Blotechnol. Rcg. 1992, 8, 501-507
101
Effects of Glucose Fluctuations on Synchrony in Fed-Batch Fermentation of Saccharomyces cerevisiaet Pradyumna K. Namdev) B. G. Thompson,§ Dennis B. Ward,§ and Murray R. Gray**$ Department of Chemical Engineering, University of Alberta, Edmonton, Alberta, Canada T6G 2G6, and Biotechnology Department, Alberta Research Council, Edmonton, Alberta, Canada T6H 5x2
The effects of transient conditions, due to fed-batch dynamics and fluctuations in glucose supply, on synchronous growth were evaluated by conducting fed-batch fermentation of Saccharomyces cerevisiae. Chemostat operation with dilution rates of 0.2 and 0.1 h-I produced periodic oscillations of the C02 fraction in off-gas and of the budded fraction of the cell population. A fed-batch operation with an exponentially increasing glucose feed was started from a synchronized culture established in a chemostat. The CO2 cycles continued for about five cycles (11h of feed time) and then they disappeared due to gradual decay of synchrony. The fluctuations in glucose supply due to poor mixing were simulated by an intermittent supply of nutrients with a pulse-on time of 5 s and a pulse-off time varying from 3 t o 39 s according to Monte Carlo method. Such rapid fluctuations in glucose supply during a fed-batch fermentation immediately produced a chaotic response in C02 levels which indicated a complete disruption of the synchrony established previously in a chemostat. The elimination of C02 oscillations was consistent with the measurements of the fraction of budded cells. These results suggest that synchrony cannot be established in large-scale stirred fermentors due to inherent concentration gradients.
Introduction Spontaneous generation of autonomous oscillations of variables, such as biomass, dissolved oxygen, pH, COn fraction in off-gas, flow cytometry measurements, and NADH fluorescence, has been observed in chemostat cultures of the budding yeast, Saccharomyces cerevisiae (Kuenzi and Fiechter, 1969;Myenburg et al., 1973;Meyer and Beyeler, 1984; Parulekar et al., 1986; Scheper and Schugerl, 1986; Porro et al., 1988; Strassle et al., 1989). These oscillations were shown to be due to the tendency of budding yeast to self-synchronize under uniform conditions in a chemostat culture when the dilution rate was below a critical level. If a yeast population is divided into two categories, single cells and budded cells, then all of the cells in a fully synchronized population will alternate between these two types of cells. The oscillatory behavior of budding yeast has been considered to be a severe problem in process analysis and control of the chemostat operation (Meyer and Beyeler, 1984;Parulekar et al., 1986; Porro et al., 1988). Nevertheless, growing a synchronous culture of yeast in chemostat culture has been useful in determining the metabolic events associated with various phases of the cell cycle of a yeast cell (Kuenzi and Fiechter, 1969; Strassle et al., 1988; Munch et al., 1991). Oscillations in a chemostat culture, which are not caused by imperfect control of culture conditions, could occur due to certain feedback interactions (Harrison and Topiwala, 1974). These interactions could be either between cells and environment parameters or between linked intracellular reactions. The autonomous oscillations in an aerobic glucose-limited chemostat culture of budding
* Author to whom correspondence should be addressed. t A part of this paper was presented at the American Institute of Chemical Engineers Annual Meeting, Los Angeles, November 1991. t University of Alberta. f Alberta Research Council.
87587938/92/3008-050 1$03.00/0
yeast were linked to secretion of ethanol in the buddedcell phase and subsequent utilization of the ethanol in the single-cell phase (Kuenzi and Fiechter, 1969; Porro et al., 1988). Secretion and uptake of ethanol depended on the levels of glucose and dissolved oxygen in the fermentor. The oscillations in chemostat culture were temporarily eliminated when a pulse of glucose or ethanol was added (Parulekar et al., 1986; Martegani et al., 1990). Too high or too low dissolved oxygen levels or a high steady-state glucose level resulted in a stationary steady state. Batch and fed-batch fermentations, not continuous fermentation, are used for commercial production of baker’s yeast, chemicals, and recombinant products from S. cerevisiae. The quality of baker’s yeast, indicated by its fermentative activity and storage capability, is considered to be better when the fraction of the population in the budded phase is lower (Reed, 1982). For example, during production of baker’s yeast, the supply of molasses is reduced at the end of fermentation to obtain a reduced fraction of budded cells. The optimal on-off type control policy for growth rate produced a final product with higher fermentative activity compared to an exponential fedbatch fermentation (Takamatsu et al., 1985). Growth under this optimal control strategy was shown to be equivalent to a synchronously dividing culture. Operational strategies to produce synchrony in fed-batch fermentation, therefore, could be useful in obtaining a better and more consistent quality of final product. The pseudo steady state in a fed-batch fermentation under certain conditions has been shown to be equivalent to a chemostat (Yamane et al., 1977; Lim et al., 1977). Thus, it may be possible to establish a synchronously dividing culture in fed-batch fermentation. Performance of a large-scale fed-batch fermentation depends on the degree of heterogeneity of the population and the contents of the fermentor. Poor mixing in large stirred fermentors, which can have volumes up to 300 m3,
0 1992 American Chemical Society and American Institute of Chemical Engineers
Biotechnol. Prog., 1992, Vol. 8, No. 6
502
Table I. Composition of Media Used in Fermentations with S. cerevisiae ATCC 32167. component
units
start-up
glucose (NH.dzS04 (NHdzHP04 chlorides sulfates vitamins
glL g/L g/L mL/L mL/L mL/L
2.5 0.5 0.16 2.5 2.5 0.5
chemostat 5.0 1.0 0.32 5.0 5.0 1.0
fed batch
100.0 0.0 6.4 40.0 40.0 20.0
a Stocksolutionsof chlorides,sulfates,and vitamins were prepared. The final concentrations of various components in the media were in the same proportions with respect to glucose a~ those given in D medium by Fiechter et al. (1987).
causes the formation of gradients in the concentrations of glucose and dissolved oxygen (Anderson et al., 1982). Frequent exposure of yeast cells to different concentrations during their circulation in the fermentor was experimentally simulated using a Monte Carlo method and a circulation time distribution (Namdev et al., 1991). These concentration fluctuations in an agitated fermentor and the transient nature of the fed batch may lead to difficulty in the establishment of synchrony in fed-batch fermenbation. Partially synchronized populations have been produced in batch (Wiemkenet al., 1970)and in fed-batch fermentations (Alberghina et al., 1991). As cells grew in the fermentor to a high biomass level, the synchronization became less pronounced and finally disappeared. The oscillation in the budded fraction during the fed-batch fermentation was attributed to the synchronization of a subpopulation caused by on-off control of nutrients used to achieve high cell density. The objective of the present study was to evaluate the effect of fluctuations in glucose supply due to poor mixing on a synchronously dividing culture of S. cerevisiae in fed-batch fermentations. Fed-batch fermentation with an exponentially increasing feed supply was started from a synchronous culture of S. cerevisiae established in a chemostat. Fluctuations in glucose supply were simulated for a 150m3stirred fermentor by intermittently supplying feed to the culture. The mixing effects were evaluated by comparing the performance of continuously fed with intermittently fed, fed-batch fermentations.
Materials and Methods Microorganisms a n d Medium. S. cerevisiae ATCC 32167 was maintained on YD (yeast extract and dextrose) slants for long-term storage and steaked on MYPD plates (malt extract, yeast extract, peptone, and dextrose) every 2 week A single colony was used to inoculate a shake flask with medium as follows: glucose, 15g/L; yeast extract, 7 g/L; NHdC1,2.5 g/L; KH2P04,l.O g/L; MgS0~7H20,0.3 g/L; CaC1~2Hz0,O.lg/L. The defined medium used for fermentation was based on the “D” medium (Fiechter et al., 1987). Diluted medium was used to start batch fermentation and then D medium for chemostat culture, and then a modified medium was used as the feed for the exponential fed-batch culture (Table I). RO water was used for the preparation of media. Equipment. A 2-L New Brunswick Multigenfermentor was equipped for computer monitoring and control. The fermentor off-gas was analyzed for 02, COZ,and N2 by a Dycor quadruple mass spectrometer (Model MPOOBEF, Ametek Thermox Instrument Division). The gas composition data was collected and stored on a V A X computer every 2 min. Sensorsfor pH (Phoenix autoclavable double reference electrode), temperature (Omega),and dissolved oxygen (Ingold autoclavable DO probe) were connected
to a microcomputer via an OPTO-22 data acquisition system. The values for pH, DO, and temperature were recorded every 20 s on the microcomputer. Tylan mass flow controllers and a peristaltic pump (Pharmacia P1: 0.6-500 mL/h, precision 0.5 mL/h) were used under computer control to set the flow rates of air and nutrients, respectively. Cultures a n d Assay. The inoculum (100 mL) was grown in a shake flask (500 mL) to the exponential phase for 24 h at 30 “C and 250 rpm. The fermentation was carried out with a 5 5% (v/v) inoculum and an initial content of 1 L of start-up medium at 30 OC, 900 rpm, and 1.5 L/min of air supply. These conditions ensured that the dissolved oxygen was always maintained above 20 % in all experiments. The pH was controlled at 4.9 f 0.1 by using 2 N NaOH and 2 N HC1 during batch and chemostat operation. The chemostat was started after 3 h of batch fermentation. The dilution rate was increased, in two equal steps, from 0.16 to 0.2 h-I within 4 h. A partially synchronized culture was established after 12 h of chemostat operation, which also produced periodic oscillations in COz fraction of off-gas. All fed-batch experiments were consistently started from a partially synchronizedculture after about 21 h of chemostat operation at a dilution rate of 0.2 h-l. The feed was switched on during the rising period of the cycle of COZ fraction in off-gas. An exponentially increasing feeding strategy was based on the following equation (Lim et al., 1977),
where the initial conditions used were as follows: biomass, XO,of 2.25 g/L of glucose; fermentor volume, VO,of 1L; glucose concentration in the feed, Gin, of 100 g/L. The parameters were a growth rate, p, of 0.16 h-l and a biomass yield on substrate, Yxlg,of 0.45. The pH was controlled between 4.7 and 5.0 during the fed-batch run by using 2 N NH40H, which also served as a nitrogen source during the fed-batch operation. A 5-mL sample from the fermentor was taken frequently during chemostat operation and every 1.5 h during the fed-batch experiments. Samples were withdrawn into precooled tubes and placed in an ice bath to slow yeast metabolism. The samples were immediately filtered using 0.45-pm Sartorius cellulose nitrate filters, and the supernatant was stored at -20 OC. The biomass was measured by absorbance at 620 nm using a spectrophotometer (Spectronic 21). The dry weight was obtained by drying the filtrate using a microwave oven for 10min. A correlation between absorbance and dry weight was obtained at various times during a fed-batch run. Glucose in the supernatant was analyzed using a YSI glucose analyzer (Model 271, which could detect down to 0.1 g/L. Ethanol was analyzed with an enzyme kit (Boehringer Mannheim) with a lower limit of 0.01 g/L and a precision of k0.002 g/L. The number of single cells and budded cells was counted by fixing cells with 10% formaldehyde in saline and using a hemocytometer (Pringle and Mor, 1975). From 350 to 500 cells were counted for each sample in duplicate. The budded fraction was calculated by taking the ratio of the total number of cells, and it was reproducible within f2.5%. Design of Cyclic Experiments. The effect of glucose fluctuations on synchrony was evaluated by comparing the performance of two types of experiments (1) %ontinuous” experiments, where feed was continuouslyadded to the fermentor accordingto an exponentially increasing feeding strategy, and (2) “MonteCarlo”experiments,where the feed was intermittently supplied to the fermentor
SO3
Biotechnol. Frog... 1992, Vol. 8 , No. 6
according to a Monte Carlo method (Namdev et al., 1991). These two types of experiments simulated the performance of a large fed-batch fermentor under perfect and poor mixing conditions, respectively. For Monte Carlo experiments, the feed into the fermentor was switched on and off on the basis of a Monte Carlo algorithm (Namdev et al., 1991). For a duration of feeding (tp) of 5 s, the feed was supplied a t a flow rate given by
C t F,, = - F ( t ) (2) tP where F ( t ) followed eq 1. The feed supply was switched off for a period, (tc-tp),where t , varied from 8 to 44 s. The magnitude oft, was selected from a log-normal circulation time distribution with t , = 20 s and u = 8.9 s, according to the Monte Carlo method. The log-normal distribution used in the present study characterizes the circulation time distribution for a typical 150 m3 stirred fermentor containing Newtonian broth (Bajpai and Reuss, 1982). During a typical Monte Carlo run of 14 h, approximately 2500 cycles of feed-on and feed-off were repeated. A fedbatch fermentation with periodic feeding with a constant period of 15 s was also carried out to evaluate the effect of distribution of circulation times. The total volumes of feed added to all fed-batch experiments were equal to within h5 mL. Results and Discussion Characterization of Oscillations. Chemostat experiments were carried out at two dilution rates ( D )of 0.1 and 0.2 h-l. Steady oscillations of the C02 fraction in off-gas, dissolved oxygen (DO), and biomass were established after 15 h of chemostat operation at a dilution rate of 0.2 h-' and after 30 h at 0.1 h-l. Typical cycles of C02 fraction and dissolved oxygen are shown in Figure 1. The ratio of budded cells to total number of cells or budded fraction (BF) measured for a complete cycle of CO2 is shown in Figure 2. The cycle of budded fraction was in phase with the COS cycle for both dilution rates. The fraction of cells exhibiting buds never fell below 10-15 % ; therefore, a nonsynchronous subpopulation was always present. The increase in period from 9.8 h for a dilution rate a t 0.1 h-' compared to 2.7 h at 0.2 h-l was consistent with the results of Strassle et al. (1989) with the same medium and strain of yeast, as were the changes in shape of the C02 curves as a function of dilution rate. Spontaneous initiation and maintenance of oscillations have been attributed to switchingfrom respirofermentative growth on glucose and stored carbohydrates during the budded-cell phase to aerobicgrowth on glucose and ethanol during the single-cell phase (Kuenzi and Fiechter, 1969; Porro et al., 1988). According to this model, simultaneous utilization of glucose and stored carbohydrates saturates the respiratory capacity and leads to excretion of ethanol. The bypass of glucose into ethanol provides a fast supply of ATP needed to complete the budded-cell phase in a constant duration. The self-synchronizationof the culture is induced by acceleration of the growth of single cells due to simultaneous uptake of glucose and ethanol. Any perturbations in glucose and dissolvedoxygen levels during the single-cell phase will, therefore, affect the degree of synchrony. In the present study, ethanol accumulated up to 0.015 mg/L for a period of 0.5 h when a dilution rate of 0.1 h-l and 5 g/L of glucose in the feed were used. Poro et al. (1989) showed that ethanol accumulated up to 0.037 g/L for a period of 2.5 h when a dilution rate of 0.106 h-' and 5 g/L of glucose in the feed were used. Lower levels of
DO
?J1sc
- 95
D = 0.2 h-'
0.7 0.6
z W
2 w
W
h
- 990 0
K
E
I
no
w
0.5
D
= 0.1 h-'
h
I
R
0
,i
I
0.1 0
200
400
600
800
, 1000
75 1200
TIME (min)
Figure 1. Oscillations in dissolved oxygen in the fermentor and COz concentration in off-gas during chemostat operation at dilution rates of 0.1 and 0.2 h-l. The point where t = 0 was selected arbitrarily.
ethanol and net excretion for a shorter duration in the present study could be due to its consumption by the nonsynchronized subpopulation. Figure 1 shows that dissolved oxygen oscillated between 93% and 97% at a dilution rate of 0.2 h-l and between 91 5% and 96% at 0.1 h-l. These near-saturation levelsof dissolved oxygen were higher than in previous studies, wherein oscillations disappeared when dissolved oxygen was higher than 7090% (Parulekar et al., 1986; Porro et al., 1989). The observed oscillationsin C02 and the budded fraction of cells were similar to results from Strassle et al. (1989) with the same strain; therefore, the existence of periodic cycles in the C02 concentration was taken to indicate synchrony of at least a portion of the culture. Without synchrony, periodic oscillations in C02 would not be observed. Conversely, the C02 peaks would reach a maximum, relative to the base line, as the fraction of synchronized cells in the population increased toward 100%. We assumed, therefore, that the area under the CO2 peaks divided by the total area was proportional to the fraction of the cell population that was synchronized. Fed-Batch Experiments. Fed-batch experiments using glucose feeds were performed to study the effect of transient conditions on the synchrony of S. cerevisiae. Figure 3 shows a profile of C02 fraction in off-gas during a typical complete fermentation experiment. Fermentations were started from partially synchronized cultures established in chemostat culture at a dilution rate of 0.2 h-l. The chemostat culture was shut off after three COZ cycles, and the fed batch was started during the rising portion of a COz cycle. The oscillations in C02 continued
Biotechnol. Prog,, 1992, Vol. 8,
504 TIME (min) 100
50
0
TIME (h)
200
150
No. 6
250
20
300
0.7
25
,.
I
35
30
40
I
0.6
0.5
5 F V
0.4
2
0.3
w a
a
m 0.2
0 BUDDED FRACTION
- co, I
0.0
I
I
0.5 1
I
I
I
(Z)
10.1
I
I
I 0.0 1 0.7
1
I
I
I
'" 0.0
p
FED-BATCH I
CONTINUOljS
'
0.3
h
6:
v
N
0 V
0.2
c
0.1
b
I
FED-BATCH START
v Atom 0.0
1
0
I
1
I
!
I
100
200
300
400
500
10.0 600
TIME (min)
operation at dilution rates of 0.1 and 0.2 h-l.
FED-BATCH STOP
3.5
*
3.0
2.5 h
6:
v
2.0 ON
V
1 .5
FED-BATCH
-
CHEMOSTAT
0.5 0.0
v
0
5
20
CARW
I
I
I
25
30
35
40
TIME (h)
Figure 2. Correlation between budded fraction of yeast population and C02 concentration in off-gas during chemostat
1 .o
0.0
I
I
I
I
I
I
10
15
20
25
30
35
i
40
TIME (h)
Figure 3. COz concentration in the off-gas as a function of time for a complete experiment consisting of batch, chemostat, and fed-batch operations.
during the fed-batch fermentation for about five cycles (11 h of feeding), after which they disappeared (Figure 3). Similar oscillations in dissolved oxygen levels were also observed. Ethanol, biomass, and glucose were analyzed every 1.5 h during fed-batch operation. No detectable ethanol was observed in the fermentor for the initial 12 h of feeding, after which accumulation of 0.7 g/L in the
Figure 4. Comparison of profiles of the fraction of yeast cells exhibiting buds in the continuous fed-batch fermentation and the Monte Carlo fed-batch fermentation. last 2.5 h of feeding was detected. Glucose accumulated from 0.12 to 0.35 g/L during the fed-batch operation. Biomass accumulation from 2.25 to 10.5 g/L in 14.5 h of fed-batch operation was detected. The profile of the budded fraction measured during a fed-batch experiment is shown in Figure 4a. The amplitude of the first cycle of the budded fraction immediately after the start of fed-batch operation was 13 5% larger than the amplitude of a typical cycle observed in chemostat culture. The amplified cycle of the budded fraction was followed by four cycles with reduced amplitudes. About 90% of the population was in the single-cell phase (nonbudded) at the end of 12 h of feed supply. These variations in the oscillations of budded fraction indicated that the degree of synchronywas affected by the dynamics of fed-batch operation. Similarly, the fraction area under the C02 peaks for all five cycles in Figure 3 reached a maximum of 0.3 at the second cycle after the fed-batch operation was started and then decreased to 0. These data for budded fraction and C 0 2 indicated an initial increase in the fraction of cells in synchrony and a later decrease to extinction. The switch to fed-batch operation also affected the relative size and number of the maxima within a single C 0 2 cycle. The second maximum increased relative to the first in the initial three COZcycles during fed-batch operation and then disappeared in later cycles (Figure 3). A third maximum appeared between the original two maxima during the third cycle of fed-batch operation.
Biotechml. Prog., 1992, Vol. 8, No. 6
Strassle et al. (1989)observed that whenever oscillations disappeared from a chemostat after several weeks of operation, the last few cycles had only two peaks even though the initial cycles had three. They did not observe this transition in response to feed perturbations. Parulekar et al. (1986)observed complete elimination of oscillations in chemostat cultures following a step-up (0.2to 0.25 h-l) or step-down (0.2 to 0.05 h-l) in dilution rates. The oscillations returned on the reverse step changes. Unlike the disturbance caused by step changes in dilution rate on the synchronized culture, the exponential fed-batch fermentation in the present study produced a gradual variation in "equivalent" dilution rate. From Lim et al. (1977),the exponential fed-batch fermentation based on eq 1 would be similar to a constantvolume chemostat with the dilution rate steadily varying from 0.008to 0.052 h-' over 14 h of feeding. The transient condition introduced by the exponential fed-batch strategy, therefore, would be insufficient to immediately disrupt the synchrony. Fluctuations in Glucose Supply. The effect of concentration gradients due to poor mixing was simulated by supplying the glucose feed intermittently according to a Monte Carlo method (Namdev et al., 1991). During Monte Carlo experiments, feed was switched on for 5 s and switched off for an interval in the range of 3-39 s. The Monte Carlo fed-batch experiment was compared with continuously fed fermentation according to an exponential feeding strategy. In addition to these two types of fedbatch fermentations, a periodic fed-batch fermentation was carried out where feed was added every 15 s for 5 s. The biomass profiles were similar during the first 12 h of feed supply for all three types of experiments. No ethanol was detected while the glucose level remained below 0.35 g/L for the same period. The biomass and ethanol profiles for these three experiments diverged after 12 h of fedbatch operation. About 10% lower biomass and ethanol accumulations up to 1.7 g/L were observed for periodic and exponential fed-batch experiments during the last 2.5 h of a 14.5-hexperiment. The profiles of C02 and dissolved oxygen for all three types of fed-batch experiments are shown in Figures 5 and 6,respectively. Continuous and Monte Carlo experiments were repeated in triplicate, and reproducible resulta were obtained. The C02 and dissolved oxygen cycles were eliminated within the first 4 h of the Monte Carlo experiments. Fourier transform analysis showed no dominant frequency in the CO2 fluctuations during Monte Carlo experiments, implying a chaotic respiratory response of the yeast culture. In similar media, a step change in pH from 4 to 6 and back changed C02 by 13%, while changes between pH 4.7 and 4.9 changed COZby
