Long-term and homogeneous regulation of the Escherichia coli araBAD promoter by use of a lactose transporter of relaxed specificity

Abstract

Expression systems based on the Escherichia coli arabinose operon P_BAD promoter exhibit the all-or-nothing (autocatalytic) induction of expression that was first documented in the lac operon. Under conditions of subsaturating levels of inducer, some of the cells of the population are fully induced, whereas other cells remain uninduced. Recently, a new AraE transporter system was reported to have circumvented the problem of autocatalytic expression in the pBAD expression vectors and to provide graded and homogeneous cell-to-cell expression in the presence of variable inducer concentrations [Khlebnikov, A., Risa, O., Skaug, T., Carrier, T. A. & Keasling, J. D. (2000) J. Bacteriol. 182, 7029–7034]. However, we report that nonuniform gene expression in the AraE system was readily detectable by the use of mutant green fluorescent proteins that are rapidly degraded in E. coli. We report an approach to avoid all-or-nothing induction of the pBAD promoter; the use of a mutant LacY transporter in a strain deficient in both arabinose transport (araE araFGH) and degradation (araBAD). This mutant LacY protein performs facilitated diffusion of arabinose resulting in homogeneous expression of an unstable GFP that is maintained over extended incubation times at subsaturating levels of inducer. This approach is readily adapted to other sugar-regulated expression systems.

Expression of a gene of interest under control of an inducible promoter is a powerful molecular genetic tool. Such expression systems generally utilize strongly inducible promoters that facilitate high levels of protein expression, but which often become problematic in studies requiring either complete repression or partial induction of the gene of interest. An ideal promoter system should be fully repressible in the absence of the inducer but also should allow for levels of gene expression that vary directly and linearly with the extracellular concentration of the inducing ligand. However, most bacterial expression systems suffer from the “all-or-nothing” induction of gene expression first demonstrated in the lac operon by Novick and Weiner more than forty years ago (1). At inducer concentrations below saturation levels, a heterogeneous population results in which some cells are fully induced, whereas other cells undergo little or no induction (1). In the lac system, all-or-nothing induction results from random bursts of transporter induction caused by spontaneous dissociation of the LacI repressor from the lacZYA operator. In the ara system, bursts of arabinose transporter synthesis probably result from a stochastic change in conformation of AraC proteins bound to the promoters. In both cases, those cells that contain sufficient transporters to accumulate inducer at the time the sugar is added induce the synthesis of additional transporters that catalyze uptake of more sugar leading to further induction. This autocatalytic induction cycle continues until this subpopulation of cells becomes fully induced. In contrast, cells having transporter levels that fall below the threshold number of transporters required for inducer accumulation never capture sufficient sugar to become induced. Because the fraction of fully induced cells increases linearly with inducer concentration, a smooth increase of gene expression with increasing inducer concentration is seen at the population level. Such results have often led to the assumption that there is little cell-to-cell variation in induction level. However, this interpretation is incorrect; expression levels vary greatly within the cell population, making physiological interpretations problematic.

The araBAD regulon of Escherichia coli has been shown to exhibit all-or-nothing induction (2). By analogy with the lac system, it was suggested that the all-or-nothing response of the pBAD expression vectors is caused by the accumulation of arabinose via the arabinose-inducible active transport proteins (2). Various expression vectors that use the P_BAD promoter and araC have been developed, one of which, the pBAD vectors, has been widely used in E. coli and in other organisms (3–5). However, the autocatalytic induction of the operon has precluded their intended use in fine modulation of gene expression (2). Recently, an approach was reported to circumvent the all-or-nothing induction character of the pBAD expression vectors by uncoupling the induction of arabinose transport from that of the P_BAD promoter (6). In this system, a plasmid-borne AraE transporter was expressed from a heterologous promoter in a strain lacking functional chromosomal arabinose transport genes. Green fluorescence protein (GFP) then was expressed under pBAD control in a compatible vector. This system showed a homogenous distribution of expression of GFP, whereas the wild-type strain exhibited the expected all-or-nothing response (6). Despite the reported success of the AraE system, we envisioned that because the original arabinose transport-deficient strains retained functional araBAD genes, the inducer would be depleted by metabolism, which would lead to a progressive loss of induction at low arabinose concentrations. However, the resulting heterogeneity among the cells of the population would be unobservable because of the stability of the GFP reporter. We report a different approach to coping with all-or-nothing induction of the arabinose regulatory system. A mutant LacY protein (LacY A177C) that performs facilitated diffusion of arabinose (7) is expressed in a strain deficient in both arabinose transport and arabinose degradation. By the use of unstable mutants of GFP (8), we show that reporter gene expression is homogeneous during extended growth periods in the presence of low arabinose concentrations.

Materials and Methods

Media and Reagents.

The rich medium used in this study was rich broth (RB) and contained 10 g/liter tryptone, 5 g/liter NaCl, and 1 g/liter yeast extract. In addition, a modified low salt RB (LSRB) medium (4 g/liter NaCl) was used for the growth of strains harboring plasmids that encoded GFP, because E. coli strains that produce GFP are osmotically sensitive (8). MacConkey Agar Base (Difco) and minimal medium M9 supplemented with 0.4% l-arabinose were used to test sugar utilization. To select for plasmid maintenance, antibiotics (Sigma) were used at the following concentrations: ampicillin, 100 μg/ml; tetracycline, 20 μg/ml; kanamycin, 50 μg/ml; and chloramphenicol, 34 μg/ml. Timenten (25 μg/ml) was used in the place of ampicillin in liquid cultures. Isopropyl β-d-thiogalactoside was used at a concentration of 100 μM.

Bacterial Strains and Plasmids.

The E. coli strains used and their relevant markers were CW2553, araE201 ΔaraFGH∷kan (and its recA56 derivative CW2587; ref. 9); CAG12093, car∷Tn10; BW25113, lacI^q rrnBT14 ΔLacZ_WJ16 hsdR514 ΔaraBAD_AH33ΔrhaBAD_LD78 (10) and MC1061, araD139Δ(ara-leu)7696 (11). P1 phage transduction and other bacterial genetic techniques were performed by standard methods (12). Strains RMK1, RMK 11, and RMK12 were constructed via P1 transduction of BW25113, MC1061, and CW2553 by using P1 lysates grown on MG1655, CAG12093, and RMK11, respectively. The P_tac∷araE plasmid pAK01 was the gift of J. Keasling (Univ. of California, Berkeley). The lacY plasmids were the gift of H. R. Kaback (Univ. of California, Los Angeles), and the set of pJBA plasmids encoding GFPs of varying stabilities was the gift of B. Andersson (Technical Univ. of Denmark, Lyngby). The pBAD vectors were from J. Beckwith (Harvard Medical School, Boston). The pBAD plasmids encoding the various GFPs were constructed by ligating the 0.72-kb XbaI-HindIII fragment of a pJBA plasmid to either pBAD33 or pBAD24 digested with NheI and HindIII followed by screening transformants with a long wavelength UV lamp (BioRad) after growth on LSRB-arabinose plates.

Fluorescence Induction Experiments.

Fluorescence measurements were performed on a Beckman Coulter EPICS XL flow cytometer equipped with an argon laser (excitation wavelength of 488 nm) and a band pass filter of 525 nm. Samples were diluted with the growth medium to obtain an OD₆₀₀ ≃ 0.05 before fluorescence measurements. For each sample, 30,000 events were collected. For fluorescence induction experiments, overnight cultures (grown in LSRB supplemented with 0.02% glucose and appropriate antibiotics) were diluted 100-fold in the same medium lacking glucose and were allowed to grow at 37°C until early-log phase (OD₆₀₀ ≃ 0.15). In long-term experiments, the cultures were periodically diluted with fresh medium to prevent entry into stationary phase. Expression of GFP was induced by the addition of varying concentrations of l-arabinose. To investigate GFP expression during colony formation, overnight liquid cultures were diluted 10⁶-fold and plated on LSRB plates containing the appropriate antibiotics and varying concentrations of l-arabinose. After 16 h of growth at 37°C, well-isolated single colonies were scraped from the plate, resuspended in 5 ml of LSRB, and examined by flow cytometry.

Results and Discussion

GFP Expression in Wild-Type and ΔaraBAD Strains.

During construction of the pBAD-GFP plasmids, transformation of strains having wild-type arabinose genes with ligation products failed to yield fluorescent colonies on LSRB plates supplemented with 0.2% l-arabinose; however, parallel transformations in a strain blocked in arabinose metabolism (araBAD) produced many fluorescent colonies (data not shown). These results indicated that GFP expression from the P_BAD promoter was appreciably greater in the ΔaraBAD strain vs. the wild-type strain even at saturating levels of arabinose. Thus, we compared the effects on GFP expression in isogenic strains either in a wild-type strain or in a ΔaraBAD strain under saturating levels of inducer (0.2% l-arabinose). GFP expression was assayed from pBAD plasmids expressing either the wild-type GFP or one of three altered GFPs that differ in their rates of intracellular degradation because of different extensions of the C terminus (8). We have adopted the prior designations (8) of the GFP(LAA) as the “fast degrader”, the GFP(AAV) as the “medium degrader”, and GFP(ASV) as the “slow degrader”. Expression of GFP was significantly lower in the wild-type vs. the ΔaraBAD strain, particularly in cells that harbored one of the degrader GFPs (Fig. 1 A and B). Furthermore, GFP expression in wild-type cultures that expressed one of the three unstable GFPs showed a transient increase in fluorescence followed by a rapid decline (Fig. 1C). This decrease was most pronounced when the fast and medium degrader GFPs were used in the wild-type strain, where fluorescence intensity began to decline after only 90 min of induction (Fig. 1C). In contrast, GFP expression in the ΔaraBAD strain expressing either the fast or medium degraders exhibited increased fluorescence values for up to 3 h, whereas the wild-type and slow degrader GFPs exhibited a steady rise in fluorescence yields for the duration of the experiment (6 h; Fig. 1 B and D). The differential loss of fluorescence at the level of the least stable GFPs was clearly because of arabinose metabolism by the wild-type strain, resulting in inducer concentrations that were sufficiently low that the rates of GFP degradation exceeded GFP synthesis. Furthermore, these data demonstrate that loss of induction cannot be assayed with the wild-type reporter but is readily detected by using the unstable GFPs. We also tested the effects of external arabinose concentration in wild-type and ΔaraBAD strains expressing either the wild-type or fast degrading GFPs (Fig. 2). Both strains required an arabinose concentration of 0.2% for maximal fluorescence yields; however, the ΔaraBAD strain consistently gave greater fluorescence yields than the wild-type strain at all inducer concentrations tested (Fig. 2A). Furthermore, an arabinose concentration of 1.5 μM (0.0002%) was sufficient to induce expression in 100% of the cell population of the ΔaraBAD strain. In contrast, the wild-type strain required 15 mM l-arabinose for full induction (Fig. 2B). The weak pBAD induction observed in the wild-type strain RMK1 was also noted in another widely used wild-type strain, MG1655 (data not shown). Lastly, we measured rates (t_1/2 values) of GFP degradation in strains carrying the pBAD33-derived GFP plasmids. Degradation rates were estimated according to Andersen et al. (8) in log phase cultures that had been induced with 0.2% l-arabinose for 90 min before the degradation assay. The half times were estimated as 56, 34, and 23 min for the slow, medium, and fast degraders, respectively, whereas the wild-type GFP showed no detectable degradation (data not shown). Our rates are somewhat faster than those reported (8). This discrepancy is probably because of different levels of GFP expression, as high levels of GFP expression can saturate the proteolytic degradation systems (8). In support of this hypothesis, we found that the turnover rates of both the medium and fast degrader GFPs were about 15% slower when expressed from the higher copy number vector, pBAD24.

Figure 1.

Induction of fluorescence in the wild-type strain RMK1 (A and C) vs. the ΔaraBAD strain, BW25113 (B and D), after the addition of 0.2% l-arabinose to cultures in early-log phase. ■, pRK6, expressing wild-type GFP; ●, pRK7, expressing the slow degrader GFP(ASV); ▴ pRK8, expressing the fast degrader GFP(LAA); ▾, pRK9, expressing the medium degrader GFP(AAV). (A and B) Values represent the mean ± SE of three independent experiments. (C and D) Values represent the means normalized to maximal fluorescence yields.

Figure 2.

Effect of arabinose concentration on expression in wild-type (black symbols) and ΔaraBAD (white symbols) strains. Early-log cultures were induced with varying concentrations of l-arabinose, and fluorescence was measured after 3 h of induction. ■, □, pRK6, expressing wild-type GFP; ●, ○, pRK8, expressing the fast degrader GFP(LAA). Values represent the mean ± SE of three independent experiments.

Test of the AraE System.

To test our hypothesis that the AraE system would fail to ensure population homogeneity in cultures induced with low concentrations of arabinose and/or at longer incubation times, either the wild-type or the fast degrader GFPs were expressed in this system under experimental conditions that were comparable to those reported (6). GFP expression was induced in early-log phase cultures with 0.02% l-arabinose, and the homogeneity of GFP expression within the population was monitored as fluorescence emission by flow cytometry (Fig. 3A). Six hours after induction, the population of cells expressing the wild-type GFP appeared homogeneous (Fig. 3A Left), as reported by Khlebnikov et al. (6). In contrast, cultures expressing the fast degrader GFP exhibited a less uniform distribution after only 3 h of induction and showed very heterogeneous expression, with many subpopulations being apparent 6 h after induction (Fig. 3A Right). Thus, the apparent homogeneity of GFP expression in cultures expressing the wild-type GFP is caused by the persistence of the long-lived wild-type GFP synthesized early after induction, coupled with the low rate of GFP synthesis late in the induction period because of arabinose depletion. Thus, the tight distribution of GFP content observed 6 h after induction with 0.02% arabinose by Khlebnikov et al. (6) is, to a large extent, an artifact of the long-lived nature of the GFP reporter used.

Figure 3.

Expression of GFP in an arabinose transporter-deficient strain (CW2587) harboring plasmids encoding either the AraE or LacY transporters. Fluorescence induction was measured in early-log phase cultures after the addition of 0.02% arabinose. (A) The distribution of GFP expression as a function of induction time in the AraE system harboring a pBAD reporter plasmid that expressed either the wild-type GFP (Left) or the fast degrader GFP(LAA) (Right). Values to the left denote induction time in hours. (B) Expression of the wild-type GFP under pBAD control in CW2587 harboring plasmids encoding wild-type LacY (■), LacY(A177C) (●), or LacY(A177V) (▴). Values represent the mean ± SE (n = 3). (C) The distribution of GFP expression as a function of induction time in the LacY A177C transporter system harboring a pBAD reporter plasmid that expressed either wild-type GFP (Left) or the fast degrader GFP(LAA) (Right).

Use of a Mutant Lactose Permease as an Arabinose Transporter.

Goswitz and Brooker (7) reported that a mutant A177V LacY permease protein functioned efficiently as an arabinose transporter. Additional work has shown that LacY A177V allows downhill transport (13) of maltose, arabinose, palatinose, sucrose, and cellobiose, but active transport of these sugars was not observed (14). Therefore, we transformed E. coli strain CW2587, which lacks both arabinose transport systems (9), with plasmids expressing either the wild-type LacY protein or either of two LacY mutant proteins, A177V or A177C. Expression of either mutant LacY protein allowed growth of these strains on arabinose as sole carbon source and gave red colonies on arabinose-MacConkey indicator plates, whereas expression of the wild-type LacY failed to allow metabolism of arabinose (data not shown; ref. 7). A plasmid that encoded the wild-type GFP construct then was introduced into the LacY plasmid-containing strains. Cultures of strains expressing either LacY A177V or A177C transporters exhibited significantly higher fluorescence yields than cultures expressing the wild-type lactose permease (Fig. 3B). Furthermore, 100% of the cells in cultures expressing either of the mutant LacY proteins were induced in less than 2 h and continued to be reliably induced for the duration of the experiment, whereas less than 10% of the cells expressing the wild-type LacY were induced (Fig. 3B). Expression of LacY A177C produced a 3-fold higher fluorescence yield than expression of the A177V mutant (Fig. 3B) and, thus, was used in subsequent experiments. Relaxed sugar specificity of this protein had not previously been reported but would be expected because substitutions of residue 177 with residues having hydrophobic side chains larger than that of Ala are known to relax the sugar specificity of LacY (15).

The LacY A177C system then was compared with the AraE system in cultures induced with 0.02% l-arabinose. In cultures harboring the wild-type GFP, expression levels were comparably homogenous after 6 h of incubation (Fig. 3 A and C, Left). However, in the presence of the fast degrader GFP, the LacY transporter system exhibited a more uniform distribution than the AraE system, particularly after 6 h induction (Fig. 3 A and C, Right). Thus, although the AraE transporter system has an advantage over the prior pBAD-expression systems, the LacY A177C transporter provides a significant improvement at extended incubation times. Lastly, it should be noted that, in cultures harboring either the AraE or the LacY A177C transporter systems and the fast degrader GFP exhibited very low induction levels at lower arabinose concentrations (data not shown), which precluded the use of this mutant GFP under these conditions.

Use of the LacY A177C Transporter in a ΔaraBAD Strain.

It seemed likely that a major cause of the observed heterogeneity in GFP expression in both transporter systems was arabinose consumption by the enzymes encoded by the araBAD operon. To eliminate this complication, we constructed strain RMK12, a ΔaraBAD derivative of the araE araFGH strain, and monitored GFP expression in cultures that harbored plasmids encoding either the AraE or the LacY A177C transporters plus a GFP reporter plasmid. First, we performed a very stringent test by assay of the distribution of GFP expression in cells of colonies formed on solid medium (thereby mimicking conditions used in mutant selections). Overnight cultures were diluted 10⁶-fold, plated, and allowed to grow for 16 h on LSRB plates supplemented with varying concentrations of l-arabinose. Well-isolated colonies then were scraped from the plate, and fluorescence induction was analyzed by means of flow cytometry. As expected, with the wild-type GFP, the long half-life of the reporter gave relatively homogenous distributions regardless of arabinose concentration and arabinose degradation (data not shown) for both transporter systems. Therefore, we chose the medium degrader GFP as a more accurate assessment of the uniformity of gene expression. As mentioned above, cultures of the strain with wild-type arabinose catabolism (CW2553) expressing either transport system had very low levels of GFP expression even at saturating l-arabinose concentrations, indicating that the rate of GFP degradation can exceed the rate of GFP synthesis because of the consumption of arabinose during extended time periods (Fig. 4 A and B, Left). Surprisingly, cultures of the ΔaraBAD strain harboring the AraE transporter plasmid had a highly nonuniform distribution of GFP expression at arabinose concentrations lower than 0.2% (Fig. 4A Right), and thus blocking arabinose degradation did not remedy the defects of this system. In marked contrast, cultures of the LacY A177C-producing strain deficient in both arabinose metabolism and transport maintained a uniform distribution of GFP expression over a 100-fold range of arabinose concentrations (Fig. 4B Right). Thus, these data demonstrate that LacY A177C-mediated transport in a ΔaraBAD araE araFGH strain provides a very tight distribution of gene expression over extended time periods even at low arabinose concentrations. This finding is especially noteworthy given the very different environments of cells in the center of the colony vs. those at the colony periphery. Lastly, we characterized the uniformity of expression of the medium degrader GFP in the ΔaraBAD araE araFGH strain expressing the LacY A177C transporter in liquid medium and over the shorter time periods commonly used in most gene expression experiments. Cultures in early-log phase were induced with varying levels of arabinose and GFP expression was monitored for up to 6 h after induction (Fig. 5). As expected, the cultures had a very tight distribution of graded gene expression over a 100-fold range of arabinose concentrations.

Figure 4.

Uniformity of GFP expression in overnight cultures of strain CW2553 (araE araFGH) vs. RMK12 (ΔaraBAD araE araFGH). Cultures harbored either the AraE (A) or the LacY A177C (B) transporter plasmids and a reporter plasmid expressing the medium degrader GFP(LVA) under pBAD control. Fluorescence was assayed from cultures grown for 16 h on LSRB plates supplemented with various concentrations of l-arabinose. Values to the left denote arabinose concentrations in percent. (Left) Results obtained in strain CW2553 (araE araFGH). (Right) Results obtained in strain RMK12 (ΔaraBAD araE araFGH).

Figure 5.

Effect of arabinose concentration on cell-to-cell expression of GFP in strain RMK12 (araBAD araE araFGH) harboring pLacYA177C and the medium degrader GFP(LVA). Cultures in the early-log phase of growth were induced with variable arabinose concentrations, and fluorescence yield was monitored over 6 h. Arabinose concentration in percent is denoted on the left and induction time in hours is denoted above each panel.

Conclusions

Expression of mutant LacY transporters that perform facilitated diffusion of diverse sugars provides an almost ideal solution to the problem of all-or-nothing induction of the arabinose operon, whereas the AraE system suffers from cell-to-cell heterogeneity even when arabinose degradation is blocked. Given the wide variety of monosaccharides and disaccharides transported via the mutant LacY transporters, our approach should be easily adapted to other sugar-regulated promoters. In support of this, preliminary results in our laboratory indicate that this transporter system is applicable to the very tightly regulated P_rhaB (16) promoter of the E. coli rhamnose operon.

Abbreviations

GFP

green fluorescent protein

RB

rich broth

LSRB

low salt rich broth