 glucose.Kanamycin was used as selection marker at a concentration of 30 mg L-1. Aspre-culture a single colony of E.coli was grown to stationary phase in a 15 mL test tube with 4 mL LB medium at 200 rpm and 30 °C. For the seed culture 500mL Erlenmeyer flasks without baffles containing 100mL defined medium were inoculated with 4 mL of LB pre-culture. The cells were grown to stationary phase at 250rpm and 37°C on a rotary shaker. All bioreactor cultivations were inoculated with 10% (v/v)seed culture. Cultivation systems,process monitoring and control The reference cultivations on aliter-scale were carried out at 30°C in a 5L stirred tank bioreactor equipped with three Rushton turbines (Labfors, Infors HT, Bottmingen, Switzerland) operated with a working volume of 2L. The pH was controlled topH 7.0 with 12.5% (v/v) NH3. The DO was maintained above 25 % by increasing the stirrer speed up to 1,200rpm and the aeration up to 4.0vvm. The inlet and efflux of medium was realized by peristaltic pumps (Ismatec BVP, IDEX Health and Science GmbH, Wertheim, Germany). The parallel cultivations on a milliliter-scale were carried out withsterile single-use stirred tank bioreactors with an initial reaction volume of 10mL at 30°C (bioREACTOR, 2mag AG, Munich, Germany). DO and pH were monitored by fluorimetric sensors immobilized at the bottom of each single-use bioreactor applying fluorimetric readers (MCR 8*2 v5, PreSens GmbH, Regensburg, Germany).

7

The pH in each single reactor were controlledtopH 7.0+/- 0.2 with 12.5% (v/v) NH3 by a liquid handling system (Genesis, Tecan GmbH, Crailsheim, Germany) making use of the software fed-batchXP (DASGIP – an Eppendorf company, Jülich, Germany).The software offers a PI-controller with dead-band (pH 0.2) to avoid over titration (Kusterer et al. 2008).The minimal volume of base addition by the pipette tips was 10 µL (0.1 % of the reaction volume). As the medium (according to Riesenberg et al., 1991) is well buffered, one base addition step causes a pH increase of about pH 0.1. Sufficient oxygen supply was achieved using gas-inducing stirrers (Puskeiler et al. 2005; Weuster-Botz et al. 2005). The speed of the gas-inducing impellers was set to3,000rpm. The headspace of each milliliter reactor was rinsed with 0.1Lmin-1sterile air to guarantee sufficient oxygen supply and carbon dioxide stripping. The rotating stirrers generate a vacuum at the bottom of their hollow axis. The percentage of air, which is sucked into the medium by the gas-inducing stirrers, is a function of stirrer speed.The sterile air was saturated with water in a 1L bottle at room temperature (23° C). As the headspace cooling of each reactor was set to 20°C,liquid volume loss by evaporation was avoided.The inlet and efflux of medium was realized by peristaltic pumps (IsmatecReglo ICC, IDEX Health and Science GmbH, Wertheim, Germany; 205 U, Watson Marlow GmbH, Rommerskirchen, Germany). For this purpose, a topping was designed and manufactured of polyetheretherketone. It can be integrated into the exhaustion unit of the bioreactor system and offers apertures for the pipet tip of a liquid-handling system (3.5 mm diameter) as well as for two bent micro-pipes (0.5 mm outer diameter, 0.25 mm inner diameter; UniMed, Lausanne, Swiss). One micropipe dipping into the medium was used for feeding (length 111.9 mm).The other 8

micro-pipe for the efflux (length 101.9 mm) was fixed at a defined height (73mm from the exhaustion unit) to keep a constant reaction volume of 10 mL (Figure 1). Cultivation protocols A batch cultivation (12mL) with 20gL-1glucose was carried out to determine maximal growth rates and biomass yield of E.coliBL21(DE3) pET28a-PAmCherry. The protocols for continuous cultivations were equivalent at mL- and L-scale. A batch phase (20gL-1glucose) of 15h for biomass formation was followed by continuous cultivations at different dilution rates(25gL-1glucose). Monitoring of cell and glucose concentrations For at-line measurement of optical density (OD600) at milliliter-scale, samples were consecutively diluted 1:10 and 1:50 with PBS (8gL-1NaCl, 0.2gL-1KCl, 1.44gL-1 Na2HPO4, 0.24gL-1KH2PO4, pH7.4) by the liquid-handling system. A MTP reader (FLUOstar Galaxy, BMG LABTECH GmbH, Ortenberg, Germany) was used for atline determination of the OD600. Samples from liter scale were diluted manually and measured with a GENESYSTM 10S UV-Vis Spectrophotometer (Thermo Electron Scientific Instruments LLC, Madison, USA). A correlation between both systems allowed determination of OD600 at milliliter scale. Dry cell weight (DCW) was measured gravimetrically by centrifugation (13,000rpm, 5min) and drying (80°C, for at least 24h) of 2mL culture broth. Samples for DCW were taken frequentlyat L-scale, whereas DCW in the mL-bioreactors during the cultivation was estimated by a correlation between OD600 measured by the MTP readerand the DCWdue to the small reaction volume. At steady state,DCW 9

wasdetermined gravimetricallyat the end of the cultivationson a mL-scale.Glucose concentration was measured enzymatically using a D-Glucose-kit (R-Biopharm AG, Darmstadt, Germany). Quantification of culture volume and flow rate The culture volume on a milliliter-scale was estimated by weighing of the final liquid phase of each individual reactor.The flow rate was determined gravimetrically by weighing the substrate supply container of every bioreactor and estimating the volume with the predetermined density of 1.03gL-1. Sample volumes and addition of titrants were considered for estimation of the flow rates.

Modelling

Model equations Based on mass balances for continuous cultivations in ideal stirred-tank bioreactors with constant volume, the equations (1) and (2) were used to describe the biomass and glucose concentrations (cx and cS)in the reactor: (1) (2) where D denotes the dilution rate, YXS,µ the yield coefficient for the conversion of glucose to biomass, mSthe substrate consumption for maintenance metabolismand cS,F the glucose concentration in the feed. The growth rate µ is described by the Monod model for bacterial growth (saturation kinetics) with the maximum growth rate µmax and thesubstrate affinity constant KS(Equation 3).

10

(3) Equations (1) and (2) wereevaluated for their steady state solution (: This yields the steady state concentrations for substrate and biomass (Equation4 and 5 for dilution rates below the washout, , otherwise). (4) (5) Parameter estimation and evaluation The parameters of the differential equations were estimated by minimizing the residual function (Equation 6) using the nonlinear local optimization routine lsqnonlin from the Optimization Toolbox in MATLAB 2014b (64bit) with the Trust-Region-Reflective algorithm. (6) where r is the residual function, ci,meas is the measured concentration of i and ci,sim is the simulated concentration of i. Substrate and biomass measurements for the estimation of the kinetic parameters were measured at ten different dilution rates. Initial guesses for the parameters were taken from literature reports (Jenzsch et al., 2006; Riesenberg et al., 1991; Kovarova-Kovar and Egli, 1998; Senn et al., 1994; Wunderlich et al., 2014). They are listed in Table I,along with the optimization constraints. For estimationof the parameter uncertainties a Monte Carlo Bootstrap analysis was performed (Kremling, 2014). Therefore, artificial measurements were generated by adding a normally distributed error with a standard deviation of 5 % (estimated 11

maximum measurement error) to the measured concentrations. This process was repeated 100,000 times and yielded different parameter sets, which were evaluated for their mean and standard deviation, taken as a measure for the parameter uncertainty. To analyse the influence of KS, the effect on the growth rate was estimated. Consequently, an error propagation was used to calculate the influence of the uncertainty () on the error of the growth rate (, Equation 7). (7) In a further validation of the model, the results of the simulation were also compared to additional measurements taken from miniature bioreactors as well as cultivations on a liter-scale.

Results and Discussion

Chemostat cultivations of E. coli on a milliter-scale To characterize the suitability of the parallel milliliter-scale stirred-tank bioreactors for continuous cultivation, eight-fold fermentations with E. coli were carried out in parallel (Figure 2). The continuous mode was started at a dilution rate D=0.2h-1 with 25gL-1 glucose in the feed medium after a batch phase of 15 h . A characteristic increase of the DCW until 1.5 residence times was observed. Afterwards, a constant biomass concentration of 10.5±0.24 g L-1 was measured in the eight parallel bioreactors. The mean DO (coefficient of variation < 5%) also indicates a steady state after two residence times. The pH control by intermittent dosage of ammonia with a liquid-handling system(coefficient of variation <1% at steady state) could successfully be established. Glucose concentrations in the efflux were below detection limit (25 mg 12

L-1) at any time. The reaction volume (10.5±0.2mL) and dilution rate (0.19±0.01h-1) also showed high reproducibility and precision in the eight parallel stirred-tank bioreactors. In summary, the continuous cultivation of E. coliclearly shows the suitability of the modified small-scale stirred-tank bioreactors for continuous operations. The simultaneous fermentations also indicate the highreproducibility of the parallel milliliter scale bioreactors operated at controlled process conditions. Furthermore, the so far highest DCW ever reported in milliliter scale continuous cultures was made possible without any oxygen limitation (Nanchen et al., 2006; Seletzky et al., 2007). Parameter identification of E. coligrowth kinetics based on one set of parallel chemostat studies The dilution rates for parallel chemostat cultivations of E. coli were chosen empirically after estimation of the maximum growth rate based on batch process data (µmax,batch = 0.36h-1) : D1 = 0.08 h-1, D2 = 0.17 h-1, D3 = 0.19 h-1, D4 = 0.25 h-1, D5 = 0.27 h-1, D6 = 0.33 h-1, D7 = 0.44 h-1, and D8 = 0.45 h-1. D1 was limited by the lowest reliable feeding rate of the pump. The resulting biomass and glucose concentrations at steady state (at least after five residence times)were measured at the preset dilution rates.As no washout occurred at D8, the dilution rate in two bioreactors was increased to D9 = 0.52 h-1 and D10 = 0.90 h-1 after five residence times. Thus, ten steady states were realized within a total process time of 80h.The corresponding DCW and glucose steady state concentrations were used to estimate the parameters of the kinetic model(Figure 3).

13

The values for the estimated kinetic parameters (Table II)indicate high equivalence to literature data. Hence, µmax and YXS,µ could be identified with high precision, whereas KSand mSshowed higher parameter uncertaintiesdue to the lack of data at high dilution rates nearby the washout (KS) and low dilution rates(mS) caused by the physical limitation of minimal feeding rates by the peristaltic pumps. The parameter uncertainty of KS shows high influence on the calculation of µ at glucose concentrationsin the order of KS(Figure 4). However, this influence is reduced to a coefficient of variation of less than 1 % at glucose concentrations above 1.86 g L1.

Validation results are shown in figure 5. DCW and glucose concentrations of four further steady states in milliliter-scale stirred-tank bioreactors and the steady state data of three dilution ratesmeasured on a liter-scale (V = 2 L) demonstrated high prediction accuracy of the simulation within the estimation error with the exception of dilution rates near washout (Dwashout = 0.55 h-1) due to the uncertainty of KS and mS estimations. Scalability of continuous E. coli fermentations based on the maximal oxygen transfer ratebetween milliliter- and liter-scale stirred-tank reactors is demonstrated despite the totally different stirrers on both scales (gas-inducing stirrer on milliliter-scale and rushton turbines on liter-scale). These results confirm the outcome of previously reported scale-up / scale-down studies of batch and fed-batch fermentation processes (Knorr et al., 2007; Kusterer et al. 2008; Hoefel et al., 2010; Hortsch et al., 2010).

Conclusions

14

The integration of continuous feeding and harvesting devices into a pH-controlled parallel stirred-tank bioreactor system allows the parallel continuous cultivation of microorganisms on a milliliter scale with high reproducibility.Due to direct scalability of fermentations in stirred-tank bioreactors between the milliliter- and liter-scale, the experimental effort for continuous cultivation can be reduced significantly with continuously operated small-scale stirred-tank bioreactors. The small reaction volume of 10 mL drastically reduces costs compared to standard stirred-tank bioreactorson a liter-scale usually applied. This is of special interest if cost-intensive media compounds like isotopically labeled substrates are used for e.g. steady state estimation of intracellular flux distributions (Niklas et al., 2010). The modified parallel stirred-tank bioreactor system offers the highest oxygen transfer rates (kLaof up to 0.4s-1) of miniaturized bioreactors reported in literature (Kirk and Szita, 2013). Thus, higher cell densities can be achieved compared to other continuous cultivation systems on a milliliter scale (Nanchen et al., 2006; Seletzky et al., 2007). Parallel chemostat studies of E. coli, which have been studied as an application example for the continuously operated stirred-tank bioreactors on a 10 mL-scale, enable the fast identification of kinetic model parameters. A drawback of parallel steady state experiments is the necessity of pre-selecting dilution rates without or with minor knowledge of the washout. This results in less accurate estimation of kinetic parameters like KS. A solution may be the increase of the number of parallel-operated chemostats, which increases the probability of experimental studies with dilution rates near washout. Another possibility may be the sequential application of parallel chemostat studies with successive parameter estimations followed by the application of methods of statistical design (e.g. D-optimal design) to select a set of optimal dilution 15

rates for the identification of the kinetic parameters after each parallel run (Takors et al.,1997).

Acknowledgements

The authors gratefully acknowledge the excellent technicalassistance of Florentine Haug, BjörnEckhardt and Norbert Werth(Institute of Biochemical Engineering, TechnischeUniversitätMünchen, Garching, Germany) and the provision of the PAmCherry gene by Clontech (Saint-Germain-en-Laye, France).The support of Andreas Schmideder by the TUM Graduate School (TechnischeUniversitätMünchen, Germany) is acknowledged as well.

Nomenclature

ci,meas

Measured concentration of i (g L-1)

ci,sim

simulated concentration of i (g L-1)

cS

substrate concentration (g L-1)

cS,F

substrate concentration in feed (g L-1)

cx

biomass concentration (g L-1)

D

dilution rate (h-1)

DCW

dry cell weight (g L-1)

16

DO

dissolved oxygen (%)

E. coli

Escherichia coli

kLa

oxygen transfer coefficient (s-1)

KS

substrate affinity constant (g L-1)

µ

growth rate (h-1)

µmax

maximal growth rate (h-1)

mS

maintenance coefficient (g g-1 h-1)

OD600

optical density at 600 nm (-)

qS

substrate uptake rate (h-1)

r

residual function

t

time (s or h)

τ

residence time (h)

YXS,µ

biomass yield coefficient (g g-1)

17

References

Akgun A, Muller C, Engmann R, Büchs J. 2008. Application of an improved continuous parallel shaken bioreactor system for three microbial model systems. Bioprocess BiosystEng 31(3): 193-205. Bendig C, Weuster-Botz D. 2013. Reaction engineering analysis of cellulase production with Trichodermareesei RUT-C30 with intermittent substrate supply. BioprocBiosys Eng 36:893-900. Faust G, Janzen N, Bendig C, Römer L, Kaufmann K, Weuster-Botz D. 2014. Feeding strategies enhance high cell density cultivation and protein expression in milliliter scale bioreactors. Biotechnol J 9:1293-1303. Gebhardt G, Hortsch R, Kaufmann K, Arnold M, Weuster-Botz D. 2011. A new microfluidic concept for parallel operated milliliter-scalestirred tank bioreactors. BiotechnolProg 27:684-690. HöfelT, Wittmann E, Reinecke L, Weuster-Botz D. 2010. Reaction engineering studies

for

the

production

of 2-hydroxyisobutyric acid with recombinant

Cupriavidusnecator H 16. ApplMicrobiolBiotechnol 88: 477-484. Höfel T, Faust G, Reinecke L, Rudinger N, Weuster-Botz D. 2012. Comparative reaction engineering studies for succinic acid production from sucrose by metabolically engineered Escherichia coli in fed-batch operated stirred tank bioreactors. Biotechnol J 7:1277-1287. 18

Hortsch R, Stratmann A, Weuster-Botz D. 2010. New milliliter-scalestirred tank bioreactors for the cultivation of mycelium formingmicroorganisms. BiotechnolBioeng 106:443-451. Hortsch R, Weuster-Botz D. 2010. Milliliter-scale stirred tank reactors for the cultivation of microorganisms. AdvApplMicrobiol 73:61-82. Hortsch R, Krispin H, Weuster-Botz D. 2011. Process performance of parallel bioreactors for batch cultivation of Streptomyces tendae. BioprocBiosysEng 34:297304. Hoskission PA, Hobbs G. 2005. Continuous culture making a comeback? Microbiology 151(Pt10):3153-3159. Jenzsch M, Gnoth S, Beck M, Kleinschmidt M, Simutis R, Lübbert A. 2006. Openloop control of the biomass concentration within the growth phase of recombinant protein production processes. J Biotech 127:84-94. Kirk TV, Szita N. 2013. Oxygen transfer characteristics of miniaturized bioreactor systems. BiotechnolBioeng 110(4):1005-1019. Klein T, Schneider K, Heinzle E. 2013. A system of miniaturized stirred bioreactors for parallel continuous cultivation of yeast with online measurement of dissolved oxygen and off-gas. BiotechnolBioeng 110(2):535-542. Knorr B, Schlieker H, Hohmann HP, Weuster-Botz D. 2007. Scale-down and parallel operation of the riboflavin production process with Bacillus subtilis. BiochemEng J 33:263-274.

19

Kovarova-Kovar K, Egli T. 1998. Growth kinetics of suspended microbial cells: from single-substrate-controlled growth to mixed-substrate kinetics. MicrobiolMolBiol Rev 62(3):646-666. Kremling A. 2014. Systems biology: mathematical modeling and model analysis. Boca Raton:CRC Press. 97p. Kusterer A, Krause C, Kaufmann K, Arnold M, Weuster-Botz D. 2008. Fully automated single-use stirred-tank bioreactors for parallelmicrobial cultivations. Bioprocess BoiosystEng 31:207-215. Lee KS, Boccazzi P, Sinskey AJ, Ram RJ. 2011. Microfluidic chemostat and turbidostat with flow rate oxygen, and temperature control for dynamic continuous culture. Lab Chip 11:1730-1739. Monod J. 1950. Continuous culture technique: Theory and applications. Ann Inst Pasteur 79(4):390-410. Nanchen A, Schicker A, Sauer U. 2006. Nonlinear dependency of intracellular fluxes on growth rate in miniaturized continuous cultures of Escherichia coli. Appl Environ Microbiol 72(2):1164-1172. Niklas J, Schneider K, Heinzle E. 2010. Metabolic flux analysis in eukaryotes. CurrOpinBiotechnol 21(1):63-69. Novick A, Szilard L. 1950. Experiments with the chemostat on spontaneous mutations of bacteria. ProcNatlAcadSci USA 36(12):708-719. Puskeiler R, Kaufmann K, Weuster-Botz D. 2005b. Development, parallelization and 20

automation of a gas-inducing milliliter-scale bioreactor for high-throughput bioprocess design (HTBD). BiotechnolBioeng 89:512-523. Puskeiler R, Kusterer A, John GT, Weuster-Botz D. 2005a. Miniature bioreactors for automated high-throughput bioprocess design (HTBD): Reproducibility of parallel fedbatch cultivations with Escherichia coli. BiotechnolApplBiochem 42:227-235. Riesenberg D, Schulz V, Knorre WA, Pohl H-D, Korz D, Sanders EA, Roß A, Deckwer W-D. 1991. High cell density cultivation of Escherichia coli at controlled specific growth rate. J Biotech 20:17-28. Schmidt M, Weuster-Botz D. 2012. Reaction engineering studies of acetone-butanolethanol fermentation with Clostridium acetobutylicum. Biotechnol J7:656-661. Seletzky JM, Noak U, Fricke J, Welk E, Eberhard W, Knocke C, Büchs J. 2007. Scale-up from shake flasks to fermenters in batch and continuous mode with Corynebacteriumglutamicum on lactic acid based on oxygen transfer and pH.BiotechnolBioeng 98(4):800-811. Senn H, Lendenmann U, Snozzi M, Hamer G, Egli T. 1994. The growth of Escherichia coli in glucose-limited chemostat cultures: a re-examination of the kinetics. BiochimBiophys Acta 1201:424-436. Takors R, Wiechert W, Weuster-Botz D. 1997. Experimental design for the identification of macrokinetic models and model discrimination. Biotech Bioeng 56: 564-576. Vester A, Hans M, Hohmann P, Weuster-Botz D. 2009. Discrimination of riboflavin producing Bacillus subtilis strains based on their fed-batch process performances on a 21

milliliter-scale. ApplMicrobiolBiotechnol 84: 71-76. Weuster-Botz D. 2005. Parallel reactor systems for bioprocess development. AdvBiochemEngBiotechnol 92:125-143. Wunderlich M, Taymaz-Nikerel H, Gosset G, Ramirez OT, Lara AR. 2014. Effect of growth rate on plasmid DNA production and metabolic performance of engineered Escherichia coli strains. J BiosciBioeng 117(3):336-342. Zhang Z, Boccazzi P, Choi H-G, Perozziello G, Sinskey AJ, Jensen KF. 2006. Microchemostat-microbial continuous culture in a polymer-based, instrumented microbioreactor. Lab Chip 6:906-913.

22

23

24

25

26

Table 1: Initial parameter guesses and constraints used for the parameter estimation (growth of E. coli with glucose as limiting substrate). Parameter

Initial

Lower

Upper

µmax (h-1)

0.5

0

1

YXS,µ, (g g-1)

0.5

0

1

KS, (g L-1)

0.2

0

100 27

mS, (g g-1 h-1)

0.02

0

0.1

Table 2: Estimated kinetic parameters for the growth of a recombinant E. coli BL21 strain on glucose in comparison to exemplary literature data for E. coli (Jenzsch et al., 2006; Riesenberg et al., 1991; Kovarova-Kovar and Egli, 1998; Senn et al., 1994; Wunderlich et al., 2014). Parameter

Estimatedvalue

Parameter Uncertainty

Literature

µmax

0.55 ± 0.008 h-1

1,53 %

0.45-0.67 h-1

YXS,µ

0.51 ± 0. 019 g g-1

3.76 %

0,49 g g-1

KS

90 ± 19 mg L-1

21.8 %

5-99 mg L-1

mS

46 ± 15 mg g-1 h-1

32.0 %

32 mg g-1 h-1

28

Figure captions Figure 1: (A) Schematic setup of the new bioreaction system. Shown are two units out of eight. (B) Top view of the constructed topping.It can be integrated into the exhaustion unit of the bioreactor system, with apertures for a pipette tip (diameter 3.5 mm) and two bent micro-pipes (diameter 0.5 mm) for the realization of feeding and medium removal on a milliliter-scale. (C) Photograph of the modified system. Figure 2: Means of DCW*, which was estimated based on the measured OD600, DO, pH (black) with standard deviations (grey) as function of residence times: continuous cultivations of Escherichia coli in 8 parallel operated stirred-tank bioreactors (V = 10 mL, D = 0.2 h-1, τ = 5 h, cS,F = 25 g L-1 glucose, T = 30 °C). Figure 3: Parallel chemostat studies with E. coli (V = 10 mL, cS,F = 25 g L-1 glucose, T = 30 °C, pH = 7.0): Steady state data of DCW () and glucose () concentration and simulations as function of dilution rate. Figure 4: Coefficient of variation of calculated µ as function of glucose concentration for normally distributed parameter KS. The vertical grey line indicates the determined value of KS (90 mg L-1). Figure 5: Parallel chemostat studies with E. coli (V = 10 mL, cS,F = 25 g L-1 glucose, T = 30 °C, pH = 7.0) and sequential chemostat studies (V = 2 L, cS,F = 25 g L-1 glucose, T = 30 °C, pH = 7.0): Steady state data of DCW (black) and glucose (grey) concentrations on a milliliter- () and liter-scale () and simulations as function of dilution rate.

29

30