Escherichia coli host cell modifications in continuous culture affecting

Escherichia coli host cell modifications in continuous culture affecting heterologous protein overproduction: a population dynamics study. Jeffrey Fu,...
0 downloads 0 Views 920KB Size
Biotechnol. Prog. 1992, 8, 340-346

340

Escherichia coli Host Cell Modifications in Continuous Culture Affecting Heterologous Protein Overproduction: A Population Dynamics Study Jeffrey Fu,tJ A. Paul Togna,+g§Michael L. Shuler,*9+and David B. Wilson11 School of Chemical Engineering and Section of Biochemistry, Molecular and Cell Biology, Cornel1 University, Ithaca, New York 14853

There are many published studies of plasmid segregational instability in Escherichia coli in the literature. However, the formation of plasmid-free segregants can be controlled by the addition of selective chemical agents like antibiotics. This solution has become commonplace in both the laboratory and industry. On the other hand, host cell modifications, which result in low production of plasmid-encoded protein and lead to loss of culture productivity, have not been adequately addressed. Continuous culture of an inducible (ptac) Escherichia coli vector containing strain, RB79l(pKN), was characterized by strong dynamic changes in the cell population and product (P-lactamase) expression. Long-term cultivation resulted in the loss of high-level production of 6-lactamase. Loss of productivity was not due to the formation of plasmid-free cells or structural modifications to the plasmid; instead, continuous operation resulted in a culture dominated by irreversibly altered, low-producing cells. Two distinct classes of lac- mutants which inhibited induction were identified (Y- and Is).

Introduction Various types of recombinant bacterial culture instability have been well documented. The two most common types of plasmid instability, structural and segregational, have been studied in great detail [e.g., Ensley (1986) and Zabriskie and Arcuri (1986)l. Structural plasmid instability is caused by point mutations, insertions, or deletions within plasmid DNA which result in decreased or even total loss of recombinant target protein production. Segregational plasmid instability refers to the complete loss of plasmid from a population of plasmid-bearing bacteria. Segregational instability is caused by uneven partitioning of plasmids between daughter cells upon cell division. If the plasmid-free cells have a higher growth rate than plasmid-bearing cells or cells with an altered plasmid, then instability is quickly manifested (growth-rate dependent instability). Missegregation is affected by plasmid copy number, culture conditions, and the genes which are found on the plasmid (Le., the par locus). Growth-rate dependent instability is related to the extent of metabolic stress that both the plasmid itself and protein synthesis from the plasmid have on the host cell. In the most severe cases, production of cytotoxic substances and overexpression of certain recombinant target proteins can result in complete loss of viable plasmid-bearing cells [e.g., Cheng et al. (1990) and Uhlin and Nordstrom (1978)l. These factors have led to the development of controllable plasmid promoter systems in which production of the target protein is induced either by chemical means or by a shift in temperature (Amann et al., 1983; Goeddel et al., 1980; Shatzman et al., 1983). There are many reports in the literature which describe

* Corresponding

author.

t School of Chemical Engineering. t Present address: Merck, Sharp & Dohme Research Laboratories, West Point, PA 19486. Present address: Envirogen, Inc., Lawrenceville, N J 08648. 11 Section of Biochemistry, Molecular and Cell Biology.

segregational plasmid instability and loss of target protein expression during continuous culture [e.g., Weber and San (1990)and Kim and Shuler, (1990)l. The majorityofthese reports describe the dynamics of segregational plasmid instability as a function of the metabolic burden the plasmid places on the host strain [e.g., Bentley and Kompala (1990) and Kim and Shuler (1990)l. In most cases, this type of instability can be overcomeby adding an antibiotic to the medium and inserting within the plasmid a gene encoding resistance to this antibiotic. Another common method of overcoming segregational plasmid instability is to use an amino acid auxotroph as host and to add the gene for the defective biosynthetic enzyme to the plasmid. However, these approaches do not affect structural instability of the plasmid or the appearance of host cell variants which cause low production. In this paper, we investigate the effects of host cell mutations on a culture’s ability to produce plasmidencoded proteins from an inducible promoter. We have previously developed a chemically inducible Escherichia coli host-plasmid expression system for p-lactamase using the tac promoter in which selective excretion of the target protein from the periplasm is observed without cell lysis (Georgiou et al., 1985, 1988). Here we use excretion to indicate that the protein is in the extracellular compartment while we reserve secretion to indicate passage of the protein across the cytoplasmic membrane; a secreted protein could be either localized in the periplasmic space or extracellular. Excretion of p-lactamase is closely linked to high-level expression of the target protein (Georgiouet al., 1988). During continuous culture, we have observed loss of high-level expression and of excretion of P-lactamase shortly after induction, even though antibiotic resistance is used as a selection against plasmid-free segregants (Chalmers et al., 1990). Under some conditions, this instability is observed even in batch culture but this instability is not observed in uninduced, nonproducing, cultures. In this paper, we show that this cultural instability is not caused by loss of plasmid or changes in

8756-7938/92/3008-0340$03.00/0 0 1992 American Chemical Society and American Institute of Chemical Engineers

Bbtechnol. Prog., 1992, Vol. 8, No. 4

plasmid structure but is instead caused by mutations within the host cell chromosome that affect expression from the plasmid.

Materials and Methods Bacterial Strain and Plasmid. E. coli RB791, a lacZqLa derivative of W3110, harboring the plasmid pKN, a pBR322 derivative, was employed for this study (Chalmers et al., 1990). Plasmid pKN contains the tac promoter upstream of the 8-lactamase gene and its signal sequence and can be induced by the addition of isopropyl 8-D-thiogalactopyranoside (IPTG). Plasmid pKN also confers resistance to the antibiotic neomycin. Medium Formulation and Culture Conditions. For continuous culture, RB79UpKN) was grown a t 20 "C in Tanaka medium (Tanaka et al., 19671, pH 7.2, supplemented with 4.0 g/L glucose (Sigma Chemical Co., St. Louis, MO) and 250 mg/L neomycin sulfate (Sigma ChemicalCo.). For shakeflaskexperiments, RB79l(pKN) was cultured at 20 "C in Tanaka medium (Tanaka et al., 1967),pH 7.2, supplemented with 2.0 g/L glucose and 2.0 g/L casein amino acids, acid hydrolysate (Sigma Chemical Co.). Cells were induced by adding lo4 M IPTG (Sigma ChemicalCo. or Pharmacia LKB Biotechnology,Inc., Piscataway, NJ) to the medium. The inoculum for all experiments came directly from a master frozen stock stored in glycerol at -70 "C. Assays and Cell Counting Techniques. Assays for 8-lactamase activity were performed in 50 mM phosphate buffer a t pH 7.0 using penicillin G (Sigma Chemical Co.) as substrate and monitoring the rate of decrease in absorbance at 240 nm (Samuni, 1975)with a Beckman ACTA MVI spectrophotometer (Beckman Instruments, Inc., Fullerton, CA). A unit of activity is defined as 1mmol of penicillin G hydrolyzed per minute at 25 "C using an extinction coefficient of 0.57 A240 unit/mmol (mL of penicillin G)-I. The specific activity of 8-lactamase under these conditions is 3550 units/mg (Georgiou et al., 1988). The total 8-lactamase (internal and external) was determined by rupturing cells in a Carver Model C laboratory press (Fred S. Carver Inc., Menomonee Falls, WI) at 20 000 psi and separating the soluble lysate from the cell debris by centrifugation at 13600g for 5 min in a Fisher Model 235C microcentrifuge (Fisher Scientific, Pittsburgh, PA). The culture supernatant was assayed directly for external 8-lactamase after cells were removed by centrifugation a t 13600g for 5 min. @-Galactosidasewas assayed by measuring the rate of hydrolysis of O-(nitrophenyl) 8-D-galactoside (ONPG) (Sigma Chemical Co.) in z-buffer (Miller, 1972). A unit of activity is defined as 1mmol of ONPG hydrolyzed per minute at 28 "C using an extinction coefficient of 4.5 A420 units/mmol (mL of 0NPG)-l. Under these conditions, the specific activity of 8-galactosidase was determined to be 330 units/mg. Total soluble protein of pressed cell samples was measured using the Bio-Rad Protein Assay (Bio-Rad Laboratories, Richmond, CA) with bovine serum albumin as the standard. For viable cell counts, the standard dilution plating method was used. Quintuplicate plates were incubated at 37 "C for 16-24 hours. Only those plates which contained between 30 and 300 colonies were counted. Nonselective plates contained LB medium (10 g/L Bacto-tryptone, 5.0 g/L Bacto-yeast extract, 5.0 g/L NaC1) with 15 g/L Bacto-agar (Bacto-products from Difco Laboratories, Detroit, MI). Selective plates were supplemented with 50 mg/L neomycin sulfate.

a41

A Coulter counter Model ZMwith Coulter Channelyzer Model 256 (Coulter Electronics, Inc., Hialeah, FL) was used to determine total cell counts. Cells were diluted in Isoton I1 (Coulter Diagnostics, Hialeah, FL) before counting. The optical density a t 600 nm, ODw, was measured using a Beckman ACTA MVI spectrophotometer (Beckman Instruments). One ODw corresponds to 0.5 g dry cell weight/L culture. Plasmid DNA Manipulations. Plasmid was isolated by the alkaline lysis method of Birnboim and Dolly (Maniatis et al., 1982). Purified plasmid was digested with EcoRI and loaded onto either a 0.6 % or 0.8 % agarose gel containing0.5 mg/mL ethidium bromide (Sigma Chemical Co.). Known weight standards and aHindIIIIEcoRI digest of X DNA (InternationalBiotechnologies,Inc., New Haven, CT), were also loaded. Gels were run a t 30 V in TBE buffer (Maniatis et al., 1982) containing 0.5 mg/mL ethidium bromide for approximately 10-14 hand photographed on a UV transilluminator using Type 55 Polaroid film (positive/negative) and a yellow filter. The intensities of bands on the negative were analyzed with a Quick Scan Flur-Vis densitometer (Helena Laboratories, Beaumont, TX). Plasmid copy numbers (plasmids per cell) were determined by dividing the number of plasmid molecules per milliliter of culture by the total number of cells per milliliter of culture on the basis of particle count. Cells were cured of plasmid using a modification of the acridine orange method of Hirota (1960). This method, originally used to cure E. coli cells of F plasmids, has been used to cure E. coli cells of other types of plasmids (Stafford and Swartz, 1984). Cells were grown at 37 "C in medium (pH 7.6) containing 10 g/L Bacto-peptone, 10 g/L Bacto-yeast extract, and 20 mg/L acridine orange (Sigma Chemical Co.), starting with an inoculum size of approximately lo4 cells/mL. After growth, cells were replicated onto selective and nonselective plates. Using this procedure, approximately 2% of the colonies failed to grow on selective plates and it was confirmed by agarose gel electrophoresis that these cells had lost the plasmid. Cells were transformed with plasmid using the standard calcium chloride procedure (Maniatis et al., 1982). [ 14C]TMG Uptake Experiments. The method used for measuring uptake of methyl P4C1-8-thiogalactopyranoside ([l4C1TMG)from cured cells was essentially the same as the method used by Wilson (1974) to measure galactose transport in E. coli. Individual colonies were grown overnight at 37 "C in 2 mL of LB medium. A total of 250 pL of each overnight culture was transferred to 25 mL of Tanaka medium supplemented with 2.0 g/L fructose (Mallinckrodt, Inc., Paris, KY). Cells were grown at 37 M IPTG was added "C with agitation. After 4.5 h, 2 X to fully induce the lac operon. Cultures were allowed to grow to an O D w of 0.8 as measured with a Gilford spectrophotometer 240 (Gilford Instrument Laboratories Inc., Oberlin, OH). Cells (20 mL of each culture) were centrifuged at 7000g for 7 min at 4 "C. The cells were then resuspended in 10 mL of 0.15 M NaCl and recentrifuged. The final cell pellets were each resuspended in 2 mL of Tanaka medium and stored on ice for up to 4 h. An amount of cell suspension (200 pL) that would take up less than 1% of the [WITMG was transferred to 400 pL of Tanaka medium containing 3.0 g/L fructose. Cells were incubated for 10 min a t room temperature (21 "C), and 18 pL of 1 X M [l4C1TMG (14 mCi/mmol) (New England Nuclear Research Products, Boston, MA) was added to obtain a concentration of 3 X M. Portions

Biotechnol. Prog., 1992,Vol. 8, No. 4

342 4 ~

3

5 0 iD 0

0

c 2

I

:et9 io3

1

0

1oc

200

300

200

303

430

4co

hours a f t e r induczion

Hours a f i e r i n d u c t i o n

Figure 1. Optical density at 600 nm of RB79l(pKN)in a chemostat under continuous induction (D = 0.03 h-l, 20 "C).

of 200 PL were removed and filtered onto 25-mm diameter Millipore 0.45-mm Type HA Filters (presoaked in 0.15 M NaCl) (Millipore Corp., Bedford, MA) 30 and 60 s after addition of [WITMG. The filtered cells were washed with 4 mL of 0.15 M NaCl at room temperature. Filters were dried a t 70 OC and were each added to counting vials containing 4 mL of scintillation fluid (4 g of New England Nuclear Omnifluor/L of toluene). Radioactivity was measured in a Beckman LS 7500 liquid scintillation counter (Beckman Instruments). All data are reported as nanomoles of [WITMG taken up per minute per milligram of total soluble protein. Chemostat Operation. The chemostat experiments were conducted in a Bioflo (New Brunswick Scientific Co., Inc.) culture apparatus (330-mL working volume). Temperature was maintained at 20 f 0.2 "C. The pH of the culture was maintained at 7.2 f 0.1 by the controlled addition of 0.5 N HC1 or 2 N NaOH (Cole-Parmer Model 5997-30 pH controller; Cole-Parmer Instruments Co., Chicago, IL). A drop of UCON LB-625 antifoam (Monsanto Co., Chesterfield, MO) was automatically added to the culture every 4-5 h using a timer. After inoculation, the culture was allowed to grow in batch mode for at least 36 h. The feed flow rate was then set to a dilution rate of 0.030 f 0.003 h-1 two-thirds of the maximum growth rate). Steady state was determined by allowing the uninduced culture to grow for at least five residence times and monitoring the cell concentration and size until a constant cell number and size were observed for over one residence time. The culture was induced by simultaneously adding IPTG to the culture medium and feed reservoir in a step change to yield a concentration of M.

Results Chemostat Culture Response to IPTG Induction. Measurements of the average cell mass (OD600) and cell number (Coulter counter) from a representative chemostat culture indicated that the culture experienced an unstable period approximately 24 h after induction (Figures 1 and 2). Culture density reached a minimum after approximately 65 h or two residence times. The O D w dropped from a steady-state, uninduced value of 3.5 to a minimum of 1.8, roughly corresponding to a loss in biomass of 0.85 g dry weight/L. The loss in cell number corresponded to 4.0 x 109 cells/mL (from 5.7 x lo9 to 1.7

Figure 2. Total cell count (Coulter counter) of RB79l(pKN) under continuous induction in a chemostat (D= 0.03 h-l, 20 "C). X lo9cellsiml). The culture density reached a new steady state of approximately 3.0 OD600 units or 5.5 X lo9 cells/ mL. The initial response of the culture to induction resulted in a sharp peak in total p-lactamase specific activity centered about 35 h, or roughly one residence time, after induction (Figure 3). Total specific activity peaked at 440 unitsimg of protein which represents roughly 13% of the total soluble cellular protein. Expression of p-lactamase remained elevated over the course of 90 h after induction. Interestingly, a second peak of p-lactamase specific activity, comparable in amplitude to the first peak, was observed 100 h later. However, the second peak was much broader, with elevated production spanning a period of 150 h. After prolonged cultivation in the presence of inducer (300 h or nine residence times), p-lactamase activity was reduced to the uninduced, basal level. Nearly 100% excretion was observed at the top of the first peak of 6-lactamase production. The level of p-lactamase excretion was lower in the second peak (75%), and excretion gradually was lost altogether in synchrony with the decrease in total activity. The average plasmid copy number of pKN was found to increase after IPTG induction (Figure4). Similar results have been observed in batch culture (Togna et al., 1992). Figure 4 shows the change in average copy number with the extent of p-lactamase total expression. Under the operating conditions of the experiment, at balanced growth, the uninduced average copy number of pKN varies between 10 and 20 copies/cell. Induction resulted in a maximum 10-fold increase in copy number. The average plasmid copy number gradually decreased after the first peak of /3-lactamase activity and was only about 3 times greater than the uninduced level a t the second peak of activity. The gradual loss of specific activity after the second peak in production was coupled with a decrease in average copy number, and after prolonged cultivation (350 h), the copy number of pKN was similar to that found for uninduced cells. Subpopulation Analysis. Four types of cell counts were measured: (1)the absolute cell or particle count; (2) the viable nonselective plate count; (3) the selective plate count, yielding an estimate of the number of viable plasmid-bearing cells; and (4) cells capable of forming visible colonies on neomycin plates containing 3 x M IPTG. Georgiou has determinedthat unaltered RB79l(pKN) cells are incapable of forming visible colonies on plates containing IPTG (Georgiou, 1987).

Blotechnol. hog., 1992, Vol. 8, No. 4

349

10 6 0

100

200

300

400

100

Hours a f t e r i n d u c t i o n

200

300

400

Hours after i n d u c t i o n

Figure 3. Total production (0) and excretion (A)of P-lactamase in a continuously induced chemostat (D= 0.03 h-l, 20 "C).

Figure 5. Estimate of plasmid retention in RB79l(pKN) based on selective phenotype (plating) during continuous culture in the presence of IPTG. Symbols: [O, nonselective count; A, selective count (D = 0.03 h-l, 20 "C).

4

, m 4

U

0

100

200

300

400

Hours a f t e r i n d u c t i o n

Figure 4. Average copy number (numbers beside points) variation associated with expression of (3-lactamase in a chemostat (D= 0.03 h-l, 20 "C).The steady-state average copy number of uninduced cells is 13.

The ability of the host strain to retain plasmid pKN in the presence of neomycin and IPTG was estimated by monitoring the viable cell concentrations on the basis of selective and nonselective plate counts. It is evident from Figure 5 that the number of cells capable of forming visible colonies on plates containing neomycin was about 10-fold lower than that of nonselectively plated cells in the region 50-100 h after induction, indicating the appearance of plasmid-free cells. This phenomenon was transient and occurred concurrently with the decrease in total cell density (see Figure 2). Throughout the rest of cultivation, the selective and nonselective plate counts were approximately equal indicating a negligible number of plasmid-free cells. Figure 6 shows that a subpopulation of variant cells capable of forming visible colonies on plates supplemented with both IPTG and neomycin exists in the culture prior to induction. This subpopulation initially represents 0.001% of the viable population and has been shown to be present in the master stock of RB79l(pKN) stored at -70"C(data not shown). This subpopulation can be easily selected in batch culture in medium containing 3 X M IPTG by using a small inoculum (10' cells/mL). Within

0

100

200

300

400

Hours a f t e r i n d u c t i o n

Figure 6. Population dynamics of induced RB79l(pKN) in a chemostat. Symbols: 0, particle count; A, selective count; +, IPTG-selective count (D= 0.03 h-l, 20 "C).

the chemostat, this subpopulation increased exponentially for about 35 h after induction (Figure 6). Exponential displacement of the culture by this variant continued for about 100 h. At this point, approximately 20% of the viable population consisted of IPTG/neomycin-resistant cells. The number of variant cells increased in a linear fashion until the selective and variant cell counta converged about 300 h after induction. In summary, the population dynamics can be divided into three events. (1)The first is the exponential increase in variant cell population, which roughly coincides with the drop in total cell concentration following induction and the first peak in 0-lactamase specific activity. (2) The second is the linear increase in variant cell concentration, which takes place during the second peak in 0lactamase production. (3) The third event is the convergence of the variant and selective cell counts, which coincides with the loss of high-level expression and the complete loss of excretion. Two distinct colony types, large and small, were observed on IPTG/neomycin plates. Although quantitative record-

344

keeping was not maintained, it was observed that the fraction of large-type variant colonies steadily increased throughout the experiment and gradually dominated the culture by the end of the run. Isolation of RB79l(pKN) lac Permease Mutants. IPTGineomycin-resistantvariants were isolated from the chemostat inoculum as described above. In shake flask experiments, these variant colonies produced 3-6 times less 0-lactamase than nonvariant colonies when induced with low4M IPTG (100 unitsimg versus 300-600 units/ mg). Low-producing cells would not recover the capacity to produce high levels of p-lactamase even after 44 generations before reinduction. We suspected that the variant cells within the inoculum contained a mutation that was causing low-level p-lactamase production. To determine whether this mutation was within the plasmid or within the host chromosome, plasmid was removed from variant cells and used to transform nonvariants. Six transformants were selected and tested for their ability to produce P-lactamase in batch culture. All the transformants produced between 400 and 800 units of 6-lactamasei mg of total protein. In a related experiment, five variant cell colonies and four nonvariant cell colonies were each cured of their plasmid using the acridine orange procedure. A single cured cell colony was selected from each of the nine acridine orange cultures, and these cells were transformed with plasmid pKN isolated from nonvariant cells. All four of the nonvariant cell colonies that had been cured of plasmid and then transformed with fresh plasmid pKN produced between 300 and 600 units of 0-lactamaseimg of total protein while all five of the transformed variant cell colonies produced only 150 unitsimg. These results show that a mutation within the host cell chromosome was responsible for causing low-level 0-lactamase production by these variants isolated from the chemostat inoculum. Mutant RB79l(pKN) cells isolated from the inoculum produced high levels of P-lactamase in batch culture when they were induced with 100 times the level of IPTG normally used, i.e., lo-* M rather than M (data not shown). It is known that lac permease of E. coli will concentrate @-D-galactosides(including IPTG) inside the cell to levels more than 50-fold higher than levels found outside the cell (Miller, 1972). Some 6-D-galactosidessuch as IPTG also enter the cell by nonspecific uptake in the absence of a functional lac permease. With nonspecific uptake of IPTG the concentration inside the mutant cell can be no higher than the IPTG concentration outside, and higher levels of inducer must be added to the medium to achieve full induction. To determine if the mutant RB79l(pKN) cells isolated from the chemostat inoculum were in fact lac permease mutants, uptake of methyl [W]-p-thiogalactopyranoside ([WITMG) by both original and mutant RB79l(pKN) cells was compared. TMG has a structure similar to that of IPTG, and like IPTG, TMG is a gratuitous inducer of the lac operon (Miller, 1978). TMG has been used extensively in the characterization of lac permease (Dills et al., 1980). Three mutant RB79l(pKN) cell colonies and two original RB79l(pKN) cell colonies were each cured of their plasmid. A single cured colony was selected from each of the acridine orange cultures, and [ 14C]TMGuptake experiments were performed with these cells as described in the Materials and Methods section. The original cell colonies took up 0.52 and 0.68 nmol of [l4C1TMG/min (mg of total protein)-l while the mutant colonies took up only 0.04, 0.04, and 0.05 nmolimin mg-', on average, 14fold less than the original cells. Therefore the variants

Biotechnol. Prog., 1992,Vol. 8, No. 4

isolated from the chemostat inoculum were lac permease mutants of RB79l(pKN). Isolation of Presumed RB79l(pKN) lac Superrepressor Mutants. A sample of the chemostat culture was collected after 350 h of operation and analyzed to determine the phenotype of the steady-state final population. The culture sample was inoculated into a shake flask containing Tanaka medium supplemented with 4.0 g/L glucose and 250 mg/L neomycin and cultivated at 20 "C to late exponential phase. Cells were spread on LB plates supplemented with 50 mg/L neomycin sulfate and either 3 x M or M IPTG; the selective plate count was also determined. The results showed that after 350 h the chemostat culture primarily consisted of cells resistance to both 3 X M and M IPTG. One colony from each of the plates containing M IPTG (five plates) was randomly selected and tested in batch culture for the ability to produce p-lactamase and 0-palactosidase. Neither P-lactamase nor @-galactosidase activity could be induced with M IPTG from these isolates. Under similar conditions, induction causes a 30fold increase in 0-lactamase specific activity with unaltered RB79l(pKN) cells and a 5-fold increase in 0-lactamase specific activity with RB791(pKN) lac permease mutants. @-Galactosidasespecific activity in wild-type E. coli has been observed to increase by 1000-fold in the presence of IPTG (Schleif, 1986). A l a d s , superrepressor, RB791(pKN) mutant would exhibit a phenotypic response to IPTG similar to that of the IPTG/neomycin-resistantcells isolated from the chemostat culture after 350 h of operation, Le., insensitivity to IPTG concentration and the inability to induce expression of both p-lactamase and 0-galactosidase. The repression of 0-galactosidase production precludes the possibility of promoter/operator or structural gene mutations since simultaneous mutations within both the p-lactamase and @-galactosidasegenes, which are plasmid- and chromosome-encoded,respectively, are likely to be extremely rare. Discussion The loss of high-level p-lactamase expression during continuous cultivation of E . coli RB79l(pKN) in a singlestage chemostat and within immobilized cell systems has been discussed previously (Chalmers et al., 1990; Georgiou et al., 1985). However, the mechanism responsible for this phenomenon was not established. Growth and product formation in a chemostat can be considered an extreme test of the robustness of a host-plasmid construction due to the strong selective nature of the device (Dykhuizen and Hartl, 1983). The mechanism for the instability of induced RB791(pKN) is evidently cell death. The appearance of IPTGresistance mutants can be considered a related phenomenon. In theory, if IPTG-resistant cells were absent from the culture, only the first peak in P-lactamase expression would occur followed by washout of the culture. Cell death, in this case, is characterized as loss of the ability to form visible colonies on nutrient agar plates (Figure 5). Furthermore, it appears that "dead" cells undergo some limited replication since the loss of biomass immediately after induction (Figures 1and 2) is slower than simple washout of nonreplicating cells (see Figure 7). The IPTG-resistant mutants cannot wholly explain the limited replication of cells during the washout since the mutant subpopulation accounts for no more than 4% of the total population during this period. No interpretation of whether nonplatable cells are capable of sustained target protein synthesis can be offered on the basis of the available data.

345

Blotechnol. Bog., 1992, Vol. 8, No. 4

I

"."

. 0

1

2

3

Residence times

Figure 7. Actual loss of biomass after induction compared to

predicted total washout (i.e., no growth). Symbols: 0,ODm/ (0Dw)o; -, washout (D = 0.03 h-l, 20 "C).

However, if cells are capable of some limited replication, then it is probable that some target protein is being synthesized. Cell death is likely associated with excessive plasmid DNA copy number (Figure 4) and/or overproduction of 0-lactamase. Togna et al. (1992) have demonstrated that transcriptional readthrough into the plasmid origin of replication from the tac promoter initiates runaway plasmid replication and is amplified by the dilution of lac1 repressor molecules as the copy number increases. Segregational plasmid loss does not significantly contribute to the observed culture instability. The brief plasmid instability associated with the rapid loss of biomass after induction can be attributed to washout of dead plasmid-containing cells and a basal level (e.g.,