COMMUNICATION TO THE EDITOR

Simple Technique for SimultaneousOn-Line Estimation of Biomass andAcetate from Base Consumption andConductivity Measurements in High-CellDensity Cultures of Escherichia coli

Frank Hoffmann, Michael Schmidt,* Ursula Rinas

Biochemical Engineering Division, GBF German Research Centre forBiotechnology, Mascheroder Weg 1, 38124 Braunschweig, Germany;Telephone: 0049-531-6181-126; Fax: 0049-531-6181-111; e-mail: URI@gbf.de

Received 3 October 1999; accepted 11 June 2000

Abstract: In high-cell density cultures of Escherichia coli,the amount of ammonia required to maintain a constantpH was an effective on-line measure of biomass. Whenacetic acid formation occurred, biomass estimates werecorrected by additional measurements of the cultureconductivity. In the absence of acetic acid excretion, theculture conductivity decreased approximately linearlyduring the fed-batch phase of the high-cell density culti-vation. Upward deviation of the culture conductivityfrom this trend was proportional to the acetate concen-tration in the medium. Thus, simultaneous determina-tion of the acetic acid concentration and correction ofbiomass estimates from ammonia consumption werepossible by conductivity measurements. © 2000 JohnWiley & Sons, Inc. Biotechnol Bioeng 70: 358–361, 2000.Keywords: on-line; estimation; biomass; acetate

INTRODUCTION

Successful process monitoring and control of bacterial cul-tivations often demands on-line available information aboutthe physiological state of the cells. The required quantitiesinclude specific rates, rather than volumetric rates, of bio-mass formation, substrate consumption and product forma-tion. Readily accessible and reliable measurements shouldform the basis for the estimates. Simple means of biomassestimation may, under certain circumstances, fail to giveaccurate results.

Optical density is a fast and easy method for determina-tion of cell density in many instances. However, the corre-lation of cell density and optical density of the culture maychange during production of proteins that aggregate andform inclusion bodies (Hwang and Feldberg, 1990). Esti-

mation of biomass and growth rate from respiratory activity(Gbewonyo et al., 1989; Estler, 1995; Petkov and Davis,1996) requires constant yield coefficients with respect tooxygen. Induction of recombinant genes, however, imposesa metabolic burden on the cell and reduces yield coefficients(Andersson et al., 1996; Schmidt et al., 1999a,b).

An alternative approach uses the ammonia consumptionof the biomass as a measure of cell growth (Schmidt et al.,1999b). This is possible because ionic concentrations haveonly a minimal effect on the rate of ammonia addition forpH control (Siano, 1995). Even after induction of recombi-nant protein production at high cell densities, the biomassyield coefficient with respect to ammonia does not varysignificantly and, thus, ammonia consumption can be usedfor on-line calculation, e.g., of the specific growth rate(Schmidt et al., 1999b), as long as no organic acids areexcreted. Changes of process parameters, such as the feed-ing rate, during process optimization potentially give rise tothe formation of acetic acid. When estimating biomass fromammonia consumption, it is thus important to accountfor base added to neutralize acetic acid and maintain a con-stant pH.

It is feasible to measure acetate on-line by flow injectionanalysis (Becker et al., 1993) and various HPLC methods(Turner et al., 1993), e.g., ion-exclusion chromatography(Scheller et al., 1994) or cation-exchange chromatography(von Zumbusch et al., 1994). All these methods requireadditional equipment and a sterile, protein-free samplestream, which can be difficult to obtain from high cell den-sity cultures. Anderson and Flickinger (1993) claim thatgrowth and acetic acid excretion can be monitored by mi-crocalorimetry. However, there is a poor correlation be-tween the specific heat evolution and acetate accumulationrate.

In this study, we use ammonia consumption and cultureconductivity to simultaneously estimate acetate concentra-

Correspondence to:U. Rinas*Present address:M. Schmidt, Christian Hansen Nienburg GmbH,

Abteilung Prozeßentwicklung, Große Drakenburger Str. 93-97, 31582Nienburg

© 2000 John Wiley & Sons, Inc.

tion and biomass in high-cell density cultures of recombi-nant E. coli TG1:plFGFB producing the human basic fi-broblast growth factor.

MATERIALS AND METHODS

Strain and Plasmid

Escherichia coli TG1 supE hsdD5 thi D(lac-proAB)F8[traD36 proAB+ lacIq lacZDM15] (Carter et al., 1985)carrying the plasmid plFGFB encoding human basic fibro-blast growth factor (hFGF-2) under the control of the tem-perature-induciblelPRPL/cI857 promoter-repressor system(Seeger et al., 1995) was stored as glycerol stock culture at−70°C.

High-Cell Density Cultivation: Culture Conditions,On-Line Analysis, and Calculations

High-cell density cultivations were carried out in a 50-Lbioreactor (Type U50K; B. Braun Diessel Biotech GmbH,Melsungen, Germany) essentially as described previously(Schmidt et al., 1999a,b). The preparation and the compo-sition of the batch and feed medium has been describedbefore (Korz et al., 1995). After initial batch phase (startingvolume 30 L) and fed-batch phase at 30°C, heterologousgene expression was induced by a temperature shift to 42°C.The feeding ratemGlc of the carbon substrate during thefed-batch phase of the cultivation was controlled by asimple feed-forward strategy (Korz et al., 1995), aiming ata constant specific growth rate ofmset 4 0.12 h−1 beforeand ofmset 4 0.08 h−1 after induction (eq. 1),

mGlc~t! = Smset

YX/S+ mSD ? ~X ? V!f ? exp~mset? ~t − tf!! (1)

YX/Sdenotes the biomass yield coefficient (0.5 g g−1), mS themaintenance coefficient (0.025 g g−1 h−1), X the cell density(g L−1), V the culture volume (L), andt the time (h). Theindex f indicates the respective values at the feeding start.Values entered for the standard protocol are given in brack-ets. The pH was kept constant at pH 6.7 by the addition ofaqueous ammonia and the dissolved oxygen concentrationwas maintained at 40% of air saturation by increasing stirrerspeed and/or aeration rate (Schmidt et al., 1999a). If re-quired, the inlet air was enriched with pure oxygen.

The specific culture conductivityk (mS cm−1) was ob-tained with a conductivity probe 2EL6-1-Pt100 and a trans-mitter 2AML2-70/010/03 (Jumo, Fulda, Germany). Theprobe was modified to fit into a[25 mm port. Cultureturbidity was measured by a probe AF-44 S coupled to theturbidity meter AS-82 (Aquasant Messtechnik AG, Buben-dorf, Switzerland). Placing the reservoirs for ammonia andfeeding solutions on balances enabled on-line balancing ofsubstrate and base consumption. The actual total biomass inthe bioreactorXtot was then calculated according to Eq. 2:

Xtot 4 YX/NH3?cNH3

? (mNH3− Dk?kHAc) (2)

HeremNH3is the mass of the ammonia solution added into

the bioreactor (g),cNH3the concentration of ammonia in this

solution (g g−1), and YX/NH3the biomass yield coefficient

with respect to ammonia (YX/NH34 7 g g−1; Schmidt et al.,

1999b).The difference between the actual specific culture con-

ductivity k and a linear extrapolation of the specific cultureconductivity from the beginning of the feeding phase I,Dk,was used to correct the biomass estimate by the empiricalfactor kHAc (Eq. 3). We thereby assume that at pH 6.7 onemol of ammonia is required to neutralize one mol of aceticacid.

kHAc = 25g

mS cm−1 ?MNH3

MHAc ? cNH3

(3)

HereM is the molecular mass of ammonia (NH3) and aceticacid (HAc), respectively. The specific growth ratem wascalculated from on-line estimates of the biomassXtot (g)according to Eq. 4:

m =d(ln(Xtot/g))

dt(4)

To reduce the noise by calculating the specific growth ratefrom the slope of ln(Xtot/g) versus time, linear regressionover the last 16 data points (corresponding to 30 min) wascarried out. This procedure delayed the on-line estimation ofthe specific growth rate by 18 min, but considerably reducedthe scattering of the estimates.

Off-Line Analytical Methods

Cell dry mass was determined by centrifuging (18,000g) 1mL culture samples in preweighed 1.5 mL microcentrifugetubes. The pellets were dried under vacuum at 40°C. Thevolume of the cell suspension in the reactor was taken fromthe signal of the bioreactor balance (density of the cell sus-pension4 1 ± 0.01 g mL−1 for cell densities from 0.05 −100 g L−1). Determination of acetate and glucose concen-trations by HPLC was described earlier (Seeger et al.,1995).

RESULTS AND DISCUSSION

Biomass Estimate from Ammonia Consumption

High cell density cultivations ofE. coli TG1:plFGFB con-sisted of a batch and fed-batch phase I at 30°C, after whichthe temperature was shifted to 42°C (fed-batch phase II) toinduce the synthesis of the recombinant protein hFGF-2.Glucose feeding rates aiming at specific growth rates ofmset

4 0.12 h−1 andmset 4 0.08 h−1 during fed-batch phases Iand II, respectively, were employed. These culture condi-tions did not result in the excretion of acetic acid. Due to theconstant yield coefficient of biomass per nitrogenYX/N forvarious recombinantE. coli strains (Schmidt et al., 1999b),the biomass formed during high cell density cultivation ofE. coli TG1:plFGFB was proportional to the amount of

COMMUNICATION TO THE EDITOR 359

aqueous ammonia consumed to maintain a constant pH(Fig. 1a). From the biomass estimates, the specific growthrate was calculated (Fig. 1a). The biomass estimate as wellas the estimate of the specific growth rate fitted well withthe values determined off-line from cell dry mass measure-ments. A good fit of both on-line estimates with off-linedata was even observed at higher cell densities and aftertemperature-induced production of hFGF-2. In contrast, on-line culture turbidity measurements were not a suitablemethod for biomass monitoring at higher cell densities andduring temperature-induced protein production (Fig. 1b).The low cell density in the batch phase hampered accuratedetermination of the specific growth rate from ammoniaconsumption (Fig. 1a) because the base was added intermit-

tently. Nevertheless, during both fed-batch phases accurateestimates for the biomass and specific growth rate wereobtained.

Time-Course of Culture Conductivity in theAbsence of Acetic Acid Excretion

In the absence of acetic acid excretion, the specific cultureconductivityk decreased approximately linearly during thefed-batch phase of the cultivation, when the dissolved oxy-gen concentration was mainly controlled by increasing thestirrer speed (Fig. 2). A negative deviation from this trendwas observed when control of the dissolved oxygen con-centration occurred by additionally increasing the air flowrate. A mean slope of dk/dt 4 −0.66 ± 0.04 mS cm−1 h−1

was determined by linear interpolation from 12 independenthigh-cell density cultivations. The time course of the spe-cific culture conductivity followed a linear trend in spite ofthe exponential feed and exponential cell growth becausethe specific conductivity depends not only on the concen-tration of electrically conductive ions, but is also influencedby, e.g., the presence of bubbles.

Correction of the Biomass Estimate for AceticAcid Excretion Using Culture Conductivity

The feeding rate after the temperature shift was varied fromthe protocol shown in Figure 1 by inserting a smaller yieldcoefficient in the feeding formula of Eq. 1. With a doubledfeeding rate, the excess glucose resulted in the formation ofacetic acid (Fig. 3b) and a subsequent enhanced addition ofbase (ammonia) for pH control (Fig. 3a). At the same time,the specific conductivity of the culture broth increased

Figure 1. Estimation of biomass and specific growth rate from ammoniaconsumption.E. coli TG1:plFGFB was grown in batch and fed-batchphase I (mset 4 0.12 h−1) at 30°C. The temperature was then shifted to42°C to induce synthesis of the recombinant protein hFGF-2, and thefeeding rate was reduced (mset 4 0.08 h−1; fed-batch phase II). Transitionbetween the phases are indicated by vertical lines. Restart of the feedingafter a period of frequent sampling is indicated by an arrow.A: (d)Biomass (cell density as determined from dry mass measurements multi-plied with the respective culture volume) and (h) specific growth rate weredetermined off-line. Lines represent on-line estimates of biomass and spe-cific growth rate from ammonia consumption by a procedure describedbefore (Schmidt et al., 1999b).B: (j) Cell density as determined from drymass measurements and (——) on-line measured culture turbidity.

Figure 2. Specific culture conductivityk in fed-batch cultivations with-out acetic acid excretion. Two representative fed-batch cultivations ofE.coli TG1:plFGFB, culture conditions as in Figure 1. The dissolved oxygenconcentration was either mainly controlled by increasing the stirrer speed(——) or by additionally increasing the air flow rate (– –.). After thetemperature upshift, the inlet air was enriched with pure oxygen and thedissolved oxygen concentration was mainly controlled by stirrer speed inboth cultivations. Time is given relative to the temperature upshift.

360 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 70, NO. 3, NOVEMBER 5, 2000

steeply (Fig. 3a), due to the increasing concentrations ofelectrically conductive ammonium and acetate ions in themedium. The time course of the specific conductivity devi-ated significantly from the linear extrapolation in the ab-sence of acetic acid excretion (Fig. 3a). The differenceDkbetween the actual conductivityk and the linearly extrapo-lated conductivity was proportional to the concentration ofacetate, and thus can be used for on-line monitoring ofacetic acid excretion (Fig. 3b). By subtracting from the am-monia consumption the amount of ammonia required toneutralize acetic acid (Eq. 2), a corrected estimate for bio-mass was achieved that corresponded well with the off-linedetermined values (Fig. 3a).

CONCLUSION

Biomass ofE. coli in high-cell density cultivation can beestimated on-line from ammonia consumption. Upon acetic

acid excretion, biomass estimates can be corrected and ac-etate concentration can be determined on-line using the spe-cific conductivity of the culture broth. Hence, this simplemethod seems suited to obtain on-line information about thephysiological state of the bacteria, unless extreme distur-bances of the culture such as sudden reductions of the feed-ing rate, excess antifoam addition, and acetate concentrationexceeding 25 g L−1 with concomitant glucose accumulationand inhibition of respiratory activity affect the accuracy ofthe described method.

References

Anderson CR, Flickinger MC. 1993. Monitoring growth and acetic acidsecretion by a thermotolerantBacillususing conduction microcalorim-etry. J Ind Microbiol 12:114–120.

Andersson L, Yang S, Neubauer P, Enfors SO. 1996. Impact of plasmidpresence and induction on cellular responses in fed batch cultures ofEscherichia coli.J Biotechnol 46:255–263.

Becker T, Kittsteiner-Eberle R, Luck T, Schmidt HL. 1993. On-line de-termination of acetic acid in a continuous production ofAcetobacteraceticus.J Biotechnol 31:267–275.

Carter P, Bedouelle H, Winter G. 1985. Improved oligonucleotide site-directed mutagenesis using M13 vectors. Nucleic Acids Res 13:4431–4443.

Estler MU. 1995. Recursive on-line estimation of the specific growth ratefrom off-gas analysis for the adaptive control of fed-batch processes.Bioprocess Eng 12:205–207.

Gbewonyo K, Jain D, Hunt G, Drew SW, Buckland BC. 1989. On-lineanalysis of Avermectin fermentation cell growth kinetics in an indus-trial pilot plant. Biotechnol Bioeng 34:234–241.

Hwang SO, Feldberg RS. 1990. Effect of inclusion body production onculture turbidity and cell dry weight in growing bacterial cultures.Biotechnol Prog 6:48–50.

Korz DJ, Rinas U, Hellmuth K, Sanders EA, Deckwer WD. 1995. Simplefed-batch technique for high cell density cultivation ofEscherichiacoli. J Biotechnol 39:59–65.

Petkov SB, Davis RA. 1996. On-line biomass estimation using a modifiedoxygen utilization rate. Bioprocess Eng 15:43–45.

Scheller U, Siedenberg D, Hu¨bner U, Siebold M, Kretzmer G, Schu¨gerl K.1994. Ion chromatography for monitoring biotechnological processes.Part II. Applications for industrial cultivation media. Anal Chim Acta286:179–187.

Schmidt M, Babu KR, Khanna N, Marten S, Rinas U. 1999a. Temperature-induced production of recombinant human insulin in high-cell densitycultures of recombinantEscherichia coli.J Biotechnol 68:71–83.

Schmidt M, Viaplana E, Hoffmann F, Marten S, Villaverde A, Rinas U.1999b. Secretion-dependent proteolysis of heterologous protein by re-combinantEscherichia coliis connected to an increased activity of theenergy-generating dissimilatory pathway. Biotechnol Bioeng 66:61–67.

Seeger A, Schneppe B, McCarthy JEG, Deckwer WD, Rinas U. 1995.Comparison of temperature- and isopropyl-b-D-thiogalacto-pyranoside-induced synthesis of basic fibroblast growth factor in high-cell-density cultures of recombinantEscherichia coli.Enzyme MicrobTechnol 17:947–953.

Siano SA. 1995. On the use of the pH control reagent addition rate forfermentation monitoring. Biotechnol Bioeng 47:651–665.

Turner C, Thornhill NF, Fish NM. 1993. A novel method for the on-lineanalysis of fermentation broth using a sampling device, microcentri-fuge and HPLC. Biotechnol Tech 1:19–24.

von Zumbusch P, Meyer-Jens T, Brunner G, Ma¨rkl H. 1994. On-linemonitoring of organic substances with high-pressure liquid chroma-tography (HPLC) during the anaerobic fermentation of waste-water.Appl Microbiol Biotechnol 42:140–146.

Figure 3. Simultaneous estimation of biomass formation and acetic acidexcretion from ammonia consumption and specific conductivity. The feed-ing rate after the temperature shift was twice the feeding rate used in thecultivation shown in Figure 1.A: Biomass estimates from ammonia con-sumption (........) were corrected using the differenceDk between the actualspecific conductivity of the culture broth (— .) and a linear extrapolation ofthe specific conductivity from the beginning of the fed-batch phase I (– – –).Off-line determined biomass values (j) are compared to the correctedestimates (——).B: The differenceDk (line) was proportional to theacetate concentration (m) in the culture broth. At the end of the cultivation,glucose (s) accumulated in the medium.

COMMUNICATION TO THE EDITOR 361