Spontaneous cAMP-dependent derepression of gene expression in stationary phase plays a role in recombinant expression instability

Gene 209 (1998) 95–103

Spontaneous cAMP-dependent derepression of gene expression in stationary phase plays a role in recombinant expression instability Trudy H. Grossman, Ernest S. Kawasaki, Sandhya R. Punreddy, Marcia S. Osburne * Procept, Inc., Department of Molecular Biology, 840 Memorial Drive, Cambridge, MA 02139, USA Received 19 June 1997; received in revised form 2 December 1997; accepted 8 December 1997; Received by J.A. Engler

Abstract E. coli recombinant expression systems that utilize lac operon control elements to modulate gene expression are known to produce some amount of uninduced ( leaky) gene expression. Previously, we showed that high levels of uninduced gene expression was a major cause of instability in the pET expression system. We show here that the pET system, in which the phage T7 RNA polymerase gene is expressed via lac operon control elements, exhibits leaky expression that increases markedly as cells grown in complex medium enter stationary phase. Moreover, we found that this phenomenon occurs with the chromosomal lac operon as well. Further investigation revealed that stationary phase leaky expression requires cyclic AMP, and that substantial leaky expression could be effected in log phase cells by adding cyclic AMP and acetate at pH 6.0. Finally, a comparison of otherwise isogenic cya and wild-type hosts showed that expression stability and plasmid maintenance in the cya host is greatly enhanced, even when cells are passaged repeatedly in non-selection medium. These findings both provide a method to enhance the stability of lac-based recombinant expression systems, and suggest that derepression of the lac operon in the absence of inducer may be part of a general cellular response to nutrient limitation. © 1998 Elsevier Science B.V. Keywords: T7 expression vectors; Leaky gene expression; lac Operon regulation

1. Introduction E. coli lac operon control elements are widely used to regulate recombinant gene expression systems. The pET system, for example, which exploits the high processivity of T7 RNA polymerase to optimize recombinant gene expression (Studier and Moffatt, 1986) is ultimately regulated by the lac repressor protein, LacI, which controls expression of the T7 RNA polymerase gene. Although pET vectors have been used extensively to produce high levels of recombinant protein in E. coli (Makrides, 1996; Kelley et al., 1995), problems of * Corresponding author. Tel.: +1 617 4940400; Fax: +1 617 491 5310; E-mail: [email protected] Abbreviations: Ap, ampicillin; cya, a mutation in the gene encoding adenylate cyclase; cAMP, cyclic adenosine 3∞, 5∞-monophosphate; CFU, colony-forming units; E. coli, Escherichia coli; IPTG, isopropylb--thiogalactoside; Km, kanamycin; LBAp, Luria–Bertani medium containing 50 mg of Ap per milliliter; ONPG, o-nitrophenyl b--galactopyranoside; A , optical density at 600 nanometer wavelength; 600 SDS–PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis; Tc, tetracycline; wt, wild type; X-gal, 5-bromo-4-chloro-3 indolyl b--galactoside. 0378-1119/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PII S 03 7 8 -1 1 1 9 ( 9 8 ) 0 0 0 20 - 1

plasmid or expression instability do sometimes arise (Novagen, 1995; Kelley et al., 1995). The lacUV5 promoter, which drives expression of the T7 RNA polymerase gene, is known to allow a ‘basal level’ of transcription in uninduced cells (Dubendorff and Studier, 1991), resulting in the transcription of genes whose products may be toxic. In such instances, cells with intact expression systems may be counterselected. Measures to control uninduced gene expression include additional lac control elements (Dubendorff and Studier, 1991), or the gene encoding T7 lysozyme (Moffatt and Studier, 1987). Despite these measures, expression stability remains a problem in some instances. Recently, we showed that high levels of uninduced protein production could be obtained from certain pET vectors, in some cases higher than that seen following IPTG induction ( Kelley et al., 1995). This phenomenon occurred in stationary phase for cells growing in some complex media. Our current results show that conditions under which uninduced (‘leaky’) expression of the pET system was enhanced also led to increased leaky expression of the chromosomal lac operon, that leaky expression in both systems was maximal at the onset of

96

T.H. Grossman et al. / Gene 209 (1998) 95–103

stationary phase in complex medium, and that cyclic AMP (cAMP) was required for this phenomenon to occur. The requirement for cAMP in leaky expression was further substantiated by showing that a cya mutation in the host strain conferred increased expression stability.

2. Materials and methods 2.1. E. coli strains and plasmids E. coli strains BL21(DE3) (F−, ompT, hsdS , gal, B dcm) and BL26(DE3) (BL21(DE3), Lac− (uncharacterized mutation)), and plasmid pET-11a (Apr), were obtained from Novagen (Madison, WI ). Plasmid pIQ2 ( Tcr), encoding a lacI q allele, was a gift from Philip Silverman (Grossman and Silverman, 1989). All restriction endonucleases and chemicals were from standard commercial sources, unless noted. Plasmid pET-11alacZ contains a 3.4 kb NdeI/BamHI lacZ fragment, generated by PCR amplification of the full-length lacZ gene of plasmid pSVb (Invitrogen, San Diego, CA), and ligated into the NdeI and BamHI sites of plasmid pET-11a. The resulting plasmid conferred a blue colony phenotype to strain BL26(DE3), but not to strain BL26, in the presence of 40 mg/ml X-gal (Sigma, St. Louis, MO), confirming the dependence of b-galactosidase production on T7 RNA polymerase. Transcription of the recombinant gene ligated into pET-11a is from the ‘T7lac’ promoter, which consists of a T7 promoter immediately followed by the lac operator. pET-11a also encodes the lacI gene to provide additional lac repressor. Plasmid pET-11a-CD58 contains a 300 bp DNA fragment encoding the adhesion domain of the human CD58 receptor protein, huCD58 ( Wallner et al., 1987), inserted into the NdeI and BamHI sites of pET-11a. Expression of the huCD58 gene was also dependent on the DE3 lysogen in host strain BL21(DE3). Strains BL21rpoS::kan(DE3), BL26rpoS::kan(DE3), BL21cya::kan (DE3) and BL26cya::kan(DE3) were constructed by P1 transduction as described by Miller (1992), using ZK1000rpoS::kan (Bohannon et al., 1991, obtained from R. Kolter) and SP850cyaA1400::kan (Shah and Peterkofsky, 1991; obtained from Coli Genetic Stock Center, New Haven, CT ) as donor strains. P1kc+ was isolated from E. coli lysogen KL739 (Coli Genetic Stock Center). Transductants of BL21(DE3) and BL26(DE3) were selected by resistance to kanamycin ( Km). The RpoS− phenotype was verified by the decreased ability of mutants to hydrolyze H O 2 2 (Lange and Hengge-Aronis, 1991). 2.2. Media, growth conditions and transformations Cells were grown at 37°C with vigorous shaking. LB medium was from BBL (Cockeysville, MD). 4xYT

medium contained per liter: 32 g Bacto tryptone (Difco, Detroit, MI ), 20 g Bacto yeast extract (Difco, Detroit, MI ) and 5 g NaCl. Where stated, cultures were buffered with 0.05 M Tris and adjusted to either pH 6.0 or pH 7.0 with hydrochloric acid or sodium hydroxide. Ampicillin (Ap, 50 mg/ml ), tetracycline ( Tc, 10 mg/ml ), and Km (25 mg/ml ) were added as required. Cultures were routinely inoculated with seed stocks derived from early log phase cells (A =0.1–0.5) frozen in 15% glycerol at 600 −80°C. 4xYT medium was assayed for lactose content using the Lactose/-Galactose detection kit (Boehringer Mannheim, Indianapolis, IN ). Lactose was determined to be below detectable levels (<0.002%). To destroy any lactose present in the medium, 1000 units of E. coli bgalalactosidase (Sigma, St. Louis, MO) were added per 100 ml of 4xYT. Treatment, determined to be sufficient to digest 0.1% lactose added to 100 ml 4xYT, was for 3 h at room temperature, with gentle shaking. The treated medium was then autoclaved for 20 min and passed through a 0.2m filter to remove protein aggregates. No detectable b-galactosidase activity remained in the medium after autoclaving. This treated medium was then used to grow cultures, as described. Transformations using calcium chloride were carried out according to Miller (1992). Where acetate was added to cultures, potassium acetate was used. Induction of transcription from the lac promoter/operator was with 0.5 mM IPTG (Sigma, St. Louis, MO). 2.3. SDS–PAGE analysis Whole cell lysates were analyzed by SDS–PAGE using either 15% polyacrylamide gels (BioRad, Hercules, CA) as described by Kelley et al., 1995, or Novex pre-poured 4–20% gradient gels (Novex, San Diego, CA). 2.4. Acetate and protein assays Acetate in unfiltered conditioned medium was determined using an acetic acid assay kit (Boehringer Mannheim, Indianapolis, IN ). Protein in cell pellets lysed in 0.1% SDS was determined using the BioRad Bradford assay (BioRad, Hercules, CA). 2.5. b-Galactosidase assays b-Galactosidase assays were carried out using a modification (Rothstein et al., 1980) of the method described by Miller (1992). Duplicate samples containing 0.2 ml of cells were pelleted briefly in a microfuge, media was aspirated, and cell pellets were frozen at −80°C. For assay, pellets were thawed on ice. The assay was initiated by vortexing cell pellets or pellet dilutions in 1 ml of Z-buffer containing o-nitrophenyl b--galactopyranoside (4 mg/ml, ONPG), sodium deoxycholate

T.H. Grossman et al. / Gene 209 (1998) 95–103

(0.1 mg/ml ) and hexadecyltrimethylammonium bromide (0.2 mg/ml ). Assays were terminated with 0.5 ml of 1 M Na CO . Cell debris was pelleted in a microfuge, 2 3 and yellow color was quantitated at 420 nm. Units of b-galactosidase activity and the specific activity of bgalactosidase (mmol o-nitrophenol/min/mg total cell protein) were as defined by Miller (1992).

3. Results 3.1. Leaky recombinant expression is maximal at the onset of stationary phase in 4xYT medium Strain BL21(DE3) (pET-11a-CD58) was grown in the absence of inducer and analyzed for production of huCD58 protein at various times in the growth cycle. For cells grown in 4xYT medium, uninduced huCD58 protein first became detectable during late log phase and was present at high levels in overnight cultures (Fig. 1). In contrast, no leaky production of CD58

97

protein could be detected at any stage in the culture grown in LB medium. Although growth rates in both media were similar (Fig. 1A), final cell density in the richer 4xYT medium was typically 3–4-times higher than in LB. When a non-leaky LB culture was supplemented with yeast extract (6% final concentration) during late log phase (A =2.28, 5.75 h of growth, 600 Fig. 1A) and grown overnight, the supplemented culture produced CD58 at the same level as the culture grown in 4xYT ( Fig. 1B, lane 14), demonstrating the dependence of leaky expression on medium composition. Uninduced expression in cells grown in minimal defined medium (M9 or M63, Miller, 1992) was low and variable (not shown). 3.2. Leaky expression of the chromosomal lac operon also occurs at the onset of stationary phase in complex medium Leaky expression was analyzed further using both the chromosomal lac operon or otherwise isogenic Lac−

A

B

Fig. 1. Leaky recombinant expression increases at the onset of stationary phase and is dependent on medium composition. (A) BL21(DE3) (pET-11aCD58) cells were grown in LB, (#) or 4xYT ($). Samples were removed for SDS–PAGE analysis. (B) 4–20% gradient SDS–PAGE. Lanes 1–6, BL21(DE3) (pET-11a-CD58) cells grown in 4xYT ( lane 1, 4.5 h; lane 2, 5.25 h; lane 3, 5.75 h; lane 4, 6.75 h; lane 5, 7.75 h; lane 6, 24 h); lane 7, BL21(DE3) cells; lanes 8–13, BL21(DE3)(pET-11a-CD58) cells grown in LB ( lane 8, 4.5 h; lane 9, 5.25 h; lane 10, 5.75 h; lane 11, 6.75 h; lane 12, 7.75 h; lane 13, 24 h); lane 14, BL21(DE3) (pET-11a-CD58) cells grown in LB with the addition of 6% sterile yeast extract at A =2.28 (5.75 h 600 of growth). Arrow indicates huCD58 protein.

98

T.H. Grossman et al. / Gene 209 (1998) 95–103

host strain BL26(DE3) containing a plasmid encoding a full-length lacZ gene (pET-11a-lacZ). Leaky expression could thus be quantitated by measuring b-galactosidase activity, a more sensitive assay than Coomassie staining of proteins. Fig. 2A shows that the kinetics of leaky recombinant expression from pET-11a-lacZ were similar to those seen from pET-11a-CD58, indicating that leaky expression was a property of the expression system, rather than of a particular recombinant protein. In addition, Fig. 2C shows that the stationary phase increase in b-galactosidase expression was not due to a general increase in total cell protein, but was specific for b-galactosidase. b-Galactosidase expression from the chromosome of strain BL21(DE3) grown in 4xYT medium was also measured. Surprisingly, we found similar patterns of leaky expression at the onset of stationary phase (Fig. 2B). Leaky expression of both chromosomal and plasmid lacZ genes was reduced considerably by the introduction of plasmid pIQ2, which overproduces LacI protein (data not shown). To address the possibility that yeast extract may contain sufficient lactose to allow lac operon induction, we treated 4xYT medium with b-galactosidase to digest any lactose present in the medium, as described in Section 2.2. Treated medium was then autoclaved to destroy the added b-galactosidase, and used to grow cultures as before. b-galactosidase treatment of the medium did not alter levels of leaky expression from either the chromosome or the plasmid (not shown), indicating that lactose was not responsible for uninduced expression. 3.3. Regulation of leaky expression is independent of RpoS Since many stationary phase phenomena are known to be transcriptionally regulated by the stationary phase sigma factor RpoS ( Kolter et al., 1993; Loewen and Hengge-Aronis, 1994), we measured leaky expression in an rpoS mutant. We found that the rpoS mutation had no effect on leaky expression (data not shown). 3.4. Conditioned medium induces leaky expression in early log phase cells We next investigated the ability of conditioned medium obtained from cells at later stages of growth to stimulate leaky expression in early log phase cells. Medium samples from strain BL26(DE3) (pET-11alacZ), grown in 4xYT medium, were withdrawn and filter sterilized at various points in the growth cycle. Samples were numbered 1 through 7, according to the growth stage from which they were derived (Fig. 3A). Fresh cultures of strains BL26(DE3) (pET-11a-lacZ) and BL21(DE3) (pET-11a) were then grown to early

log phase (A =0.2). Note that strain 600 BL21(DE3) (pET-11a) was used to measure uninduced expression from the chromosomal lac operon because conditioned medium contained ampicillin. Aliquots of each culture were pelleted and resuspended in an equal volume of the various conditioned medium samples, grown for 1 h, and assayed for b-galactosidase activity. Cells doubled no more than twice in conditioned medium. Results show that leaky expression from both the chromosomal lac operon and the pET plasmid occurred in early log phase cells that were transferred to conditioned medium derived from later stages of growth (Fig. 3B and C ). The conditioned medium samples that yielded the highest levels of leaky lacZ expression were at a pH of approximately 6.0 and contained up to 20 mM acetate (not shown). 3.5. Leaky expression requires cyclic AMP To test the idea that depletion of a nutrient might trigger uninduced expression, we added back to late log phase cells various medium components that might have been depleted. We found that whereas addition of 2% yeast extract had no effect, addition of 1% glucose to late log phase cells virtually eliminated leaky expression (not shown). This result led us to test a possible role for cAMP regulation in leaky expression, using the cAMP-deficient mutant strain BL21cya::kan(DE3). We found that leaky expression levels were substantially reduced in cya cells, but could be restored by adding exogenous cAMP ( Table 1). Table 2 shows that pH 6, 30 mM acetate, and exogenous cAMP could each partially stimulate leaky expression, and that all three conditions stimulated leaky expression synergistically. 3.6. Comparison of cell viability, plasmid stability, and protein production stability using wt and cya expression strains To further assess the role of cAMP in leaky expression and on recombinant expression stability, we compared properties of strain BL21cya(DE3)(pET-11a-CD58) with those of the wt host. Fig. 4 shows that leaky expression of the CD58 adhesion domain also required a wt host strain ( lanes 1 and 2). The wt and cya hosts had equivalent plating efficiencies during log phase ( Table 3). Following overnight growth, however, the cya strain exhibited a much higher plating efficiency, consistent with leaky stationary phase CD58 production by the wt host, leading to eventual cell death (inability to form a colony). To analyze plasmid and expression stability, cells were grown uninduced overnight in Ap-containing 4xYT medium and then passaged three times in 4xYT without ampicillin (1:1000 dilution of each successive overnight culture). After each passage, colony-forming ability,

T.H. Grossman et al. / Gene 209 (1998) 95–103

99

A

B

C

Fig. 2. Expression of b-galactosidase in uninduced BL26(DE3) (pET-11a-lacZ ) and BL21(DE3) cells grown in 4xYT. (#), b-galactosidase Miller units; ($), A . (A) BL26(DE3) (pET-11a-lacZ ); (B) BL21(DE3); (C ) Specific activity (mmol o-nitrophenol/min/mg total cell protein) of b600 galactosidase in strain BL26(DE3) (pET-11a-lacZ ).

100

T.H. Grossman et al. / Gene 209 (1998) 95–103

A

B

C

Fig. 3. Conditioned medium induces leaky expression in early log phase cells. (A) Growth curve of BL26(DE3) (pET-11a-lacZ) cells in 4xYT medium from which conditioned media samples (1–7) were derived. (B) b-Galactosidase production in BL21(DE3)(pET-11a) cells; and (C ) BL26(DE3) (pET-11a-lacZ ) cells transferred to conditioned medium. Cells were grown to early log phase (A =0.2), transferred to conditioned 600 medium (1–7), and aerated at 37°C for 1 h. ‘No transfer’ indicates cells resuspended in their original growth medium. Results of b-galactosidase assays are the average of duplicate samples, with the standard deviation indicated.

plasmid retention, and leaky production of CD58 were assessed. Table 4 shows that the viability of the wt strain decreased markedly over time, many of the survivors having lost the plasmid or the ability to produce unin-

duced CD58. In striking contrast, the cya mutant retained the plasmid over time, even without Ap, and cell viability remained relatively constant. Following the third passage, both strains were diluted

Table 1 Exogenous cAMP restores leaky expression to strain BL26cya::kan (DE3) (pET-11a-lacZ ) Strain

Addition of 8 mM cAMP

b-Galactosidase (Miller units)a

BL26(DE3) (pET-11a-lacZ ) BL26(DE3) (pET-11a-lacZ ) BL26cya::kan (DE3)(pET-11a-lacZ) BL26cya::kan(DE3) (pET-11a-lacZ )

− + − +

2744±305 3319±323 392±35 5209±956

aCells grown overnight in 4xYT medium with or without cyclic AMP were assayed for b-galactosidase. Results are the average of duplicate samples, with standard deviations indicated.

101

T.H. Grossman et al. / Gene 209 (1998) 95–103 Table 2 Acetate, low pH, and cAMP act synergistically to promote leaky lac expression Transferred to media containinga:

b-Galactosidase

30 mM acetate

8 mM cAMP

pH

(Miller units)

− − + + − − + +

− − − − + + + +

6 7 6 7 6 7 6 7

98±0 44±2 455±22 100±2 160±8 70±8 4305±269 195±5 46±0 70±8 37 903±2792

+

+

6

870 ± 84 7±1 9±1 2899±15

BL26(DE3) (pET-11a-lacZ ) cells

Pre-transfer No transfer Induced BL21(DE3) cells Pre-transfer No transfer Induced

aEarly log cells (A =0.2) grown in 4xYT were pelleted and resuspended in an equal volume of 0.05 M Tris-buffered 4xYT (pH 6 or pH 7) 600 containing acetate or cAMP as indicated above. Following aeration for 1 h at 37°C, cells were assayed for b-galactosidase in duplicate.

Table 3 Comparison of CFUa in strains BL21(DE3) (pET-11a-CD58) and BL21cya(DE3) (pET-11a-CD58)b

Fig. 4. Recombinant CD58 protein production stability in cya and wt strains after passages in non-selective medium. Cells were grown and passaged as described in Section 3.6, and analyzed for CD58 production by SDS–PAGE as described in the legend to Fig. 1. Lane 1, wt in 4xYT+Ap, log phase, uninduced; lane 2, cya mutant in 4xYT+Ap, log phase, uninduced; lane 3, wt in 4xYT+Ap, uninduced, stationary phase; lane 4, cya mutant in 4xYT+Ap, uninduced, stationary phase; lane 5, wt in 4xYT, no Ap, uninduced, pass 1; lane 6, cya mutant in 4xYT, no Ap, uninduced, pass 1; lane 7, wt in 4xYT, no Ap, uninduced, pass 2; lane 8, cya mutant in 4xYT, no Ap, uninduced, pass 2; lane 9, wt in 4xYT, no Ap, uninduced, pass 3; lane 10, cya mutant in 4xYT, no Ap, uninduced, pass 3; lane 11, cya mutant in 4xYT+Ap, uninduced, pass 4; lane 12, cya mutant in 4xYT+Ap, induced, pass 4.

1000-fold into 4xYT medium containing Ap (passage 4), induced with IPTG and tested for CD58 production. Whereas the cya strain grew up readily in LBAp and produced CD58 following induction in log phase

Strain

CFUa-LB

Log phasec wt cya

3.6×108±1.6×108 1.5×108±0.6×108

Stationary phase (overnight) wt cya

1.4×108±0.4×108 1.2×1010±0.1×1010

aColony forming units, CFU, are given as titer per milliliter of culture per A unit (at 600 nm). CFU were determined on plates containing LB medium without ampicillin. Data are the average of two separate experiments. bCells were grown uninduced at 37°C in 4xYT medium containing Ap (50 mg/ml ). cA approximately 0.5. 600 Table 4 Comparison of CFUa and plasmid stability in strains BL21(DE3) (pET-11a-CD58) and BL21cya(DE3) (pET-11a-CD58) following passaging in non-selective mediumb Pass number

Strain

CFU-LB

CFU-LBAp

1

wt cya wt cya wt cya

4.8×108 8.2×108 4.8×108 9.4×108 1.7×106 9.6×108

8.0×107 8.0×108 1.0×107 1.3×109 5.0×105 7.5×108

2 3

% Plasmid retention

aCFU are in given as titer per milliliter of culture. bCells were grown and passaged as described in Section 3.6.

16.6 97.5 2.0 >100 29.4 78.1

102

T.H. Grossman et al. / Gene 209 (1998) 95–103

(Fig. 4), the wt strain was unable to grow in the presence of Ap.

4. Discussion Our efforts to optimize recombinant gene expression in complex medium led to the discovery that genes regulated by lac control elements were expressed at high levels in uninduced cells approaching the stationary phase. Leaky expression was readily detectable using the highly amplified pET system, but occurred from the chromosomal lac operon as well. Because levels of leaky expression were strongly influenced by medium composition and cell growth stage, we surmised that leaky expression was at least partially regulated by nutrient availability. Guided by the composition of conditioned 4xYT medium from early stationary phase cells, we found that log phase cells supplemented with acetate and cAMP at pH 6.0 exhibited high levels of leaky expression of the recombinant lacZ gene. Under the same conditions, log phase cells were also derepressed for the chromosomal lac operon in the absence of exogenous inducer. The requirement for cAMP in promoting high levels of leaky expression from both the wt lac promoter and lacUV5 promoter (theoretically not subject to cAMP regulation) is puzzling. Earlier in vitro studies by others showed that stimulation of the lac and lacUV5 promoters occurred in the presence of cAMP and CRP ( Eron and Block, 1971). Although the stimulatory effect on the lacUV5 promoter was small (2 to 3-fold ), any small increase could be highly amplified in the T7 recombinant expression system, in which a gene on a multicopy plasmid is transcribed by the highly processive T7 RNA polymerase. However, the requirement for cAMP in leaky expression may not be directly at the lac promoter. Expression of a number of survival-related gene systems are now known to be accompanied by elevated cyclic AMP levels during stationary phase. These include colicin production ( Eraso et al., 1996), SOS-induction ( Taddei et al., 1995), and oxidative stress responses (Lange and Hengge-Aronis, 1991; for review see Botsford and Harman, 1992, and Saier et al., 1996). Control of such regulons could be mediated through other effector molecules that may be regulated by cAMP. The finding that leaky lac expression is independent of RpoS function is consistent with the observation that as cells become nutrient limited, high affinity sugar transport systems are turned on before some RpoSdependent functions (Notley and Ferenci, 1996). Certain carbohydrate scavenging genes, such as mal, gal, mgl, and lamB are known to be induced during conditions of glucose limitation in the absence of an external inducer (Death and Ferenci, 1994; Death et al., 1993; Notley and Ferenci, 1995; Notley and Ferenci, 1996).

Such induction has been shown to be dependent on the synthesis of endogenous inducers and elevated cyclic AMP levels. Whether an endogenous inducer is required for lac derepression under the conditions described here is not yet known. Our results led us to test the suitability of a cya mutant as an expression host. The mutant showed greatly enhanced expression and plasmid stability as compared with the wt host presumably because it is not counterselected in early stationary phase. Although early stationary phase cells are not normally used for protein production, stationary phase is usually reached on agar plates during the initial selection of transformant colonies, and by some fraction of the cell population during growth, induction, or seed stock formation. For these reasons, the presence of the cya mutation should be particularly desirable for production of toxic proteins, or in other situations that confer plasmid or expression instability. In summary, we found that uninduced expression of lac-controlled genes occurs as cells approach stationary phase in complex medium. Cyclic AMP, acetate, and low pH are required to effect high-level expression in the absence of IPTG. These are novel findings with regard to regulation of lac expression, and suggest that the lac operon may be part of the class of high-affinity carbohydrate transport genes that are turned on in the absence of inducer under carbon-limiting conditions.

Acknowledgement We thank K. Kelley for his early contributions to these studies. We also thank D. Rothstein for help with the manuscript and D. Rothstein and B. Magasanik for helpful discussions.

References Bohannon, D.E., Connell, N., Keener, J., Tormo, A., Espinosa-Urgel, M., Sambrano, M.M., Kolter, R., 1991. Stationary-phase-inducible gearbox promoters: differential effects of katF mutations and role of s70. J. Bacteriol. 173, 4482–4492. Botsford, J.L., Harman, J.G., 1992. Cyclic AMP in prokaryotes. Microb. Rev. 56, 100–122. Death, A., Ferenci, T., 1994. Between feast and famine: endogenous inducer synthesis in the adaptation of Escherichia coli to growth with limiting carbohydrates. J. Bacteriol. 176, 5101–5107. Death, A., Notley, L., Ferenci, T., 1993. Derepression of LamB protein facilitates outer membrane permeation of carbohydrates into Escherichia coli under conditions of nutrient stress. J. Bacteriol. 175, 1475–1483. Dubendorff, J.W., Studier, F.W., 1991. Controlling basal expression in an inducible T7 expression system by blocking the target T7 promoter with lac repressor. J. Mol. Biol. 219, 45–59. Eraso, J.M., Chidambaram, M., Weinstock, G.M., 1996. Increased production of colicin E1 in stationary phase. J. Bacteriol. 178, 1928–1935.

T.H. Grossman et al. / Gene 209 (1998) 95–103 Eron, L., Block, R., 1971. Mechanism of initiation and repression of in vitro transcription of the lac operon of Escherichia coli. Proc. Natl. Acad. Sci. USA 68, 1828–1832. Grossman, T.H., Silverman, P.M., 1989. Structure and function of conjugative pili: inducible synthesis of functional F pili by Escherichia coli K-12 containing a lac–tra operon fusion. J. Bacteriol. 171, 650–656. Kelley, K.C., Huestis, K.J., Austen, D.A., Sanderson, C.T., Donoghue, M.A., Stickel, S.K., Kawasaki, E.S., Osburne, M.S., 1995. Regulation of sCD4-183 gene expression from phage-T7-based vectors in Escherichia coli. Gene 15, 33–36. Kolter, R., Siegele, D.A., Tormo, A., 1993. The stationary phase of the bacterial life cycle. Annu. Rev. Microbiol. 47, 855–874. Lange, R., Hengge-Aronis, R., 1991. Identification of a central regulator of stationary-phase gene expression in Escherichia coli. Mol. Microbiol. 5, 49–59. Loewen, P.C., Hengge-Aronis, R., 1994. The role of the sigma factor ss ( KatF ) in bacterial global regulation. Annu. Rev. Microbiol. 48, 53–80. Makrides, S.C., 1996. Strategies for achieving high level expression of genes in Escherichia coli. Microbiol. Rev. 60, 512–538. Miller, J.H., 1992. A Short Course in Bacterial Genetics: A Laboratory Manual and Handbook for Escherichia coli and related bacteria. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, 1992. Moffatt, B.A., Studier, F.W., 1987. T7 lysozyme inhibits transcription by T7 RNA polymerase. Cell 49, 221–227. Notley, L., Ferenci, T., 1995. Differential expression of mal genes under cAMP and endogenous inducer control in nutrient-stressed Escherichia coli. Mol. Microbiol. 16, 121–129.

103

Notley, L., Ferenci, T., 1996. Induction of RpoS-dependent functions in glucose-limited continuous culture: what level of nutrient limitation induces the stationary phase of Escherichia coli? J. Bacteriol. 178, 1465–1468. Novagen, Inc., 1995. Hosts for Expression. pET System Manual, 5th edn, pp. 5-6. Rothstein, D.M., Pahel, G., Tyler, B., Magasanik, B., 1980. Regulation of expression from the glnA promoter of Escherichia coli in the absence of glutamine synthetase. Proc. Natl. Acad. Sci. USA 77, 7372–7376. Saier, Jr., M.H., Ramseier, T.M., Reizer, J., 1996. Regulation of carbon utilization. In: Neidhardt, F.C. ( Ed.), Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, 2nd edn. American Society for Microbiology, Washington, DC, pp. 1325–1343. Shah, S., Peterkofsky, A., 1991. Characterization and generation of Escherichia coli adenylate cyclase deletion mutants. J. Bacteriol. 173, 3238–3242. Studier, F.W., Moffatt, B.A., 1986. Use of bacteriophage T7 RNA polymerase to direct selective high-level expression of cloned genes. J. Mol. Biol. 189, 113–130. Taddei, F., Matic, I., Radman, M., 1995. cAMP-dependent SOS induction and mutagenesis in resting bacterial populations. Proc. Natl. Acad. Sci. USA 92, 11736–11740. Wallner, B.P., Frey, A.Z., Tizard, R., Mattaliano, R.J., Hession, C., Sanders, M.E., Dustin, M.L., Springer, T.A., 1987. Primary structure of lymphocyte function-associated antigen 3 (LFA-3), the ligand of the T lymphocyte CD2 glycoprotein. J. Exp. Med. 166, 923–932.