_______ iotechnology
`~-and--
`Bioengineering
`
`BEQ 1035
`Page 1
`
`

`
`l
`
`l
`
`l
`
`Editor
`Douglas S. Clark
`DeparfiiiCIII of Che111ica/ E11gineeriHI}, Uniuersity of Ca/1jomia
`201 Gi/111an Hall, Berr<eley, CA 94720-1462
`
`Associate Editors
`
`Editorial Board
`
`James D. Bryers
`Lbiiucrsity of Co1111Cctiwt
`Farlllillljloii, Collllecticllt
`
`Pauline Doran
`llllil!ersity of New So11th Wales
`Syd11ey, Australia
`
`Jonathan S. Dordick
`Re~~ssclacr Polytcclmic illslitllte
`Troy, New York
`
`George Georgiou
`lllliuersity of Texas
`At1sli11, Texas
`
`Jeffrey A. Hubbell
`ETH-Ziiric!J a11d
`Uniuersity of Ziiric!J
`Zitric!J, Switzer/mid
`
`Maria-Regina Kula
`Hcillric!J-Hcine lllliuersitiit
`Diisscldo1f
`i11 der KFA liilic!J, Ger111a11y
`
`LeeR. Lynd
`Dartllloiit!J College
`Ha11o1Jer, New HmHJlshire
`
`Tadashi Matsunaga
`Tokyo ll11iucrsity of Agriculliire
`a11d Teclmology
`Tokyo, Iapa~~
`Eleftherios T. Papoutsakis
`Nort/Jwestcm Lbiiuersity
`Eua11slo11, Illi11ois
`
`John Villadsen
`Tee/mica/ lllliuersily of De11111ark
`Ly11ghy, Dmmarl<
`
`James E. Bailey
`ETH-Ziirich
`Ziirich, Switzerla~~d
`
`Harvey W. Blanch
`lliJiuersity of Califomia
`Berkeley, Califomia
`
`!sao Karube
`lliiiuersity of Tokyo
`Tokyo, Japa11
`
`Alexander M. Klibanov
`Massach11setts Instilllle of
`Tech11ology
`Ca111bridge, Massach11sctts
`
`Bernhard 0. Palsson
`Lbliuersity of Califomia at
`Sail Diego
`La Jolla, Califomia
`
`Michael L Shuler
`Come// ll11i1Jersity
`Ithaca, New Yori<
`
`Daniel I.C. Wang
`Massachusetts l11stitute of
`Tech11ology
`Ca111bridge, Massachusetts
`
`Founding Editor: Elmer L Gaden, Jr., Uniuersity of Virgi11ia, Charlottesuille, Virginia
`
`Biotechnology and Bioengineering (ISSN, 0006-3592) Volumes 62-66 are published 28
`times a year, semi-monthly, except tri-monthly in Februaf)', May, August, and November,
`by John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158.
`
`Postmaster: Send address changes to Biotechnology and llioengineering. Caroline
`Rothaug, Director, Subscription Fulfillment and Distribution, John Wiley & Sons, lnc. 1
`605 Third Avenue, New York, NY 10158.
`
`Copyright © 1999 John Wiley & Sons, Inc. All rights reserved. No part of this publication
`may be reproduced in any form or by any means, except as permitted under section I 07
`or I 08 of the 1976 United States Copyright Act, without written permission of the pub(cid:173)
`lisher, authorization through the Copyright Clearance Center, 222 Rosewood Drive,
`Danvers, MA 01923, 508-750-8400, fax (508) 750-4470. Periodicals postage paid at New
`York, NY, and at additional mailing offices. The Combinatorial Chemistry section, vol.
`61 ( 1-4), will be sent via standard mail.
`
`The code and copyright notice appearing at the bottom of the first page of an item in the
`journal indicates the copyright holder's consent that copies may be made for personal or
`internal use, or for the personal or internal use of specific clients, on the condition that
`the copier pay for copying beyond that permitted by Sections I 07 and I 08 of the United
`States Copyright Law. The per copy fee for each item is to be paid through the
`Copyright Clearance Center. This consent does not extend to other kinds of copying,
`such as copying for general distribution, for advertising or promotional purposes, for cre(cid:173)
`ating new collective works, or for resale. Such pennission requests and other pem1ission
`inquiries should be addressed to the Permissions Dept.
`
`Subscription Price (Volumes 62-66, 1999), $3195.00 in US, $3475.00 in Canada and
`Mexico, $3713.00 outside North America. All subscriptions outside US will be sent by
`air. Personal rate, $525.00 in US, and $693.00 outside North America. Coml>i>taiOiinl
`Chemistry section personal rate, $145.00 in US, and $169.00 outside North America.
`Subscriptions at the personal rate. are available to individuals. Payment ~ust be
`made in US dollars drawn on a US bank. Claims for undelivered copies will be
`accepted only after the following issue has been received. Please enclose a copy of the
`mailing label. Missing copies will be supplied when losses have been sustained in
`tra-nsit and where reserve stock permits. Please allow four weeks for processing a
`change of address. For subscription inquiries, please call (212) 850-6645;
`F.-mail, SlJBINFO@wiley.com
`
`Advertising Sales: Inquiries concerning advertising should be forwarded to Advertising
`Sales, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158 1
`(212) 850-8832. Advertising Sales, European Contacts, Jackie Sibley,
`John Wiley & Sons, Ltd., Baffins Lane, Chichester, West Sussex P019 IUD, England.
`TeL 44 1243 770 351 1 Fax, 44 1243 770 432 1 e-mail, adsales@wiley.co.uk.
`Reprints: Reprint sales and inquiries should be directed to the Customer Service
`Department, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158. Tel:
`(212) 850-8776.
`
`Manuscripts for Biotrclmology & Biomgi>~mil1g should be submitted to the Editor, Professor
`Douglas S. Clark, Dept. of Chemical Engineering, University of California, 20 I Gilman
`Hall, Berkeley, CA 94720-1462. Manuscripts for Coi»Vi>lntorin/ Clmuistry should be
`submitted to the Editor, Dr. David J. Ecker, Isis Pharmaceuticals, Inc., 2292 Faraday
`Avenue, Carlsbad, CA 92008.
`
`AU Other Correspondence: Address to Biotechnology and Bioengineering, Publisher,
`lnterscience Division, Professional, Reference, and Trade Croup, John Wiley & Sons, lnc.l
`605 Third Ave., New York, NY I 0158.
`
`Information for Contributors appears in the first and last issue of each volume. The con(cid:173)
`tents of this journal are indexed or abstracted in Biological Abstracts, BIOSIS, Chemical
`Abstracts, Chemical Titles, Current Contents/Life Sciences, Current Contents/Agriculture~
`Biology, and Environmental Science, Engineering lnJex, Excerpta Medica, Reference
`Update, Research Alert (lSI), Science Citation Index (lSI), SCISEARCH Database (lSI),
`and Index Medicus/,\.IEDUNE.
`
`Edilorlal Production, John Wiley
`Diane C('rn
`
`@> This paper meets the requirement of ANSIINISO Z39.48-1992 (Permanence of
`Pap<W)
`
`BEQ 1035
`Page 1
`
`

`
`On-Line Detection of Acetate Formation
`in Escherichia coli Cultures Using
`Dissolved Oxygen Responses to
`Feed Transients
`
`Mats Akesson, 1 Eva Nordberg Karlsson/ Per Hagander, 1
`Jan Peter Axelsson,3 Anita Tocaj2
`
`1 Department of Automatic Control, Lund Institute of Technology,
`P.O. Box 118, SE-221 00 Lund, Sweden; telephone: +46 46 222 97 43;
`fax: +46 46 13 81 18; e-mail: mats.akesson@control.lth.se
`2Department of Biotechnology, Lund University, Lund, Sweden
`3 Pharmacia & Upjohn, Process R&D, Stockholm, Sweden
`
`Received 3 April 1998; accepted 1 February 1999
`
`Abstract: Recombinant protein production in Escherichia
`coli can be significantly reduced by acetate accumula(cid:173)
`tion. It is demonstrated that acetate production can be
`detected on-line with a standard dissolved oxygen sen(cid:173)
`sor by superimposing short pulses to the substrate feed
`rate. Assuming that acetate formation is linked to a re(cid:173)
`spiratory limitation, a model for dissolved oxygen re(cid:173)
`sponses to transients in substrate feed rate is derived.
`The model predicts a clear change in the character of the
`transient response when acetate formation starts. The
`predicted effect was verified in fed-batch cultivations of
`E. coliTOPP1 and E. coli BL21 (DE3), both before and after
`induction of recombinant protein production. It was also
`observed that the critical specific glucose uptake rate, at
`which acetate formation starts, was significantly de(cid:173)
`creased after induction. On-line detection of acetate for(cid:173)
`mation with a standard sensor opens up new possibili(cid:173)
`ties for feedback control of substrate feeding. © 1999 John
`Wiley & Sons, Inc. Biotechnol Bioeng 64: 590-598, 1999.
`Keywords: Escherichia coli; acetate; on-line detection;
`dissolved oxygen; transients; mathematical model
`
`INTRODUCTION
`
`Escherichia coli is one of the most frequently used host
`organisms for recombinant protein production. Fed-batch
`cultivation is a common method to obtain high cell densities
`and thereby high productivity. One of the problems encoun(cid:173)
`tered is the formation of byproducts such as acetate. Accu(cid:173)
`mulation of acetate has been reported to inhibit growth (Luli
`and Strohl, 1990) and to reduce recombinant protein pro(cid:173)
`duction (Bauer eta!., 1990; Bech Jensen and Carlsen, 1990).
`To reduce or avoid acetate formation, a number of substrate
`feeding strategies have been developed; see Lee (1996) and
`Yee and Blanch (1992).
`A typical problem in monitoring and control of microbial
`cultivations is that many important process variables cannot
`
`Correspondence to: M. Akesson
`Contract grant sponsors: Pharmacia & Upjohn; EO-Biotech Programme
`Contract grant number: BIO-CT96-0488
`
`be measured on-line. This has triggered much research and
`development concerning new sensors, see for instance
`(Schiigerl eta!., 1996). Another way of addressing the prob(cid:173)
`lem is to improve and to extend the use of e:dsting sensors
`(Wang eta!., 1977; Stephanopoulos and San, 1984). In this
`paper we will demonstrate how a standard dissolved oxygen
`sensor can be used for on-line detection of undesirable ac(cid:173)
`etate formation. The key idea is to exploit a characteristic
`change in the relation between oxygen uptake and glucose
`uptake at the onset of acetate formation. This change can be
`detected by superimposing short pulses in the glucose feed
`rate. An attractive feature is that no assumptions are re(cid:173)
`quired on parameters like stoichiometric coefficients.
`
`ACETATE PRODUCTION
`
`Formation of acetate, when E. coli is grown under fully
`aerobic conditions, typically occurs at high growth rates
`and/or high glucose uptake rates. The acetate production is
`thought of as an overflow phenomenon where flux of Ace(cid:173)
`tylCoA is directed to acetate, via acetylphosphate, instead of
`entering the TCA cycle. In batch and continuous cultiva(cid:173)
`tions, it was observed that the specific oxygen uptake rate
`reached an apparent maximum at the onset of acetate for(cid:173)
`mation (Andersen and von Meyenburg, 1980; Reiling eta!.,
`1985). It was suggested that the respiratory system, where
`NADH is reoxidized, has a limited capacity. As flux to the
`TCA cycle results in NADH production and as flux to ac(cid:173)
`etate does not, redirection of AcetylCoA flux to acetate
`would be necessary to avoid accumulation of NADH when
`the respiration saturates. Another explanation that has been
`suggested is that the TCA cycle has a limited capacity and
`that this limitation is reached before that of the respiration
`(Majewski and Domach, 1990). When the TCA cycle satu(cid:173)
`rates, increasing glucose uptake will again result in flux
`from AcetylCoA to acetate. In this case, NADH production
`and respiration can increase further until the maximum res-
`
`© 1999 John Wiley & Sons, Inc.
`
`CCC 0006-3592/99/050590-09
`
`BEQ 1035
`Page 2
`
`

`
`piration capacity or the maximum glucose uptake is
`reached. The experiments presented in Paalme et al. (1997),
`however, indicate that a respiratory limitation is more
`likely.
`In Majewski and Domach (1990), a flux network over
`parts of the central metabolic pathways was used to derive
`relations between triose flux and acetate production for the
`two explanations mentioned above. Assuming that the cells
`tend to maximize ATP production, a constrained optimiza(cid:173)
`tion problem was formulated. Acetate production was pre(cid:173)
`dicted when constraints in the respiration or the TCA cycle
`were reached. These ideas were extended in (Ko et al.,
`1993, 1994; Varma et al., 1993) where the flux models
`. cover larger parts of the metabolism and also describe ac(cid:173)
`etate formation due to oxygen limitation.
`
`result from the models in (Ko et al., 1993, 1994; Varma et
`al., 1993).
`The relation between specific glucose uptake, qg, and
`specific oxygen uptake, q0 , can be represented as
`q < crit
`qg
`,
`g
`q >
`crit
`g - qg
`,
`
`with the yield constant Yog- Similarly, the specific growth
`rate, J.L, can be described as
`
`yox
`G = qg xg•
`crityox + ( _ crit) yfe
`{
`J.L(
`)
`qg
`qg
`qg
`xg
`xg•
`
`The specific glucose uptake, qg, is taken to be of Monod
`type
`
`SIMULATION MODEL
`
`We will now derive a model for how dissolved oxygen in a
`bioreactor responds to transients in the feed rate. The pur(cid:173)
`pose is to obtain a good description in a time scale of sec(cid:173)
`onds to minutes. First, considerations on the cell level are
`used to derive relations between glucose uptake, acetate
`production, growth rate, and oxygen consumption. These
`relations are then incorporated into a macroscopic model of
`a bioreactor.
`
`Metabolic Relations
`
`From the analysis in Majewski and Domach (1990), it is
`straightforward to compute also the corresponding NADH
`flux for a respiratory limitation. Assuming that the oxygen
`consumption is proportional to the NADH production and
`that the glucose flux is proportional to the triose flux, rela(cid:173)
`tions between glucose uptake, acetate production, growth
`rate, and oxygen consumption can be obtained. Qualitative
`results are shown in Fig. 1. When the glucose uptake, qg,
`exceeds a critical level, qgcrit, acetate formation starts and
`the oxygen uptake saturates. Concomitantly, there is a de(cid:173)
`creased yield from glucose to cell mass. Similar relations
`
`t
`
`which describes a smoothly saturating glucose uptake.
`
`Bioreactor Model
`
`Assuming that the expressions for oxygen uptake, growth
`rate, and glucose uptake are valid in a time scale of seconds,
`these are inserted in a dynamic model of a bioreactor in
`fed-batch mode. Component-wise mass balances for the
`bioreactor give the following equations:
`
`dV
`-=F
`'
`dt
`
`d(VX)
`----;]{ = J.L( G) · VX,
`
`d(VG)
`-d-=FG;n-qg(G) · VX,
`t
`
`where V, X, G, and C0 are, respectively, the liquid volume,
`the cell concentration, the glucose concentration, and the
`dissolved oxygen concentration. Further, F, Gin• and Co*
`denote the feed rate, the glucose concentration in the feed,
`and the dissolved oxygen concentration in equilibrium with
`the oxygen in gas bubbles. To obtain good mixing in a
`reactor, the stirrer speed, N, is in practice never below a
`minimum value. For the considered range of stiner speeds,
`the volumetric oxygen transfer coefficient, KLa, is approxi(cid:173)
`mated as an increasing linear function of the stirrer speed.
`In practice, most sensors do not measure the oxygen con(cid:173)
`centration but rather the dissolved oxygen tension. The dis(cid:173)
`solved oxygen tension 0 is related to the dissolved oxygen
`concentration through Henry's law
`
`Figure 1. Relations between specific glucose uptake, qg, specific oxygen
`uptake, q0 , specific growth rate, fh, and specific acetate production, qac·
`
`O=H· C0 •
`
`AKESSON ET AL.: E. COLI: ON-LINE DETECTION OF ACETATE FORMATION
`
`591
`
`BEQ 1035
`Page 3
`
`

`
`It is also important to consider the dynamics in the dis(cid:173)
`solved oxygen probe. It is here modeled as a first-order
`system with time constant TP,
`
`~~0,[%]
`
`10o
`
`(a)
`
`2
`
`4
`
`6
`
`8
`
`···············;-:-· __ __,___··~··.········.················:·······_
`:
`:
`(
`
`:
`
`F [L/hJ
`
`0.04
`
`0.02
`0o
`
`2
`
`4
`Time [min)
`
`6
`
`8
`
`~:~:pa:
`···-~·-························:•.·······j
`~Pt?i ' ; l o, [%]
`::c~·· t d _ 1 L [L/h]
`• 3
`. j
`
`which is a reasonable approximation under normal turbu(cid:173)
`lence levels (Dang et al., 1977).
`
`KEY IDEA AND SIMULATIONS
`
`The simulation model is now employed to illustrate that a
`standard dissolved oxygen probe can be used to detect ac(cid:173)
`etate formation. The values of the model parameters that are
`used are found in Table I together with initial values for cell
`mass and volume. To obtain constant specific glucose up(cid:173)
`take rates, exponentially increasing feed rates are used in the
`simulations. This also causes an increased oxygen demand
`and the stirrer speed is therefore increased exponentially to
`avoid trends in dissolved oxygen.
`The key idea in the detection method is to exploit the
`characteristic change in the relation between oxygen uptake
`and glucose uptake at the onset of acetate formation. This
`change, and hence acetate formation, can be detected by
`superimposing short pulses in the glucose feed rate. Under
`glucose-limited conditions the feed pulses give rise to
`changes in the glucose uptake. These changes imply varia(cid:173)
`tions in the oxygen uptake that can be seen in the dissolved
`oxygen measurement.
`Figure 2 shows simulations for increasing specific glu(cid:173)
`cose uptake rates: below, at, and above the onset of acetate
`formation. Below acetate formation, a clear response in dis(cid:173)
`solved oxygen is seen to both up and down pulses in the
`glucose feed rate F, see Fig. 2a. When the onset of acetate
`formation is reached (see Fig. 2b), the response to an up
`pulse in F is absent due to the saturation in the specific
`oxygen uptake. For a glucose uptake above acetate forma(cid:173)
`tion, there is also a reduction in the response to a down
`pulse, Fig. 2c. When the glucose uptake is increased even
`further, the respiration is completely saturated and no oxy(cid:173)
`gen response will be seen.
`It is clear that the pulse responses would reveal if qg is
`above qg cnt, and thus if acetate is produced. The validity of
`the simulation results will now be examined experimentally.
`
`(b)
`
`(c)
`
`10o
`
`2
`
`4
`
`6
`
`8
`
`0 .0 2L
`0o
`
`2
`
`4
`Time [min)
`
`6
`
`8
`
`::~~ ~~~~ 1 Op [%)
`
`20
`···············:················:················:················:·······
`:
`:
`:
`:
`10o
`2
`4
`6
`8
`
`F [1/hJ
`
`4
`Time [min)
`
`6
`
`8
`
`Figure 2. Simulation of responses in measured dissolved oxygen OP to
`pulses in feed rate F. (a) Below onset of acetate formation, qg = 0.8 g/(gh)
`< qgcdt. (b) At the onset of acetate formation, qg = 1.0 g/(gh) < qgccit. (c)
`Above onset of acetate formation, qgcdt < qg = 1.1 g/(gh).
`
`MATERIALS AND METHODS
`
`Table I. Parameter values used in simulations.
`
`Parameter
`
`Value
`
`Parameter
`
`Value
`
`Microorganisms
`
`q;ax
`~~
`k,
`q~rit
`0*
`VX(O)
`
`1.34 g/(gh)
`0.50 g/g
`!Omg!L
`1.0 g/(gh)
`100%
`20 g
`
`Gin
`TP
`V(O)
`
`0.5 g/g
`0.25 gig
`14,000 (L%)/g
`500 giL
`20 s
`2.0 L
`
`Two recombinant E. coli strains with different plasmids
`were used. Experiments without induction of recombinant
`protein were performed with E. coli TOPPl (Stratagene, La
`Jolla, CA) carrying a plasmid pHD389 with a protein L gene
`fragment inserted (Tocaj et al., 1995). The second strain
`employed was E. coli BL2l(DE3) (Studier and Moffatt,
`
`592
`
`BIOTECHNOLOGY AND BIOENGINEERING, VOL. 64, NO.5, SEPTEMBER 5, 1999
`
`BEQ 1035
`Page 4
`
`

`
`1986) with plasmid pBRMX14 derived from the vector
`pET25b (Novagen, Madison, WI) and encoding a xylanase
`(Xynl~N) (Nordberg Karlsson et a!., 1998).
`
`the DO, and the feed was then added according to expo(cid:173)
`nential feed-rate profiles as described below. Expression of
`the recombinant protein in BL21(DE3) was initiated by add(cid:173)
`ing IPTG ( 1 mM).
`
`Bioreactor Equipment and Cultivation Conditions
`
`Feed-Rate Profile
`
`Media Composition
`
`The media for the batch and fed-batch cultivations were
`composed as described in Table II.
`
`The feed rate was varied to obtain different specific growth
`rates 11-set < 11-max· In each interval the feed rate was in(cid:173)
`creased exponentially as
`
`Inoculum
`
`The inoculum was prepared in 100 mL of the media in
`Table II, but without antifoam. It was grown in a baffled
`shake-flask for 10 h, at 30°C (TOPP1) or 37°C (BL21(DE3)),
`on a rotary shaker (GFL 1092, Burgwedel, Germany).
`
`Fed-Batch Cultivations
`
`The cultivations were carried out in a 3-L bioreactor (Che(cid:173)
`moferm FLC-B-3, Hagersten, Sweden) with an initial vol(cid:173)
`ume of 2.0 L. Data logging, dissolved oxygen (DO) control,
`and feed-pump control were implemented using the Satt(cid:173)
`Line control system (Alfa Laval Automation AB, Malmo,
`Sweden). The temperature was 37°C (BL21(DE3)) or 30°C
`(TOPP1), and the pH was kept at 7.0 by titration with 6.7 M
`ammonia. DO was measured with a galvanic electrode cali(cid:173)
`brated to 100% at air saturation at 30 and 37°C, respectively,
`in the cultivation media. The DO was kept at 30% by au(cid:173)
`tomatic control of the stirrer speed, using a PID controller.
`Aeration was set to approximately 1 vvm. Oxygen and car(cid:173)
`bon dioxide content in the exhaust gas stream was recorded
`by a paramagnetic oxygen analyzer (Series 1100, Servo(cid:173)
`mex) and a C02 analyzer (Binos 1.1, Leybold-Heraeus),
`respectively. The feed vessel was placed on a balance for
`registration of:"feed addition. The feed period was started
`upon depletion of the initial substrate, indicated by a peak in
`
`where t0 is the start of the interval. The value of VX(t0 ) was
`precalculated from the total amount added glucose assum(cid:173)
`ing that the biomass/glucose yield Yxg was 0.5 gig. Pulses
`with a duration of 1 min were superimposed to the nominal
`feed rate. The pulse amplitude was ±25% of the nominal
`feed rate.
`
`Analyses
`
`Sampling and Sample Treatment
`
`The samples were withdrawn through a septum at the bot(cid:173)
`tom of the fermenter, using 3-mL Venoject evacuated blood
`collecting tubes (Terumo, Madrid, Spain). Samples for glu(cid:173)
`cose or metabolite determinations were collected in tubes
`containing perchloric acid. These samples, used for enzy(cid:173)
`matic determination of glucose and acetic acid, were cen(cid:173)
`trifuged (15000g) to remove the cell fraction. Supernatants
`were kept frozen (-20°C), and prior to analysis the samples
`were thawed, heated (80°C, 15 min), and centrifuged
`(15000g).
`
`Optical Density (OD)
`
`The OD was measured at 620 nm. Samples were diluted
`with 0.9% NaCl at OD values exceeding 0.5.
`
`Acetic Acid
`
`Acetic acid concentrations were determined enzymatically
`using a test kit (no. 148261, Boehringer Mannheim).
`
`Feed
`
`500
`
`Table II. Media and feed composition.a
`
`Media component
`
`TOPP1
`
`BL21(DE3)
`
`Glucose (giL)
`(NH4)zS04 (giL)
`KH2PO 4 (giL)
`K2HP04 (giL)
`Na2HP04 ·(2H20) (g/L)
`NaH2P04 ·(H20) (giL)
`(NH4)z-H-citrate (giL)
`1 M MgS04 (mLIL)
`Trace element (mLIL)
`Ampicillin (giL)
`Adekanol (mLIL)
`
`10
`2.0
`1.6
`
`6.6
`
`0.5
`2.0
`2.0
`0.1
`0.05
`
`10
`2.0
`
`14.6
`
`3.6
`0.5
`2.0
`2.0
`0.1
`0.05
`
`50
`10
`
`Glucose
`
`aThe trace element solution used was that of Holme et al. (1970). An(cid:173)
`tifoam (Adekanol (LG-109), Asahi Denka Kogyo K.K., Japan) was also
`added when needed (0.1 mL).
`
`Glucose concentration was determined enzymatically using
`a test kit (cat. no. 716251, Boehringer Mannheim), accord(cid:173)
`ing to the analytical procedure described by Larsson and
`Tornkvist (1996).
`
`AKESSON ET AL.: E. COLI: ON-LINE DETECTION OF ACETATE FORMATION
`
`593
`
`BEQ 1035
`Page 5
`
`

`
`Op [%]
`
`G
`
`X [g/LJ
`
`~···iMJ··J.:·J···.· i·····_i: .... ~··}~··L tJ
`::~
`[mg/L] '] 1 J~jv J 1vr JJ
`~:.~~ ,::1 + oL~L }·.··r ·;~·o]
`c c I c.....__L_O~ c ~I ,· c r "~" I c I
`::!l___..l...__~ "-'-------'---~
`
`8.5
`
`9
`
`9.5
`
`10
`Time [h]
`
`10.5
`
`11
`
`11.5
`
`12
`
`Xylanase Activity
`
`Samples were taken at 30-min intervals during the induction
`phase. The xylanase activity was determined according to
`the procedure in Nordberg Karlsson eta!. (1998).
`
`Cell Dry Weight Calculations (cdw)
`
`For E. coli TOPPl, the cdw was calculated using a standard
`curve (cdw vs OD) obtained from several batch cultivations
`without induction. One OD unit was equal to 0.44 g cdw/L.
`For E. coli BL2l(DE3), the cdw was determined by cen(cid:173)
`trifuging (15000g) quadruple samples (1 mL) in preweighed
`Eppendorf tubes. The pellets were washed with 1 mL of
`0.9% NaCl and dried overnight (105°C). One OD unit was
`equal to 0.52 g cdw/L.
`
`RESULTS
`
`To verify the results from the simulations, cultivations were
`made with two recombinant strains: E. coli TOPPl and E.
`coli BL2l(DE3). Different specific glucose uptake rates
`were obtained by changing the feed rate between different
`exponential profiles. In this way, periods with and without
`acetate production could be achieved. For each exponential
`feeding period, pulse experiments were performed at well(cid:173)
`controlled oxygen levels. During the pulse experiments, the
`dissolved oxygen controller was set in manual mode and the
`stirrer speed was kept constant. This gave a decreasing trend
`in dissolved oxygen due to the increased oxygen demand
`when the feed rate F increased exponentially; see for ex(cid:173)
`ample the pulse experiments at 10 and 10.4 h in Fig. 3. In
`the experiments with E. coli BL2l(DE3) this effect was
`reduced by increasing the stirrer speed in accordance with
`the feed rate. To facilitate the comparison with the simula(cid:173)
`tions, where perfect compensation of the oxygen demand
`can be achieved, a linear trend was removed from the dis(cid:173)
`solved oxygen data. The final value is thus reset to 30% in
`the compensated data, which are shown together with the
`original measurements.
`
`Figure 3. Fed-batch part from an experiment with E. coli TOPPI. From
`top to bottom: glucose feed rate F, stirrer speed N, dissolved oxygen
`tension OP' glucose concentration G, acetate concentration HAc, and cell
`concentration X.
`
`Figure 4b shows an experiment made during acetate pro(cid:173)
`duction around 10 h. The specific glucose uptake rate is
`rather high, qg ~ 1.26 g/(gh), but the glucose concentration
`was still low (~10 mg/L). As was predicted by the simula(cid:173)
`tions, the response to the down pulse is now slightly reduced
`and no reaction is seen to the up pulse.
`
`E. coli TOPP1
`
`E. coli Bl21(DE3)
`
`The fed-batch part from a cultivation of E. coli TOPPl is
`shown in Fig. 3. Acetate produced in the initial batch phase
`was consumed during the first part of the fed-batch phase.
`The resulting glucose concentration was below 50 mg!L
`except during the period with the highest fLset when it tem(cid:173)
`porarily increased to about 350 mg!L.
`
`Pulse Experiments
`
`A pulse experiment performed at 9.1 h for a moderate glu(cid:173)
`cose uptake rate, qg ~ 0.66 g/(gh), is shown in Fig. 4a. No
`acetate production could be detected, and a clear response in
`dissolved oxygen is seen to both the up and the down pulses.
`
`The fed-batch part of a cultivation with E. coli BL2l(DE3)
`is shown in Fig. 5. The acetate concentration after the batch
`phase was negligible, and the concentrations of acetate as
`well as glucose were low throughout the fed-batch phase.
`Before the induction phase, acetate production could be
`detected for a specific glucose uptake rate around 1.22
`g/(gh) at time 7.8 h.
`Prior to induction, f.Lset was decreased to 0.25 h- 1
`• Ex(cid:173)
`pression of the foreign protein was induced after 8.6 h, and
`a reduction of the specific growth rate down to 0.11 h- 1 was
`observed, see Fig. 6. While fLset was still kept at 0.25 h-1
`,
`the specific glucose uptake rate qg increased from 0.50 to
`0.63 g/(gh). During the same period, it was also observed
`
`594
`
`BIOTECHNOLOGY AND BIOENGINEERING, VOL. 64, NO.5, SEPTEMBER 5, 1999
`
`BEQ 1035
`Page 6
`
`

`
`OP [%]
`
`F [L/h]
`
`F]L/h]o::~~
`. ti~····
`
`. .. ·•············•········ .. :.....
`1000 ........ ······ ... :..
`•
`•
`:
`•
`N [rpm]
`:
`•
`5oo:'~.f' .... : .. t~
`
`.
`
`Op [%]
`
`.
`.
`40 ·········•···· ... ;
`.
`
`.
`.
`. ............... ···•··········
`.
`:
`:
`:
`:
`:
`20~
`
`.
`:
`
`...
`
`.
`
`.
`
`::~···············;·····················································=···
`10o
`0.03 ··············· ; ................. • ................. :·················:········
`0.04r:l.
`I
`.
`.
`.
`:
`:
`0.02 ................ ; ................. · ..... ! \ ..•..................•........
`r
`r
`0.0:~! u
`
`0
`
`2
`
`
`4
`Time [min]
`
`::
`
`(a)
`
`:
`
`2
`
`:
`
`4
`
`:
`
`6
`
`:
`
`8
`
`j
`
`11
`6
`8
`
`. 1
`
`'
`
`20====~·. =
`00~1~
`:: ······· ~>T'<-.~····················· I •••••••••••••••••••••••.••
`10o
`(b) ,::t;!
`' m~~
`! 1 F ~jh]
`
`~
`
`2
`
`~
`
`4
`
`..... ~-·--·-·-·-('-
`
`6
`
`8
`
`0
`
`2
`
`4
`Time [min]
`
`6
`
`8
`
`Figure 4. Experiments with E. coli TOPP 1. Response in dissolved oxy(cid:173)
`gen OP to pulses in feed rate F. Original data (dashed) and data with linear
`trend removed (solid). (a) No acetate formation: qg ~ 0.66 g/(gh), X = 6
`giL. (b) During acetate formation: qg ~ 1.26 g/(gh), X = 9 giL.
`
`that the ratios between oxygen uptake and carbon dioxide
`production to the glucose uptake q0 /qg and qcJqg increased
`(data not shown). At approximately 11 h, a 4-fold increase
`in the acetate concentration was detected, indicating that
`acetate had been produced. In the final part of the cultiva(cid:173)
`tion, f.Lset was set to 0.15 h- 1
`. The xylanase activity obtained,
`see Fig. 5, shows that a high production level was achieved.
`
`Pulse Experiments
`Pre-induction pulse experiments gave responses similar to
`those obtained with TOPPl. Three consecutive pulse ex(cid:173)
`periments after induction, from 9.5 to 10.2 h, are shown in
`Fig. 7. In the first experiment, see Fig. 7a, qg was about 0.54
`g/(gh) and no acetate production could be observed. Clear
`responses to both up and down pulses were obtained. When
`qg increased to 0.59 g/(gh), the response to an up pulse was
`broadened and the amplitude was somewhat reduced, Fig.
`7b. Still no accumulation of acetate was detected. Figure 7c
`shows the third experiment where qg was approximately
`0.63 g/(gh). Almost no response to an up pulse resulted, and
`shortly after this experiment an increased acetate concen(cid:173)
`tration was detected.
`
`DISCUSSION
`From the experimental results, it is clear that the dissolved
`oxygen response to transients in the glucose feed rate
`
`G
`[mg/L]
`
`HAc
`[mg/LJ
`
`X [g/L]
`
`'~JJ~J.;~;;~;~~+ vj
`·:l····l OL:t~~L; h+··~·l
`t ~ ~~I~ ~I~ 0 r ~ t ~ r a l
`1---:'-:---~ t ~~~ ~ l
`~~:.~L;~r'----'------'--. -L . . - . . . lL -L -1
`
`6
`
`7
`
`.
`
`8
`Time [h]
`
`9
`
`10
`
`11
`
`12
`
`Figm·e 5. Example of fed-batch part of an experiment witb E. coli
`BL2l(DE3). From top to bottom: glucose feed rate F, stirrer speed N,
`dissolved oxygen tension OP, glucose concentration G, acetate concentra(cid:173)
`tion HAc, cell concentration X, and recombinant protein detected as
`xylanase activity Xyn.
`
`change character with the onset of acetate formation. This
`supports the idea of a change in the relation between glu(cid:173)
`cose and oxygen uptake at the onset of acetate formation.
`The results also indicate that this change takes place in a
`time scale of seconds or faster.
`An important consequence is that it is feasible to detect
`acetate formation from the dissolved oxygen response to
`superimposed transients in the glucose feed rate. This opens
`up new possibilities for feedback control of substrate feed(cid:173)
`ing. The fact that it is a qualitative change that is detected
`reduces the need for a priori knowledge, which is beneficial
`from a control perspective. Furthermore, as it is the forma(cid:173)
`tion of acetate that is detected, it is possible to avoid accu(cid:173)
`mulation of acetate at an early stage. To avoid the difficulty
`with the trends in dissolved oxygen, it is advantageous to
`perform the evaluation of the oxygen response during the
`feed pulse. The time between consecutive pulses can then be
`used to control dissolved oxygen using the stiner speed.
`This technique has been successfully applied in a novel
`
`AKESSON ET AL.: E. COLI: ON-LINE DETECTION OF ACETATE FORMATION
`
`595
`
`BEQ 1035
`Page 7
`
`

`
`w~ 40 ........ '' ...
`·:·"":.'>; ~ .......... ·~·· '' ........ ······i···' '' ... ' ........ : ....... .
`30
`>~~~~~::",·····•··········
`.. • ..
`.... -- ..... ~ .......
`20~
`10o
`
`\
`
`:
`
`6
`
`8
`
`:
`
`2
`
`:
`
`4
`
`::::NT ························s······· ... ···················j·········i
`0.04 ............... "" .................................................. .
`
`0.02
`
`' .
`
`:
`
`:
`
`:
`
`. .... )..........
`. ....
`:
`:
`;
`QL---~---L--~~--~~
`0
`2
`4
`6
`8
`Time [min]
`
`F [L/h]
`
`Op [%]
`
`F [L/h]
`
`1o
`
`2
`
`4
`
`6
`
`8
`
`0o
`
`2
`
`4
`Time [min]
`
`6
`
`8
`
`'
`40 ............. '
`.
`.
`~·· .
`30 " " . H :~~~~>_.~_,M_~--
`Op [%]
`20===·······f···········+··········· .. i·······
`1
`10o
`
`1
`
`4
`
`2
`
`6
`
`8
`
`50~:~ I' I
`40~·············,····i·······~·~·······•········
`~Ef?'-t:-y-T-
`~•~EJ•••••••••·•••~••••=r•••• ,!••••••1
`sopt:· .
`::u•• .••••...•• ; ••.••• 3 ...••••••.••.• ; •••...
`
`(a)
`
`(b)
`
`(c)
`
`4.5 ,---.,...-----,----.--,--,--,...--.---,
`
`4 "" ..... · ......... ; .. ..
`
`ln(VGr) -0.7
`(V)
`
`3.5
`
`3
`
`lnOD
`(D)
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`Time [h]
`
`Figure 6. Logarithms of total amount fed glucose VGr (''7) and optical
`density (D) in a cultivation of E. coli BL2l(DE3). At 7.8 h, f.1,et is lowered
`to 0.25 h- 1 followed by induc