Biotechnol. hog. 1992, 8, 1-4
1
ARTICLES Plasmid Stabilization of an Escherichia coli Culture through Cycling M. L. Stephens,t C. Christensen, and G. Lyberatos* Department of Chemical Engineering, University of Florida, Gainesville, Florida 32611
The problem of plasmid instability of fermentations that involve plasmid-bearing recombinant organisms is dealt with in this work. Previous theoretical work demonstrated that under certain conditions (where plasmid-bearing species are slower in responding t o changes in the fermentation environment than the wild species) the washout of the plasmid-bearing species can be prevented. In the sequel, Weber and San showed that cycling the dilution rate can delay the washout of plasmid-bearing species for a plasmidbearing Escherichia coli culture. This work shows that it is indeed possible to secure the presence of the plasmid-bearing species at all times through appropriate cycling.
Introduction Segregational instability is one of the major problems of industrial fermentations involving recombinant microorganisms. This problem is the result of the independent proliferation of plasmids during the cell’s life cycle. When a cell undergoes division, these plasmids are distributed among the resulting daughter cells. Occasionally, a daughter cell results that does not contain the plasmid and can no longer produce the desired product. These recombinant organisms are essentially thus providing their own contaminant to the reactor in which they are grown and ultimately are replaced by it if the reactor is operated in a steady continuous mode (e.g., refs 1-5). This occurs because the plasmid-bearing species are at a competitive disadvantage to their wild counterparts, since the presence of plasmid genes leads to competition for the resources that would otherwise be used exclusively for growth. Approaches that have been proposed for solving this problem include: (1) growth of the culture in an antibiotic-containing medium coupled with the use of a plasmid that provides resistance to the particular antibiotic (the problem of this approach is the expense involved with using antibiotic in the medium), (2) use of an auxotrophic host that requires the plasmid to metabolize an essential nutrient (e.g., ref 61,and (3) employment of plasmids that produce substances toxic to the plasmid-free cells (e.g., refs 7, 8). In a previous publication (91,we raised the following question. Under what conditions could periodic manipulation of a process variable bring about stabilization of the plasmid-containing population? We found that if the plasmid-free cells are faster in adapting to environmental changes, then it is indeed possible to give a competitive edge to the plasmid-containing population through cycling. In the sequel, Weber and San (10, II), working with Escherichia coli RP1 containing the plasmid pBR322 (ATCC 370171, showed that, through cycling of the dilution rate, washout of the plasmid-bearing species can be delayed. Their LB medium was also supplemented with
* Author to whom correspondence should be addressed at the Departmentof ChemicalEngineering,Universityof Patras,GR 26110, Greece. Present address: General Electric Corp., Schenectady,NY.
*
ampicillin, an antibiotic giving a competitive edge to the plasmid-bearing species. Unfortunately, one of the two possible explanations they offered is that the plasmidbearing species is faster in responding to changes, which is the exact opposite of the requirement we theoretically proved in ref 9. Furthermore, they started with a pure plasmid-bearing culture and, of course, observed a decrease of the plasmid-bearing fraction with time. This is something expected since even if stabilization is to occur through cycling, the eventual fraction will have to be less than one due to the continued reversion. In parallel with Weber and San, we investigated the feasibility of the approach, and the results are presented in the present paper. Our medium did not contain an antibiotic, and we deliberately started with a small fraction of plasmid-bearing species in order to investigate if cycling can increase this fraction over time. This is evidently the only way the feasibility of the approach can be proved. The objective of this work is to explore the feasibility of plasmid stabilization through cycling using an actual experimental system.
Materials and Methods The organisms used in this study were a strain of E. coli harboring the plasmid pBR322 and the host strain of E. coli from which the engineered strain was developed. Both strains were obtained from the American Type Culture Collection (ATCC), strain numbers 31344 and 31343, respectively. The plasmid contained in this organism (pBR322)provides the cell with resistance to the antibiotics tetracycline and ampicillin. The resistance to tetracycline was the marker used to track the fraction of the population that contains the plasmid. The microorganisms were grown on a mineral salts minimal medium containing glucose as its single source of carbon and energy. The medium was supplemented with two amino acids, proline and leucine, and a vitamin, thiamin, that the strains are unable to synthesize due to a metabolic deficiency. The complete medium formulation was as follows: K2HP04, 7 g/L; KH2P04, 3 g/L; (NH& SO1, 0.1 g/L; FeSOp7H20, 0.5 g/L; CaC12, 11.1 mg/L; MgSOp7H20,O.l g/L; proline, 20 mg/L; leucine, 20 mg/L; thiamin, 2 mg/L; and glucose accordingly. The fraction of plasmid-bearing species in the population was deter-
@ 1992 American Chemical Society and American Institute of Chemical Englneers 8~56-~938/92/3008-0001$03.00/0
Biotechnol. Prog., 1992, Vol. 8, No. 1
2
mined by carrying out serial dilutions of samples and growing them on nutrient agar plates, both with and without the addition of tetracycline. The concentration of biomass in a sample was determined by measuring the absorbance at 550 nm in a Milton Roy Spectronic 20D spectrophotometer. It was found that optical density is proportional to cell-mass density over the range of values in this study. The conversion factor from optical density to dry weight of biomass was determined by filtering samples of biomass and drying them for various absorbance values. The concentration of residual glucose in a sample was measured by an Analytic Research Model 110 glucose monitor. The inoculum cultures for all experiments were prepared by culturing from nutrient agar slants the appropriate strain of bacteria. The inocula were then transferred to 250-mL seed flasks containing minimal medium and glucose at a concentration of 3.0 g/L. These flasks were then placed in a Blue M Magniwhirl constant temperature bath at 37 "C. When the seed flasks became visibly turbid (after about 1day), a sample of the seed flask was used to inoculate the shake flask or fermentor. Shake-flask batch experiments were carried out in 500mL beakers with a working volume of 250 mL. All of the continuous experiments, those with either constant or periodic input, and some batch experiments were carried out in a 3.7-L Bioengineering KLF 2000 fermentor. The working volume in every instance was 2 L. Good mixing was accomplished by a rotation of two flat-blade agitators at 500 rpm. A baffle cage ensured thorough mixing. A constant temperature of 37 "C was maintained in the reactor by a PT 100temperature sensor, a Bioengineering K54450 controller, and an 800-W heater. An Ingold Ag/ AgCl pH electrode and a Bioengineering M7832N pH controller were used to maintain a constant pH of 7.0 environment by the addition of a NaOH solution. The dilution rate of the system was maintained by a load cell that kept the volume of the reactor constant. For periodic operation, the feed pump was controlled by a PDP 11/23 computer, switching between two levels of dilution rate at predetermined times. A detailed description of all procedures may be found in ref 12.
Results Batch Experiments. The purpose of the batch runs was to determine the best conditions under which to operate the continuous nonperiodic and periodic experiments. This included characterizing the conditions under which various growth limitations were expected to occur. The experiments also determined whether the chosen recombinant E. coli strain would demonstrate the appropriate growth characteristics to allow the possibility of maintaining the plasmid-bearing strain in the culture through cycling. The conclusions from the batch experiments can be summarized as follows. The amino acid and vitamin additions are necessary to sustain growth. Batches with the metabolite addition and an initial glucose concentration of 1000mg/L will grow to complete glucose utilization. Batches with an initial concentration of 3000 mg/L show a growth limitation caused by either product or substrate inhibition. A batch that starts with a lag phase experiences balanced growth. The wild strain seemed to be more responsiveto environmental changes than the recombinant strain, and it grew at a faster rate. A detailed description of shake-flask and bioreactor batch experiments may be found in ref 12. I t was thus seen that the chosen system had the desired properties explained in the Introduction
Table I. Apparent Steady-State Values h-1
inlet substrate concn, mg/L
av absorbance
av substrate,
0.11 0.2 0.25 0.2 0.28
1000 1000 1000 2000 2000
0.139 0.205 0.211 0.242 0.260
20.2 18.5 32.0 419.0 680.0
D,
mg/L
that allow an attempt at improvement in recombinant retention through periodic operation. Continuous Operation. The experiments to be described in this section all took place in the Bioengineering fermentor with a 2-L working volume. The runs were carried out at constant temperature, 37 "C, and pH, 7.0. The two control variables that were changed between runs were the dilution rate and the feed glucose concentration. During a run, all inputs to the system were kept constant. As previously discussed, a system of this type at steady state is eventually expected to exhibit extinction of the plasmid-bearing strain. The purpose of these runs was to verify that the loss of plasmid does indeed occur and that the fraction of recombinants in the population decreases with time. No true steady state was attained in these runs since a mixed population was always present in the reactor, but the values of absorbance and residual substrate concentration did seem to approach what could be considered constant values. This occurs since the two strains are identical except for the presence of the plasmid in the recombinant strain. It is expected that the yield factors and other growth parameters of the two strains are sufficientlysimilar that only significant changes in the total fraction of the population can result in actual differences in the two gross features (absorbance and residual glucose) of system behavior that are considered here. A listing of the so-called "steady" values is given in Table I and a plot is given in Figure 1. The data are interesting in several regards. The five sets of data involve four different values of the dilution rate and twodifferent values of the feed glucose concentration. Operation at a feed glucose concentration of 1000 mg/L results in very small residual glucose concentration regardless of the system dilution rate. The two runs carried out at 2000 mg/L have residual glucose values that indicate that significant substrate utilization is occurring. It is hard to draw conclusions on the basis of the residual substrate values for the runs carried out at 1000 mg/L since they are all very close to the same value. The absorbance values are much more conclusive. A typical Monod-type behavior predicting that the biomass concentration decreases with increasing dilution rate was not observed for this system in the range of dilution rates studied. Also in typical Monod behavior, the residual substrate concentration is not affected by the inlet substrate concentration. The runs carried out at D = 0.2 h-l, anddifferent inlet substrate values show that this too is not the case for this system. Typical results are shown in Figures 2 and 3. The run in Figure 2 started under batch conditions, and then continuous operation at a flow rate of 0.28 h-' and an inlet substrate concentration of 2000 mg/L was implemented. This run shows an overshoot to the change in operating conditions. The absorbance first increases, then it drops while the substrate increases to a large value, and then it decreases in response to the switch from batch to continuous operation. The loss of plasmid in the population with time is very clearly demonstrated in the figure. Plasmid loss is also demonstrated in Figure 3 for a dilution rate of 0.2 h-l and a substrate concentration of 2000 mg/L.
Bbtechnol. Prog., 1992, Vol. 8, No. 1 0.3
3
I
n
L
=
B
s
---
K
O
0.11
0.2
DILUTION RATE (llhr)
Figure 1. Substrate and absorbance values as a function of dilution rate for So = lo00 mg/L and SO= 2000 mg/L.
- 1.0
-
0.8
z
0
6 U
- Oe6 E5 -
-
u 0.4
z
3
- 0.2 8g 20
0
40
80
80
100-
TIME (hrs)
Figure 2. Absorbance, substrate concentrations, and recombinant fraction for D = 0.28 h-l and SO= 2000 mg/L.
p6
2
- 1.o - Pz - 0.8 6 - s
6oo
v
500
-
; 0
400-
0.8
0
g
300-
n
p 2
! TIME (hrs)
0.28
0.25
0.2
200-
0.4
100-
0.2
2m = 0
p
in 0
0
TIME (hrs)
Figure 3. Absorbance, substrate concentrations, and recombinant fraction for D = 0.2 h-l and SO = 2000 mg/L. All the continuous runs demonstrate the typical behavior of a mixed recombinant continuous culture operating under constant conditions. The fraction of recombinant bacteria in the population does decrease with time as expected. The batch experiments of this work suggested that the reversion phenomenon required a significant time span to express itself with differences in the observed growth rate and the fraction of recombinant cells in the population. The continuous r-unsdemonstrated that the reversion does indeed occur, with the most likely primary effect being the growth rate difference that further changes the population with time: Once a significant fraction of the cells are wild type, then the reversion phenomenon is overshadowed by growth rate differences and fails to make much more of a significant contribution to the increase in wild type. Thus, the problem basically becomes the problem of competition between two species for a common nutrient with the differences that no extinction of wild
Figure 4. Changein recombinant fraction for periodicoperation, batch 5. The cycling in the dilution rate is between 0.0 and 0.4
h-l. The cycling period is 2 h. The inlet substrate concentration is 2000 mg/L.
type is possible. This slowly reverting system is a good choice for demonstrating the effect of periodic operation on genetically engineered strains and on a general mixed competition problem at the same time. Having one of the strains contain a simple antibiotic marker on the plasmid allows easy differentiation between two competing organisms, much easier than a different type of differentiation, e.g., that based on differences such as cell size. Periodic Operation. The actual shift in a population from dominance by one strain to another under constant operating conditions was indeed demonstrated in the previous section. In response to this replacement of one type of strain by another less desired organism, a possible remedy was proposed in ref 9. The idea was to impose periodic switching from one level of inlet substrate concentration to another in a square-wave fashion. It was found that this method could reverse the trend of reversion from the desired to the undesired strain for certain operating conditions. I t was also found that the method was not specifically limited to a particular model form or a particular periodic input. It was found that the desired response is expected not only because of inlet substrate concentration switching but also because of a square wave implementation in the system dilution rate (which indirectly causes periodic variations of residual substrate). The latter method was chosen for reasons of simplicity of operation. The lower dilution rate value was always set to zero. This was easily done by simply having a computerinterfaced controller turn the feed pump off at the appropriate times. The load cell was used to maintain the system's constant volume (by switching the harvest pump off). At some later time, the computer switched the feed pump back on, and the harvest pump turned on shortly thereafter. The periodic experiments were carried out so that the total amount of medium fed to the system would be equal to the amount used for the continuous base case which was run a t a dilution rate of 0.2 h-l. The inlet substrate concentration, SO,for all periodic runs was set at 2000 mg/L. The desired upper value of the dilution rate was set to 0.4 h-l. In Figure 4 the effect of periodic operation is indeed seen to reverse the trend of reversion to wild type. This run was carried out by cycling in dilution rate between 0 and 0.4 h-l at a cycling period of 2 h. The system was started a t a small initial value of recombinants in the population so that any positive effect would quickly manifest itself. As explained in the Introduction, had we started with a pure culture of recombinant cells, it would
Biotechnol. Prog., 1992, Vol. 8, No. 1
4
-
E I z
o.8
a U
-
0.6 + z U.
i 1 U
0.4
0
-
0.2
U
c 0
0
I I
I 20
I
I 40
I
I
60
I
I
so
I
100
TIME (HRS)
Figure 5. Changein recombinantfraction for periodicoperation, batch 6. The cycling in the dilution rate is between 0.0 and 0.45 h-1. The cycling period is 4 h. The inlet substrate concentration is 2000 mg/L.
have been hard to draw definite conclusions regarding plasmid stabilization. As seen in Figure 4,the recombinant fraction significantly increased over the course of the experiment. A second periodic run (Figure 5) illustrates the same reversion in the trend of plasmid loss for a period of 4 h. The dilution rate for this experiment was varied between 0 and 0.45h-l. These periodic runs prove that it is indeed possible to reverse the effect of plasmid loss in recombinant populations by a simple environmental manipulation. The fact that Monod-type behavior is not followed, unfortunately, does not allow the use of the operating diagrams that were derived in ref 9. An adequate model that accurately describes the behavior of this system is needed before operating diagrams similar to those obtained in ref 9 can be constructed. Nonetheless, this work indeed proves the feasibility of the approach. Conclusions The primary conclusion of this experimental endeavor is that it is indeed possible to use simple periodic operation to prevent the expected extinction of a plasmid-containing microorganism in continuous culture. The particular system that was studied, a mixed population of E. coli containing a plasmid conferring antibiotic resistance and the resulting wild-type counterpart, exhibited the desired properties. The batch experiments demonstrated that the recombinant strain did indeed grow more slowly than the wild type. It was also found that the plasmid-bearing strain is less responsive to sudden environmental changes, a requirement, according to ref 9, for periodic operation to solve the plasmid instability problem. The batch exper-
iments also demonstrated that the fraction of recombinants in the population decreases with time and that the overall specific growth rate of the culture increases with time. It was seen that the primary effect in the decrease of plasmid with time was the growth rate difference between the two strains. The reversion phenomenon works on a time scale much longer than that allowed for the duration of the batch experiments. The nonperiodic continuous experiments verified that the plasmid-bearing fraction of the culture did decrease with time as was expected. The periodic experiments demonstrated that the periodic manipulation of a simple system-operating parameter such as dilution rate at a comfortable switching period (2-4 h) can indeed induce effective growth rate differences between the two competing speciesand prevent the extinction of the more slowly growing one. This work establishes periodic operation of plasmidbearing cultures as a promising no extra cost alternative to steady operation allowing plasmid stabilization. Further work is needed to establish the whole set of operating conditions under which plasmid stabilization can be effected through periodic operation and to relate the eventual recombinant fraction to the operating variables. Acknowledgment We gratefully acknowledge support of this work by the National Science Foundation through NSF Grant EET8657394. Also, acknowledgement is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for partial support of this research.
Literature Cited (1) Imanaka, T.; Tanaka, T.; Tsunekawa, H.; Aiba, S.J. Bacteriol. 1981,147,776. (2) Dwivedi, C. P.;Imanaka, T.; Aiba, S. Biotechnol. Bioeng. 1982,24,1465. ( 3 ) Siegel, R.;Ryu, D. D. Y. Biotechnol. Bioeng. 1985,27,28. (4) Zund, P.; Lebek, G. Plasmid 1980,3,65. (5) Heling,R.B.;Kinney, T.; Adams, J. J. Gen. Microbiol. 1981, 123,129. (6) Bailey,K.;Vieth, W. R.; Chotani, G. K. Ann. N.Y.Acad. Sci. 1987,506,196. (7) Lauffenburger, D. A. Biotechnol. h o g . 1985,1 (l),53. (8) Donoghue, D. J.; Sharp, P. A. J . Bacteriol. 1978,133,1287. (9) Stephens,M.L.;Lyberatos, G.Biotechnol. Bioeng. 1988,31, 464. (10) Weber,A.E.;San, K.-Y.BiotechnoLLett. 1988,lO(8),531. (11) Weber,A. E.;San, K.-Y. Biotechnol.Bioeng. 1989,34,1104. (12) Stephens, M.L. Ph.D. Dissertation, University of Florida, Gainesville, FL, 1989. Accepted July 12, 1991.
