Prévia do material em texto
Experimental Investigation and Modeling of Oscillatory Behavior in the Continuous Culture of Zymomonas mobilis Andrew J. Daugulis, P. James McLellan, Jinghong Li Department of Chemical Engineering, Queen’s University, Kingston, Ontario, Canada K7L 3N6; telephone: (613) 545-2784; fax: (613) 545-6637; e-mail: daugulis@qucdn.queensu.ca Received 16 August 1996; accepted 7 February 1997 Abstract: The mechanism causing oscillation in continu- ous ethanol fermentation by Zymomonas mobilis under certain operating conditions has been examined. A new term, ‘‘dynamic specific growth rate,’’ which considers inhibitory culture conditions in the recent past affecting subsequent cell behavior, is proposed in this article. Based on this concept, a model was formulated to simu- late the oscillatory behavior in continuous fermentation of Zymomonas mobilis. Forced oscillation fermentation experiments, in which exogenous ethanol was added at a controlled rate to generate oscillatory behavior, were performed in order to obtain estimates for the model parameters and to validate the proposed model. In addi- tion, data from a literature example of a sustained oscil- lation were analyzed by means of the model, and excel- lent agreement between the model simulation and ex- perimental results was obtained. The lag in the cells’ response to a changing environment, i.e., ethanol con- centration change rate experienced by the cells, was shown to be the major factor contributing to the oscilla- tory behavior in continuous fermentation of Zymomonas mobilis under certain operating conditions. © 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 56: 99–105, 1997. Keywords: oscillatory behavior; Zymomonas mobilis; fermentation; ethanol INTRODUCTION During the past 15 years,Zymomonas mobilishas been promoted as a more promising microorganism than yeast for the industrial production of ethanol.Zymomonas mobilis possesses higher ethanol tolerance and superior kinetic and yield characteristics compared to those of yeast (Beavan et al., 1982). More specifically, it exhibits higher specific rates of substrate consumption, ethanol production, and a higher ethanol yield (Rogers et al., 1982). One drawback associated with the continuous fermenta- tion of Zymomonas mobilisis the peculiar occurrence of oscillation or cycling of substrate, biomass, and product concentrations under certain fermentation conditions (Bruce et al., 1991; Ghommidh et al., 1989; Jobses et al., 1986; Lee et al., 1980; Lee et al., 1979). Oscillatory behavior means that there are periods of time during which there is de- creased ethanol productivity, and a high level of residual substrate resulting in unacceptably high substrate losses. Based on previous work (Li et al., 1995; Li, 1996), the concept of adynamic specific growth rateis proposed in this article. The dynamic specific growth rate refers to a situation in which the cells undergo a growth rate which is less than or equal to a growth rate characterized by the instantaneous culture conditions, and reflects (potentially deleterious) conditions experienced by the cells some time in the past. Experimental evidence is provided by direct estimation of dynamic specific growth rates from experi- mental data and comparisons made with the instantaneous specific growth rate at the culture conditions. With this concept, a model is formulated to describe the oscillatory behavior of continuous ethanol fermentation byZymomonas mobilis. Forced oscillation fermentation experiments were performed to validate the proposed model and a literature example of oscillation in the continuous fermentation of ethanol was used to test the model. Model Development Several earlier models have been proposed to account for the oscillatory behavior ofZ. mobilis.One model (Ghom- midh et al., 1989) later expanded to include substrate limi- tation and product inhibition (Jarzebski, 1992), conceptual- ized the cell population as consisting of three states: viable cells, non-viable cells (able to produce ethanol but not to reproduce), and dead cells. With appropriate mass balance equations, the model was able to represent the data obtained by their group during oscillatory culture ofZ. mobilis.Some experimental support for the presence of viable and non- viable, or dead cells was provided by plate count and slide culture estimations, although these assays are time-con- suming. Another model (Jobses et al., 1986), a structured two-compartment representation, considered biomass as be- ing divided into compartments containing specific group- ings of macromolecules (e.g., protein, DNA, and lipids). A common feature of the above models is that they are based upon some physiological quantities which are often Correspondence to:Andrew J. Daugulis Contract grant sponsor: Natural Sciences and Engineering Research Council of Canada © 1997 John Wiley & Sons, Inc. CCC 0006-3592/97/010099-7 difficult or time consuming to monitor during the fermen- tation process. Our approach has been more phenomeno- logical in that macroscopic variables such as ethanol, sub- strate, and biomass concentrations are the only experimental quantities which need to be measured. This is not to suggest that sub-populations do not exist withinZ. mobiliscultures, particularly in stress situations; however, such stress re- sponses may be more easily detected through the presence of filamentation, which has been previously linked to stress situations experienced by the cells (Daugulis et al., 1985; Fein et al., 1984), including exposure to high ethanol con- centrations in a short period of time (Ghommidh et al., 1989). Previous work in our laboratory (Li et al., 1995) with Zymomonas mobilishas shown that the inhibition effect of ethanol concentration history is insignificant, while the in- hibition effect of an upward ethanol concentration change rate is quite intense. Moreover, recognizing that cells are unable to respond instantaneously to changes in their envi- ronment, it was postulated that a delay exists between the time that the cells experience a change, and their metabolic response. The time delay inhibition effect was modeled as the effect of the weighted average of ethanol concentration change rate,Z: Z~t! = *−` t dP/dt|t · cz~t!dt (1) In Equation (1), the weight for the ethanol concentration change rate historycz(t) is: cz(t) 4 b 2 · (t − t) · e−b·(t−t) (2) wheret is the time in history andt is the current time. The maximum ofcz(t) occurs att 4 t − 1/b, i.e., 1/b h prior to the current timet. In other words, the ethanol concentration change rate which happened at 1/b hours ago has the most significant influence on current cell performance. The pa- rameterb indicates the magnitude of the time lag for the delayed inhibition effect. As noted, this time delay effect recognizes that cells do not respond instantaneously to changes in their environment, but require a period of time to produce a metabolic response. With Z(t) being defined in Equation (1), it is hypothesized that the growth of cells at timet is inhibited when the weighted average of ethanol concentration change rateZ(t) is positive, while a negativeZ(t) has no effect on the cell growth. Obviously, such a hypothesis about the effect of Z(t) leads to a piece-wise function, withZ(t) 4 0 being the switching point. Usually, a piece-wise continuous function is not desired in parameter estimation procedures since it is not suitably differentiable. In order to facilitate the param- eter estimation procedure, a continuously differentiable hy- perbolic function,fm (Z(t)), is introduced to approximate the effect of Z(t) on the cell growth: fm ~Z~t!! = 1 2 S1 − el · Z~t! − d − e−l · Z~t! + del · Z~t! − d + e−l · Z~t! + dD (3) l andd are parameters to be estimated from the experimen- tal data which represent the intensity of the inhibition effect of the weighted average of ethanol concentration change rateZ(t). The functionfm (Z(t)) possesses the properties: fm (Z(t)) ∈ (0,1) for Z(t) ∈ (−`, +`) (4) fm isthe inhibition factor, and is used to define the dynamic specific growth ratem(S(t), P(t), Z(t)) in terms of the con- ventional or ‘‘instantaneous’’ specific growth ratem(S(t), P(t)): m(S(t),P(t),Z(t)) 4 m(S(t),P(t) z fm(Z(t)) (5) The instantaneous specific growth ratem(S(t), P(t)) refers to the traditional or classic representation of specific growth rate well established either in textbooks (Bailey and Ollis, 1986) or the literature (Aiba et al., 1968).m(S(t), P(t)) is instantaneous in the sense that it is strictly a function of the current values ofS(t) andP(t). The dynamic specific growth ratem(S(t), P(t), Z(t)) is proposed to include the inhibitory effect ofZ(t), the weighted average of ethanol concentration change rate history experienced by cells. The dynamic na- ture of this specific growth rate arises from the fact thatm depends onZ, and is associated with the evolution of the ethanol concentration change rate experienced by the cells. Becausefm lies between 0 and 1, Equation (5) states that the dynamic specific growth ratem(S(t), P(t), Z(t)) assumes a value in the range between 0 andm(S(t), P(t)), depending upon the value offm (Z(t)). When the weighted average of ethanol concentration change rate is negative or zero, cor- responding to a sustained downward ethanol concentration change rate trend or a trend of constant ethanol concentra- tion, the dynamic specific growth ratem(S(t), P(t), Z(t)) will reach a value identical to the instantaneous specific growth rate m(S(t), P(t)). When an upward ethanol concentration change rate trend occurs (Z(t) > 0), the dynamic specific growth ratem(S(t), P(t), Z(t)) will only be able to reach a fraction fm (Z(t)) of the instantaneous specific growth rate m(S(t), P(t)). For convenience and simplicity in reporting our results, the variable namesS, PandZ will be used to representS(t), P(t) andZ(t) in the following sections. The instantaneous specific growth ratem(S,P) in Equa- tion (5) is modeled using the product and substrate inhibi- tion formula of Veeramallu and Agrawal (1990): m~S,P! = mmaxSS1 − S PPmaD aDS1 − S P − PobPmb − PobD bD Ks + S+ S~S− Si! Ki − Si (6) where P − Pob Pmb − Pob = 0 if P ø Pob S− Si = 0 if Sø Si P − Pob Pmb − Pob = 1 if P . Pmb 100 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 56, NO. 1, OCTOBER 5, 1997 The specific production rate Qp is modeled using the litera- ture formula (Lee et al., 1983; Luong, 1985), Qp = QPmax · S Kmp + S · (1 − ( P Pme )a) (7) In order to be able to undertake modeling for both batch and continuous fermentation, the complete differential equation model for the cutlure ofZ. mobiliswas formulated as: dX dt = ~m~S,P,Z! − D − s~t! · De! · X (8) dS dt = −1/YP/S · Qp · X + D · Sf − ~s~t! · De + D! · S (9) dP dt = Qp · X + s~t! · De · Pe − ~s~t! · De + D! · P (10) dZ dt = b · ~W − Z! (11) dW dt = b · ~Qp · X + s~t! · De · Pe − ~s~t! · De + D! · P − W! (12) Note that this model also has provision to account for ex- ternally added ethanol, in order to be able to model the forced oscillation experiments described later. The variable s(t) is either 0 or 1 depending on whether exogenous etha- nol is being added (i.e., whether the ethanol feed pump is off or on). This ethanol can be added at a dilution rateDe and at a concentrationPe. The first three equations in the model [eqs. (8)–(10)] are established from mass-balance considerations in the fer- mentor. Equations (11) and (12) result from differentiating Z(t) [eq. (1)] with respect tot, whereW(t) is an intermediate variable which is also a weighted average of the previous ethanol concentration change rate with weightcw(t), W~t! = *−` t dP/dt|t · cw~t!dt (13) cw~t! = b · e −b · ~t − t! (14) As will be seen, the above model [eqs. (1)–(14)] has been used to simulate batch and forced oscillation fermentation operations. It has also been used to simulate a literature example of a sustained oscillation fermentation. The parameters in the proposed model have been ob- tained from several sources. The intrinsic kinetic parameters for Z. mobilis were obtained from the literature, and are shown in Table I. The operating parameters during continu- ous and forced oscillation fermentation for substrate feed (Sf), dilution rate (D), ethanol dilution rate (De) and exog- enous ethanol concentration (Pe) were fixed at 20 g/L, 0.084 h−1, 0.015 h−1, and 75% (v/v), respectively. The ethanol yield was determined experimentally to be 0.495 g/g for conventional continuous fermentation and 0.46 g/g for the forced oscillation fermentations. The remaining parameters (b, l, d, a, b, Qpmax, and a) were obtained by multi- response nonlinear regression estimation using the experi- mental data (Li, 1996). Additional insight supporting the concept of dynamic specific growth ratem(S, P, Z) can be obtained by estimat- ing this quantity directly from the experimental data ob- tained from batch, conventional continuous, and forced os- cillation experiments. Using finite differences to represent dX/dt,the dynamic specific growth ratem(S, P, Z) at timeti can be calculated directly from the experimental data using Equation (8) as follows: m~Si , Pi , Zi! = SXi − Xi−1ti − ti−1 D Xi + D + s~t! · De (15) where, for batch fermentationD 4 0, De 4 0; for conven- tional continuous fermentation or if the ethanol feeding pump is off during forced oscillation fermentationDe 4 0. The subscripti refers to theith sample in the experimental data. Because all the effects ofS, P,andZ on the fermen- tation should be reflected in the experimental data which are used directly in Equation (15), them(Si , Pi , Zi) calculated by this equation provides information about the dynamic spe- cific growth rate as a function of these quantities. If there was no memory effect the estimated values should match the instantaneousm(S,P) calculated atSi , Pi within a rea- sonable range of experimental error. However, if the esti- mated specific growth rates deviated significantly from the instantaneous values, this would support the concept of a dynamic specific growth rate. Furthermore, the estimates of m are not dependent on any of the parameters described in the preceding paragraph. MATERIALS AND METHODS The microorganism, the medium composition, and the ex- perimental procedure were described in previous work (Li et al., 1995). Beginning with batch and conventional con- tinuous fermentation, forced oscillation fermentation ex- periments were subsequently conducted during which an oscillatory ethanol concentration trajectory was created by a pump turning on and off which pumped ethanol into the Table I. Literature values of parameters used in the modeling of oscil- latory behavior. Model parameter Veeramallu and Agrawal (1990) Lee and Rogers (1983) mmaxh −1 0.41 Pob (g/L) 50.0 Pma (g/L) 217.0 Pmb (g/L) 108.0 Pme(g/L) 127.0 Ks (g/L) 0.5 Ki (g/L) 200.0 Si (g/L) 80.0 Kmp (g/L) 0.5 DAUGULIS, MCLELLAN, AND LI: OSCILLATORY BEHAVIOR OF ZYMOMONAS MOBILIS 101 fermentor. Two such sequential fermentations (batch, fol- lowed by continuous, followed by forced oscillation) were conducted, one with an ethanol oscillation period of 40 h, and the other with a period of 50 h. The feeding substrate concentration was intentionally kept at a low level, 20 g/L glucose, to ensure that the ethanol converted from glucose would not affect the imposed ethanol concentration trajec- tory significantly. With this imposed oscillatory ethanol concentration trajectory, the fermentation was monitored by measuring the ethanol, biomass, and residual glucose con- centrations. RESULTS Forced Oscillation Fermentation Two forced oscillation experiments were undertaken, with the first being conducted in order to impose ethanol oscil- lations on the fermentation system from which estimates for the model parameters could be obtained. With these esti- mates, the model was used to simulate the data obtained from the second forced oscillation experiment (with a dif- ferent ethanol oscillation period) to help determine how wellthe model predictions provided a fit to the experimental data. In both cases batch fermentation was followed by conventional continuous fermentation to obtain a stable steady state situation before the initiation of the forced os- cillations. Figure 1 shows the results of the first forced oscillation experiment (40 h oscillation period), along with the model predictions for this run. As can be seen, the imposed ethanol oscillations caused corresponding oscillations in the cell and residual substrate concentrations. At time zero of the simu- lation (the onset of the batch fermentation), the weighted average of ethanol concentration change rates was assumed zero, i.e.,Z(0) 4 0.0,W(0) 4 0.0 which was a reasonable assumption provided that the inoculum was in the stationary phase before being added to the fermentor. As might be expected, the model simulations closely matched the experi- mental data for this run. With model parameters estimated from all phases of the experiment, simulations were run in order to examine the ability of the model to fit the data for the second experiment (50 h oscillation period), as shown in Figure 2. Similar responses were obtained for the cell and substrate concen- trations and, as can be seen, the model was able to predict all phases of the experiment exceedingly well. The dynamic specific growth ratem(S, P, Z) calculated from the experimental data obtained during the first forced oscillation fermentation experiment using Equation (15) is plotted against the ethanol concentration in Figure 3, to- gether with the instantaneous specific growth rate curve calculated by Equation (6) using the necessary estimated parameters (i.e., a4 0.3332, b4 1.14), and the necessary parameters from the literature as given in Table I. It can be seen that the instantaneous specific growth rate curve pro- vides an upper boundary for the dynamic specific growth rates as expected. The dynamic specific growth ratesm(S, P, Z) are found in an area either adjacent to or below the instantaneous specific growth rate curve depending on the value of the weighted average of ethanol concentration change rateZ. When the weighted average of ethanol con- centration change rateZ is positive, the dynamic specific growth rate is less than the instantaneous specific growth rate by a factor defined by the functionfm(Z) [eq. (3)–(5)] and it falls in an area below the instantaneous specific growth rate curve. When the weighted average of ethanol concentration change rateZ is zero or negative,fm(Z) ap- proaches 1.0, and consequently the dynamic specific growth rate approaches the instantaneous specific growth rate. As a result, the dynamic specific growth rate falls in an area adjacent to the instantaneous specific growth rate curve. Figure 3 provides a graphical illustration of the distinction between the instantaneous specific growth rate and the dy- namic specific growth rate as determined experimentally. The difference is ascribed to the earlier ethanol concentra- tion change rate experienced by the cells. Figure 1. (a) Ethanol and glucose concentrations for forced oscillations having an imposed oscillation period of 40 h. For the model predictions, the model parameters were estimated to be:b 4 0.1732,l 4 0.4557,d 4 0.5559,a 4 5.498,a 4 0.3332,b 4 1.14. (b) Ethanol and biomass concentrations for forced oscillations having an imposed oscillation period of 40 h. For the model predictions, the model parameters were the same as for (a). 102 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 56, NO. 1, OCTOBER 5, 1997 Modeling Literature Data The literature data published by Ghommidh et al., 1989, were examined to see if they supported the hypotheses and model developed in this paper. The operating conditions in their experiment wereSf 4 200 g/l andD 4 0.05 h −1, with a different strain ofZ. mobilis,NRRC B-14022, having been used. Since the residual glucose concentrations were not reported in their work, an overall product yield coefficient Yp/s was assumed (0.495), which is the same value as was obtained in our experimental work. The estimated param- eters and the model simulation for their experimental data are shown in Figure 4. As can be seen, good agreement between the simulation and their experimental data is ob- tained. As was done in the previous experimental analysis,m(S, P, Z) andm(S, P) were estimated from the data of Ghom- midh et al., and are plotted against ethanol concentration as shown in Figure 5. The characteristics of this plot are con- sistent with those seen in Figure 3, and support the proposed concept of dynamic specific growth rate. CONCLUSION It had been previously demonstrated (Li et al., 1995) that the effect of ethanol concentration history on the fermentative capability ofZ. mobilisis negligible while the upward etha- nol concentration change rate has a significant inhibitory effect on this organism. In the current work, these experi- mental observations were incorporated into a model for simulating/predicting dynamic behavior in continuous fer- mentations ofZ. mobilis. The concept of a dynamic specific growth ratem(S, P, Z), incorporating the effect of inhibitory conditions experienced at an earlier time by the cells, has been proposed, and a distinction made between this term and the traditional specific growth rate, here called the in- stantaneous specific growth ratem(S, P). A model consist- ing of ordinary differential equations was formulated, in- corporating the newly proposed concept of dynamic specific growth ratem(S, P, Z). Forced oscillation fermentations were performed in order to obtain experimental data of oscillatory behavior, to esti- mate model parameters, and to test the proposed model. It was demonstrated that the proposed model is capable of fitting the experimental data well, and of predicting the dynamic behavior of experiments conducted under different operating conditions. The model was further tested by simu- lating a literature example of another sustained oscillation Figure 3. Instantaneous specific growth rate [from eq. (6)], and estimated dynamic specific growth rates [from experimental data and eq. (15)] for the forced oscillation experiment having an oscillation period of 40 h. Figure 4. Ethanol and biomass concentrations (from the data of Ghom- midh et al., 1989), and our model predictions. Model parameters were estimated to be:b 4 0.03178,l 4 7.586,d 4 0.515,a 4 11.25,a 4 0.2038,b 4 1.715. Figure 2. (a) Ethanol and glucose concentrations for forced oscillations having an imposed oscillation period of 50 h. For the model predictions, the model parameters were estimated to be:b 4 0.1732,l 4 0.4557,d 4 0.5559,a 4 5.498,a 4 0.3332,b 4 1.14. (b) Ethanol and biomass concentrations for forced oscillations having an imposed oscillation period of 50 h. For the model predictions, the model parameters were the same as for (a). DAUGULIS, MCLELLAN, AND LI: OSCILLATORY BEHAVIOR OF ZYMOMONAS MOBILIS 103 fermentation ofZ. mobilis,and good agreement between the simulation and the experimental data was obtained. The proposed concept of a dynamic specific growth rate was illustrated by plotting the experimentally obtained dynamic specific growth rate, together with the estimated instanta- neous specific growth rate against ethanol concentration. The instantaneous specific growth rate curvem(S, P) pro- vides an upper boundary for the dynamic specific growth rate m(S, P, Z), with m(S, P, Z) falling in an area either below or adjacent to them(S, P) curve depending on the value ofZ. Recent work has focused on generating overdamped con- tinuous, damped oscillation and sustained oscillation fer- mentation data in our laboratory. The validity of the model proposed in this paper will now be tested against these additional data, which represent the spectrum of modes of continuous operation possible with this organism. The stress response of the cells (i.e., filamentation) to high ethanol concentration change rates is also being investigated, along with the relationshipof these morphological changes tom andQp values. We are grateful to the School of Graduate Studies and Research, Queen’s University, for student support for one of us (JHL). NOMENCLATURE a instantaneous specific growth rate parameter b instantaneous specific growth rate parameter D dilution rate (h−1) De exogenous ethanol dilution rate (h −1) fm inhibition factor for dynamic specific growth rate Ki substrate inhibition parameter for specific growth rate (g/L) Ks substrate saturation parameter for specific growth rate (g/L) Kmp substrate saturation parameter for ethanol production rate (g/L) P ethanol concentration (g/L) Pe exogenous ethanol concentration (g/L) Pma ethanol inhibition threshold for instantaneous specific growth rate (g/L) Pmb maximum ethanol inhibition threshold for cell growth (g/L) Pme maximum ethanol inhibition threshold for ethanol production (g/L) Pob ethanol inhibition threshold for instantaneous specific growth rate (g/L) Qp specific ethanol production rate (g/g-h) Qpmax specific ethanol production rate (g/g-h) S substrate (glucose) concentration (g/L) Si lower threshold for substrate inhibition of specific growth rate (g/L) t time (h) X biomass concentration (g/L) W first-order weighted average ethanol concentration change rate (g/L-h) Yp/s yield coefficient of ethanol vs. glucose (g/g) Z second-order weighted average ethanol concentration change rate (g/L-h) Greek Symbols a ethanol inhibition exponent for ethanol production rate b weighted history parameter for ethanol concentration change rate d ethanol concentration change rate inhibition factor parameter l ethanol concentration change rate inhibition factor parameter s variable in equations 8, 9, 10, 12 and 15 to account for exogenous ethanol addition cw weighted average of previous ethanol concentration change rate (h−1) m(S,P) instantaneous specific growth rate (h−1) m(S,P,Z) dynamic specific growth rate (h−1) mmax instantaneous specific growth rate at zero instantaneous etha- nol concentration (h−1) References Aiba, S., Shoda, M., Nagatani, M. 1968. Kinetics of product inhibition in alcohol fermentation. Biotechnol. Bioeng.10: 845–864. Bailey, J. E., Ollis, D. F. 1986. Biochemical engineering fundamentals, 2nd edition. McGraw-Hill, New York. Beavan, M. J., Charpentier, C., Rose, A. H. 1982. Production and tolerance of ethanol in relation to phospholipid fatty-acyl composition inSac- charomyces cerevisiaeNCYC 431. J. Gen. Microbiol.128: 1447– 1455. Bruce, L. J., Axford, D. B., Ciszek, B., Daugulis, A. J. 1991. Extractive fermentation byZymomonas mobilisand the control of oscillatory behaviour. Biotechnol. Lett.13: 291–296. Daugulis, A. J., Krug, T. A., Chroma, C. E. T. 1985. Filament formation and ethanol production byZymomonas mobilisin adsorbed cell bio- reactors. Biotechnol. Bioeng.28: 626–631. Fein, J. E., Barber, D. L., Charley, R. C., Beveridge, T. J., Lawford, H. G. 1984. Effect of commercial feedstocks on growth and morphology of Zymomonas mobilis. Biotechnol. Lett.6: 123–128. Ghommidh, C., Vaija, J., Bolarinwa, S., Navarro, J. M. 1989. Oscillatory behaviour ofZymomonasin continuous cultures: A simple stochastic model. Biotechnol. Lett.11: 659–664. Jarzebski, A. B. 1992. Modelling of oscillatory behaviour in continuous ethanol fermentation. Biotechnol. Lett.14: 137–142. Jöbses, I. M. L., Egberts, G. T. C., Luyben, K. C. A. M., Roels, J. A. 1986. Fermentation kinetics ofZymomonas mobilisat high ethanol concen- trations: Oscillations in continuous cultures. Biotechnol. Bioeng.28: 868–877. Lee, K. J., Rogers, P. L. 1983. The fermentation kinetics of ethanol pro- duction byZymomonas mobilis. Chem. Eng. J.27: B31–B38. Lee, K. J., Skotnicki, M. L., Tribe, D. E., Rogers, P. L. 1980. Kinetic stud- Figure 5. Instantaneous specific growth rate (from eq. 6), and estimated dynamic specific growth rates (from experimental data and eq. 15) for the data of Ghommidh et al. (1989). 104 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 56, NO. 1, OCTOBER 5, 1997 ies on a highly productive strain ofZymomonas mobilis. Biotechnol. Lett 2: 339–344. Lee, K. J., Tribe, D. E., Rogers, P. L. 1979. Ethanol production byZy- momonas mobilisin continuous culture at high glucose concentrations. Biotechnol. Lett.1: 421–426. Li, J. 1996. Experimental investigation and mathematical modelling of oscillatory behaviour in the ethanol fermentation byZymomonas mo- bilis. Ph.D. Thesis, Queen’s University, Kingston, Ontario. Li, J., McLellan, P. J., Daugulis, A. J. 1995. Inhibition effects of ethanol concentration history and ethanol concentration change rate onZy- momonas mobilis. Biotechnol. Lett.17: 321–326. Luong, J. H. 1985. Kinetics of ethanol inhibition in alcohol fermentation. Biotechnol. Bioeng.25: 280–285. Rogers, P. L., Lee, K. J., Skotnicki, M. L., Tribe, D. E. 1982. Ethanol production byZymomonas mobilis. Adv. Biochem. Eng.23: 37–84. Veeramallu, U., Agrawal, P. 1990. A structured kinetic model forZymomo- nas mobilisATCC10988. Biotechnol. Bioeng.36: 694–704. DAUGULIS, MCLELLAN, AND LI: OSCILLATORY BEHAVIOR OF ZYMOMONAS MOBILIS 105