( Springer-Verlag 1995

Appl Microbiol Biotechnol (1995) 44: 126—132

OR I G I N A L P AP E R R. O’Kennedy · C. J. Houghton J. W. Patching

Effects of growth environment on recombinant plasmid stability in Saccharomyces cerevisiae grown in continuous culture

Received: 6 September 1994/Received last revision: 1 March 1995/Accepted: 6 March 1995

Abstract A recombinant strain of Saccharomyces cerevisiae, containing a 2-lm-fragment-based plasmid (pYEaa4) was grown under non-selective conditions in continuous culture. The decrease in the population carrying the plasmid-encoded auxotrophic marker, ¸Eº2, was examined under different physiological conditions. The difference in growth rate (Dk) between plasmid-free and plasmid-containing cells and the rate of plasmid segregation (R) were determined using a non-linear regression technique. Loss rates were greater in defined glucose-limited cultures than in complex glucose-limited cultures. Plasmid loss was Dkdominated in cultures grown on defined media whereas Dk and R were co-dominant in cultures grown on complex medium. Loss rates increased with increasing dilution rate in complex glucose-limited cultures. The reverse was found in defined glucose-limited cultures. Plasmid retention and loss kinetics determined from defined magnesium-limited cultures were not significantly different from those observed in defined glucoselimited cultures. Although plasmid retention in defined phosphate-limited culture was not significantly different from that in defined glucose-limited culture, reduced R and increased Dk indicated an alternative physiological effect of phosphate limitation on plasmid stability.

Introduction Instability of recombinant vectors in commercially important organisms represents a significant obstacle to R. O’Kennedy · J. W. Patching ( ) Department of Microbiology, University College Galway, Ireland. Fax: #353 91 725005 C. J. Houghton Department of Mathematical Physics, University College Galway, Ireland

the high-level production of heterologous proteins. Structural rearrangements and segregational instability can both result in the loss of the gene of interest. Segregational instability occurs when a plasmid-bearing host fails to pass on the plasmid to daughter cells. The population, consequently, consists of two cell types, plasmid-free and plasmid-containing. The higher metabolic load resulting from plasmid replication and expression of plasmid-borne functions may reduce the growth rate of the plasmid-containing host. Because of this reduced growth rate, competition between the two subpopulations amplifies plasmid loss. Estimations of the rate of plasmid segregation (R) and the difference in growth rate (Dk) between plasmid-free and plasmidcontaining cells have been carried out on a variety of host/vector systems (for examples see: Caulcott et al. 1987; Fleming and Patching 1994). The effects of growth environment on plasmid loss kinetics have been studied less in yeast than in bacterial systems. The effect of growth rate on the loss kinetics of 2 lm-based vectors in glucose-limited chemostat cultures has been previously investigated. The full 2 lm-based vector, pJDB248, was more stable at high dilution rates than at low dilution rates (Kleinman et al. 1986). Decreased dilution rates resulted in elevated copy number in the plasmid-containing subpopulation and it was suggested that an increased rate of plasmid segregation was the cause of increased instability at the low dilution rate (Bugeja et al. 1989). Conversely, the 2 km-fragment-based vector, pLG669-z was more stable at low dilution rates, primarily because of decreased rates of plasmid segregation (Impoolsup et al. 1989). These studies appear to conflict with each other. How can host/vectors that are based on essentially the same replicon exhibit different loss kinetics? The studies on pJDB248 used defined medium whereas studies on pLG669-z used complex medium. This suggests that there may have been a substantial environmental influence on the loss kinetics of the two vectors.

127

The purpose of this study was to assess the effects of environmental conditions on the stability of a 2 lmfragment-based plasmid in Saccharomyces cerevisiae. We demonstrate that, depending on the medium used, a 2 lm-fragment-based host/plasmid system can exhibit two different responses to dilution rate. Conditions that decreased plasmid loss were found and analysis showed that Dk and R were dependent on environmental conditions and dilution rate.

Materials and methods Strain

clearing around colonies grown on YPDAmy agar. Samples removed from the start and end of chemostat and batch cultures were screened for structural instability. No evidence of segregation of the Leu` and Amy` phenotypes was found.

Batch stability and growth rate determinations Batch stability determinations were carried out in triplicate 10-ml serial cultures grown at 30° C and shaken at 200 rpm. Plasmid retention was determined for cultures grown on CCM, YPD or WDMnon-selective media containing 20 g/l glucose as a carbon source. Every 24 h, about 2]105 cells were subcultured into fresh medium. The proportion of plasmid-containing cells was determined at the start of the experiment and after 96 h (35—46 generations). Segregational loss rates (S ) over t generations were calculated from L proportions of plasmid-containing cells Eq. 1, where p`, p` are the t 0 at time"t and time"0 respectively (Longtine et al. 1993). This parameter describes the overall loss rate and so is a combination of the effects of both Dk and R.

The original host, AH22 (MATa leu2-3 leu2-112 his4-519 [cir`]), was obtained from Nadia Danash (University of Manchester, England). The host/plasmid combination, AH22 (MATa leu2-3 leu2-112 his4519 [cir`]) pYEaa4 (12.67 kb) (ALKO2802), was obtained from E. Va¨isa¨nen by kind permission of Alko laboratories (Helsinki, Finland) and VTT/Bio (Espoo, Finland). The plasmid consisted of the coding region of Bacillus amyloliquifaciens a amylase placed between the ADH1 promoter and terminator in pAAH5, a b-lactamase gene for selection in Escherichia coli and a LEU2 marker for selection in Saccharomyces cerevisiae (Ruohonen et al. 1991).

S "2 1!exp L

Media

Continuous cultures

Defined medium (WDM) consisted of full-strength Wickerham defined medium (Wickerham 1946) with the following additions (l~1) : 20 g D-glucose 20 mg L-histidine 20 g agar (where necessary) (Oxoid Bacteriological no. 3, Great Britain). YPD medium was as described by Guthrie and Fink (1991). Colonies were screened for amylase activity on YPDAmy agar, consisting of YPD plate medium containing four Phadebas amylase test tablets (Pharmacia Diagnostics AB, Sweden) /100 ml. Defined chemostat medium (DCMnon-selective) consisted of one-half-strength Wickerham defined medium (Wickerham 1946) plus L-histidine (20 mg l~1). Adjustments were made to ensure glucose, phosphate or magnesium limitation. Nutrient limitation was confirmed by the method of Goldberg and Er-el (1981). The changes applied to the medium were (l~1) glucose limitation: D-glucose 0.5 g; phosphate limitation: D-glucose 5 g, KH PO 2 mg, KNO 200 mg; 2 4 ·7H O 25 mg. 3 Commagnesium limitation: D-glucose 5 g, MgSO 4 21988) contained plex chemostat medium (CCM) (Lee and Hassan (l~1) 0.5 g D-glucose, 0.5 g bactopeptone (Oxoid), 5 g yeast extract (Oxoid). All media were adjusted to pH 6.0$0.5 with 0.1 M succinic acid and NaOH. Non-selective defined media (WDMnon-selective and DCMnon-selective) were supplemented with L-leucine (30 mg l~1). When selective conditions were required, L-leucine was omitted.

Continuous cultures (500 ml) were carried out in the apparatus described by Fleming et al. (1988). Cultures were aerated at 1.5 vvm. Temperature was maintained at 30$0.5° C. The succinic acid buffer maintained the pH of the cultures at 6.0$0.5. Vegetable oil (Dunnes Stores, Ireland) was added at regular intervals (75 ll every 6 h) to suppress foaming when using CCM. Samples were removed at intervals via a port on the side of the vessel, into sterile syringes, and the proportions of plasmid-containing cells were determined.

Determination of the proportion of plasmid-containing cells Samples were diluted to give between 100 and 200 colonies/plate and spread onto YPD plates, which were incubated for 48—72 h at 30° C. At least 100 isolated colonies were picked onto WDMselective and WDMnon-selective media to determine the proportion of plasmid-containing cells. Structural instability of pYEaa4 was screened for by checking the isolates for retention of amylase activity. Amylase activity was defined by the appearance of zones of

C

A

1n(p`/p`) t 0 t

BD

(1)

Growth rates were determined in 500-ml baffled flasks with 100 ml medium. The rate of change in absorbance at 600 nm was monitored over time. The maximum specific growth rate, k , was calculated .!9 from least-squares linear regression.

Model development and data analysis Data analysis was essentially that proposed by Davidson et al. (1990) using the plasmid loss model described by Lenski and Bouma (1987). Preliminary estimates for Dk (generation~1) and R (generation~1) were made using the linear regression method of Cooper et al. (1987). The parameters Dk, R and p` were then varied by 0 data to the model, non-linear regression to fit the experimental Eq. (2), where p` is the model prediction for the proportion of p` cells at time tt. p`(Dk#R) p`" 0 (2) t (p~Dk#R) e (Dk#R)t#Dkp` 0 0 Confidence limits on Dk (generation~1) and R (generation~1) were determined using the modified ‘‘general jack-knife’’ method of Robinson (1985) as described by O’Kennedy et al. (1994). Jackknifing ignores single data pairs in a dataset and then repeats non-linear regression. For each data pair that is ignored (generations and p` data in this case), non-linear regression generates a new set of estimates (called pseudovalues) for R, Dk and p`. Conse0 p`) will quently, a data set with N data pairs (generation versus generate N groups of estimates of R, Dk and p`. These pseudovalues 0 with N!3 degrees are then used to calculate 95% confidence limits, of freedom.

128

Results Batch cultures After 48 h batch culture under selective conditions (WDMselective) 86% of the population was plasmidcontaining. Segregational loss rates (S ) derived from L non-selective batch cultures showed that pYEaa4 was most stable in CCM (Table 1). There was no significant difference between the segregational loss rates of cultures grown on WDMnon-selective and YPD. Batch cultures of ALKO2802 grown on CCM showed a significant growth rate advantage over the original host while the growth rates of AH22 and ALKO2802 cultures demonstrated that plasmid carriage imposed a significant burden on the host during culture on WDMnon-selective medium (Table 1). Addition of leucine to CCM cultures did not significantly affect S (0.016$0.008 generation~1) demonstrating that seL lection because of leucine deficiency was not responsible for the improved stability. Plasmid carriage imposed no significant burden in cultures grown on YPD medium. Chemostat cultures

Fig. 1 Plasmid loss by ALKO2802 in glucose-limited continuous cultures on defined chemostat medium (DCMnon-selective), grown at high dilution rate (D"0.17 h~1, n) and low dilution rate (D"0.05 h~1m). Error bars standard error in the proportion of Leu` colonies. Best-fit lines are derived from Dk, R and p` deter0 mined from jack-knifed non-linear regression (see Materials and methods)

Glucose-limited defined medium (DCMnon-selective) Plasmid loss by ALKO2802 in glucose-limited DCM chemostat cultures at dilution rates of 0.17 h~1 and 0.05 h~1 was investigated (Fig. 1). Loss kinetics were Dk dominated (Table 2). Segregational loss rates and remodelling derived from Dk and R determined from non-linear regression and p`"0.86, indicated that in0 stability was greater at a dilution rate D of 0.05 h~1 (Table 2; Fig. 4). Significant variations in Dk and R were observed with changes in dilution rate. Dk (generation~1) decreased and R increased with increasing dilution rate. The zero value of R at D"0.05 h~1 indicated that the detection limit of the non-linear regression method was limited to plasmid segregation rates greater than 4.6]10~3 generation~1.

To investigate the effect of nutrient-rich medium on rates of plasmid loss, glucose-limited CCM chemostat cultures at dilution rates of 0.15 h~1 and 0.07 h~1 were monitored (Fig. 2). Segregational loss rates (S ) were L greater at the higher dilution rate (Table 2; Fig. 4). The rate of plasmid segregation (R) was approximately 30fold higher in complex glucose-limited cultures than in defined glucose-limited cultures at high dilution rates and at least 20-fold higher at low dilution rates. With complex medium, Dk was negative at both dilution rates, which indicated that plasmid-containing cells had a growth advantage over plasmid-free cells. In

Table 1 Growth rates and plasmid retention in batch culture, Maximum specific growth rates (k ) were determined during the first 8 h .!9 of non-selective culture and are shown with their SEM. Figures in parantheses show the number of determinations carried out. The percentage of the population of ALKO2802 that was Leu` at the beginning and the end of the initial 8 h-growth period (about 85% and 83% respectively) was not significantly different. Levels of

significance (in parentheses) for differences in growth rates between AH22 (the plasmid-free host) and ALKO2802 (Dk ) were deter.!9 mined by Student’s t-test. Levels of significance below 90% were considered not significant (NS). Segregational loss in ALKO2802 (S ) was calculated after 96 h of serial batch culture (see Materials L and methods). ¼DM Wickerhan defined medium, CCM complex chemostat medium, ½PD medium see Guthrie and Fink (1991)

Glucose-limited complex medium (CCM)

AH22

ALKO2802

Medium

k $SEM .!9 (h~1)

k $SEM .!9 (h~1)

102]D k .!9 (h~1)

Segregational loss (S ) (generation~1) L

WDMnon-selective CCM YPD

0.293$0.004 (4) 0.355$0.004 (4) 0.354$0.002 (4)

0.259$0.003 (8) 0.385$0.010 (8) 0.338$0.012 (8)

3.4 ('95%) !3.0 ('95%) 1.6 (NS)

0.045$0.008 0.015$0.010 0.047$0.011

129 Table 2 Plasmid loss kinetics of ALKO2802 grown in chemostat culture. Values of Dk and R are expressed as the parameter$95% confidence limits, calculated from the jack-knife pseudovalues (see Materials and methods). The segregational loss rate (S ) is L

based on non-linear regression predictions after 40 generations, where the initial proportion of plasmid-containing cells (p`)" 0 0.86. DCM defined chemostat medium, CCM complex chemostat medium

Medium and limitation

D (h~1)

S L (generation~1)

102]Dk (generation~1)

102]R (generation~1)

102]Dk (h~1)

DCM, glucose

0.17 0.05 0.15 0.07 0.16 0.17

0.079 0.169 0.131 0.064 0.126 0.098

8.35$1.06 13.7$0.20 !8.50$0.00 !10.6$0.00 11.3$0.39 8.44$1.27

0.46$0.61 0.00$0.00 13.2$0.00 9.6$0.00 0.00$0.00 0.49$1.06

2.04$0.27 0.99$0.02 !1.60$0.28 !0.95$0.16 2.56$0.09 2.07$0.30

CCM, glucose DCM, phosphate DCM, magnesium

Fig. 2 Plasmid loss by ALKO2802 in glucose-limited continuous cultures on complex chemostat medium (CCM), grown at high dilution rate (D"0.15 h~1, h) and low dilution rate (D"0.07 h~1, j). Error bars standard error in the proportion of Leu` colonies. Best-fit lines are derived from Dk, R and p` deter0 mined from jack-knifed non-linear regression (see Materials and methods)

Fig. 3 Plasmid loss by ALKO2802 in magnesium-limited (D"0.17 h~1, v) or phosphate-limited (D"0.16 h~1, j) continuous culture on defined medium (DCMnon-selective). Error bars standard error in the proportion of Leu` colonies. Best-fit lines are derived from Dk, R and p` determined from jack-knifed non-linear 0 regression (see Materials and methods)

glucose-limited complex medium (D"0.15 h~1), R was dominant over Dk and remodelling predicted complete loss of plasmid (%Leu`(5%) after 45 generations. At D"0.07 h~1, R was no longer dominant and a low level of coexistence (Leu` around 9%) was predicted by remodelling (Fig. 4). In both DCM and CCM chemostat cultures, the absolute values of Dk (DDkD) increased with decreasing dilution rate.

cultures on magnesium-limited DCM did not exhibit significant differences in plasmid loss kinetics or predicted loss (Fig. 4; Table 2). Loss kinetics in phosphateand magnesium-limited culture were dominated by Dk. Instability was higher in phosphate-limited culture than in equivalent glucose-limited culture. Instability under phosphate limitation was completely attributed to Dk where R was below the detection limits of the system.

Magnesium-limited (DCMmagnesium) and phosphate-limited (DCMphosphate) cultures on defined medium

Discussion

When compared with glucose-limited DCM cultures at equivalent dilution rates (Table 2; Fig. 3), chemostat

Batch cultures of ALKO2802 were most stable when grown on CCM medium. This correlated with a growth rate advantage for plasmid-containing cells. Improved

130

Fig. 4 Remodelled stability of ALKO2802 grown in chemostat culture under different conditions (see Materials and methods). Glucose limitation: DCMnon-selective (defined) medium, D"0.17 h~1 (n) and D"0.05 h~1 (m); glucose limitation: CCM (complex) medium, D"0.15 h~1 (h) and D"0.07 h~1(j); magnesium limitation: DCMnon-selective medium, D"0.17 h~1 (v); phosphate limitation: DCMnon-selective medium, D"0.16 h~1 (j)

stability was not related to increased nutrient levels since cultures grown on YPD, a richer medium, had a significantly higher segregational loss rate yet no significant growth rate difference between p` (plasmidcontaining) cells and the p~ (plasmid-free) host. Addition of leucine to CCM cultures of ALKO2802 did not affect S , demonstrating that leucine deficiency was not L responsible for selecting p` cells. From this we conclude that an optimal medium composition exists that will support enhanced plasmid stability. This optimum appears to be between a nutrient-poor defined medium (WDMnon-selective) and a nutrient-rich complex medium (YPD). The kinetic parameters governing plasmid loss by ALKO2802 and the influence of dilution rate on them show significant variation with medium type. For glucose-limited cultures, pYEaa4 was more stable at the higher dilution rate on complex medium (CCM) where Dk and R were codominant, while pYEaa4 was more stable at the lower dilution rate on defined medium (DCM) where Dk was the dominant kinetic parameter (Table 2; Fig. 4). A direct relationship between dilution rate and R, previously observed by Impoolsup et al. (1989), was also observed by us. This relationship was independent of the medium even though the response of S to dilution rate was different on the different L media. Many bacterial host/plasmid systems have shown improved stability at low dilution rates (Noack et al. 1981; Seigul and Ryu 1985; Caulcott et al. 1987;

Wei et al. 1989). These systems exhibited increasing Dk with decreasing dilution rate but unlike yeast/ 2 lm-based vector systems, R increased with decreasing dilution rate suggesting that causes of plasmid segregational instability in S. cerevisiae are inherently different from those in bacteria. Other 2 lm-fragment-based and full-2 lm-based vectors can exhibit increased stability with increasing dilution rate (Dibiasio and Sardonini 1986; Bugeja et al. 1989). To provide a comparison with our results, we have calculated values of Dk and R from observations of plasmid loss in S. cerevisiae S150-2B pJDB248 during continuous culture under similar glucose-limited conditions, which were made by Bugeja et al. (1989) (Table 3). As with our results, S150-2B pJDB248 exhibited a decrease in competitive ability (Dk, generation~1) with increased dilution rate, which correlated with decreased S . As with loss of pYEaa4 or pLG669-z L (Impoolsup et al. 1989), R also increased with increased dilution rate. The nature of the dilution-rate-induced change in Dk was not investigated by us, but Bugeja et al. (1989) observed a greatly increased copy number at the low dilution rate, which could have been responsible for the increased value of Dk derived from their results. This increase in copy number would explain the lower value of R derived from their observations at a low dilution rate, and could also explain the similar dilution-rate-related changes in Dk and R observed by us. The effect of medium type on plasmid loss in bacteria has also been examined by some groups who have concluded that growth on complex medium can stabilise plasmids (Summers 1991; Fleming and Patching 1994). As with batch cultures, chemostat cultures grown on CCM demonstrated a significant growth advantage for plasmid-containing cells (Table 2). This improvement was not reflected in a lower segregational loss rate (S ) as it was in batch cultures, because L it was offset by at least a 20-fold increase in R. Growth on complex medium also reduced competition (Dk, generation~1) between recombinant and plasmidfree Bacillus subtilis in starch-limited chemostat cultures (Fleming and Patching 1994), which may reflect the greater demand made by the plasmid on the pool of cellular intermediates when nutrient restriction is more severe (Matsui et al. 1990). In contrast to the effects of medium type, differences in plasmid loss rates induced by different limitations were minimal (Table 2). The similarity of segregational loss rates and the underlying kinetic parameters under magnesium- or glucose-limited culture on minimal medium would suggest that the physiological effects of the limitations were also similar. Magnesium is known to be important for the correct functioning of key enzymes in the yeast glycolytic pathway (Berry and Brown 1987) so that magnesium limitation may reduce the flow of carbon and energy within cells in the same way as external glucose limitation.

131 Table 3 Plasmid loss kinetics of S 150-2B [cir0] pJDB248 calculated from the original data of Bugeja et al. (1989).Values of Dk and R are expressed as the parameter$95% confidence limits,

calculated from the jack-knife pseudovalues (see Materials and methods). The segregational loss rate (S ) is based on non-linear L regression predictions after 40 generations where p`"0.99 0

Medium and limitation

D (h~1)

S L (generation~1)

102]Dk (generation~1)

102]R (generation~1)

102]Dk (h~1)

WDM, glucose

0.12 0.05

0.032 0.138

!6.84$0.45 12.8$0.76

3.95$0.09 0.84$0.30

!1.19$0.14 0.93$0.87

Kinetic parameters derived from phosphate-limited cultures deviated from those determined for comparable cultures under glucose limitation. Reduction in R along with increased Dk also occurs when S150-2B pJDB248 is cultured at low dilution rates in glucoselimited culture (Table 3). Carbon or phosphate limitation leads to increased instability of high-copy-number plasmids and large plasmids in E. coli hosts (Wouters et al. 1980; Noack et al. 1981; Jones and Melling 1984). Caulcott et al. (1987) demonstrated that low-copynumber vectors exhibit a different response to limitation. They suggest that high metabolic loads imposed by plasmids may be subject to certain limitations. This would suggest that phosphate limitation affects segregation mechanisms resulting in higher copy number, decreasing R but increasing the burden on the host. This study demonstrates the importance of taking environmental factors, especially medium type, into account when studying plasmid loss in continuous cultures. Published studies on S. cerevisiae would indicate fundamental differences between the stability characteristics of full-2 lm-based and 2 lmfragment-based vectors (Kleinman et al. 1986; Lee and Hassan 1988; Bugeja et al. 1989; Impoolsup et al. 1989). In this study, the loss of pYEaa4 was Dk dominated in cultures on defined medium. Growth on complex medium (CCM) alleviated this plasmid-imposed burden and plasmid-containing cells had a growth advantage in both batch and chemostat cultures. Rates of plasmid segregation (R) increased in CCM chemostat cultures, however, so that R and Dk were codominant. Comparison of our results and those of Bugeja et al. (1989) shows that the stability characteristics of full2 lm-based and 2 lm-fragment-based vectors are similar, if the same types of medium and limitation are used. Experimental systems analysed in this paper and in others (Bugeja et al. 1989; Impoolsup et al. 1989) demonstrate that R consistently increases with dilution rate. Increases in R do not necessarily increase overall segregational loss rates (S ). It would appear that if L other yeast host/plasmid systems are Dk-dominant, they will be more stable at high dilution rates, whereas systems that are R-dominant will be more stable at low dilution rates. Acknowledgements Many thanks to Dr. S. Keranen for providing information on ALKO2802 and Dr. V. Bugeja for per-

mission to use the pJDB248 plasmid loss data for non-linear regression.

References Berry DR, Brown C (1987) Physiology of yeast growth. Yeast Biotechnology. Allen & Unwin, London, pp 161—199 Bugeja VC, Kleinman MJ, Stanbury PF, Gingold EB (1989) The segregation of the 2 lm-based yeast plasmid breaks down under conditions of slow, glucose limited growth. J Gen Microbiol 135:2891—2897 Caulcott CA, Dunn A, Robertson HA, Cooper NS, Brown ME, Rhodes PM (1987) Investigation of the effect of the growth environment on the stability of low-copy-number plasmid in Escherichia coli. J Gen Microbiol 133:1881—1889 Cooper NS, Brown ME, Caulcott CA (1987) A mathematical model for analysing plasmid stability in microorganisms. J Gen Microbiol 133:1871—1880 Davidson AM, Dunn A, Day MJ, Randerson PF (1990) A nonlinear technique for the analysis of plasmid instability in microorganisms. J Gen Microbiol 136:59—64 Dibiasio D, Sardonini C (1986) Stability of continuous culture with recombinant organisms. Ann N Y Acad Sci 469:111—117 Fleming G, Dawson MT, Patching JW (1988) The isolation of strains of Bacillus subtilis showing improved plasmid stability characteristics by means of selective continuous culture. J Gen Microbiol 134:2095—2101 Fleming GT, Patching JW (1994) Plasmid instability in an industrial strain of Bacillus subtilis grown in continuous culture. J Ind Microbiol 13:106—111 Goldberg I, Er-el Z (1981) The chemostat- an efficient technique for medium optimization. Process Biochem 16:2—8 Guthrie C, Fink GR (eds) (1991) Guide to yeast genetics and molecular biology. Methods Enzymol 194:3—21 Impoolsup A, Caunt P, Greenfield PF (1989) Effect of growth rate on stability of a recombinant plasmid during continuous culture of Saccharomyces cerevisiae in non-selective medium. J Biotechnol 10:171—180 Jones SA, Melling J (1984) Persistence of pBR322-related plasmids in Escherichia coli grown in chemostat culture. FEMS Microbiol Lett 22:239—243 Kleinman MJ, Gingold EB, Stanbury PF (1986) The stability of the yeast plasmid pJDB248 depends on the growth rate of the culture. Biotechnol Lett 8:225—230 Lee FS, Hassan HM (1988) Stability and expression of a plasmidencoded killer toxin cDNA in batch and chemostat cultures of Saccharomyces cerevisiae. Biotechnol Bioeng 31:783—789 Lenski RE, Bouma JE (1987) Effects of segregation and selection on instability of plasmid pACYC184 in Escherichia coli B. J Bacteriol 169:5314—5316 Longtine MS, Enomoto S, Finstad SL, Berman J (1993) Telomeremediated plasmid segregation in Saccharomyces cerevisiae involves gene products required for transcriptional repression at silencers and telomeres. Genetics 133:171—182

132 Matsui T, Sato H, Sato S, Mukataka S, Takahashi J (1990) Effect of nutritional conditions on plasmid stability and production of tryptophan synthase by a recombinant Escherichia coli. Agric Biol Chem 54:619—624 Noack D, Roth M, Geuther R, Muller G, Undisz K (1981) Maintenance and genetic stability of vector plasmids pBR322 and pBR325 in Escherichia coli K12 strains grown in a chemostat. Mol Gen Genet 184:121—124 O’Kennedy R, Camplisson S, Patching JW (1994) Microsoft Excel Macros for analysing plasmid instability and their application to recombinant strains of Saccharomyces cerevisiae. Binary 6:204—216 Robinson JA (1985) Determining microbial kinetic parameters using nonlinear regression. In: Advances in Microbial Ecology. Plenum Press, New York, pp 61—114 Ruohonen L, Penttila M, Keranen S (1991). Optimization of Bacillus a-amylase production by Saccharomyces cerevisiae. Yeast 7: 337—346

Seigul R, Ryu DDY (1985) Kinetic study of instability of recombinant plasmid pPLc23 trpA1 in E. coli using a two-stage continuous culture system. Biotechnol Bioeng 27:28—33 Summers DK (1991) The kinetics of plasmid loss. Trends Biotechnol 9:273—278 Wei D, Parulekar SJ, Stark BC, Weigand WA (1989) Plasmid stability and a-amylase production in batch and continuous cultures of Bacillus subtilis TN106[pAT5]. Biotechnol Bioeng 33:1010—1020 Wickerham LJ (1946) A critical evaluation of the nitrogen assimilation tests commonly used in the classification of yeasts. J Bacteriol 52:293—301 Wouters JTM, Driehuis FL, Polaczek PJ, Oppernraay MLHA van, Andel JG van (1980) Persistence of the pBR322 plasmid in Escherichia coli K12 grown in chemostat cultures. Antonie van Leeuwenhoek 46:353—362