Transposon-mediated activation of the Escherichia coli glpFK operon is inhibited by specific DNA-binding proteins: implications for stress-induced transposition events

Abstract

Escherichia coli cells deleted for the cyclic AMP (cAMP) receptor protein (Crp) gene (Δcrp) cannot utilize glycerol because cAMP-Crp is a required activator of the glycerol utilization operon, glpFK. We have previously shown that a transposon, Insertion Sequence 5 (IS5), can insert into the upstream regulatory region of the operon to activate the glpFK promoter and enable glycerol utilization. GlpR, which represses glpFK transcription, binds to the glpFK upstream region near the site of IS5 insertion and inhibits insertion. By adding cAMP to the culture medium in ΔcyaA cells, we here show that the cAMP-Crp complex, which also binds to the glpFK upstream regulatory region, inhibits IS5 hopping into the activating site. Control experiments showed that the frequencies of mutations in response to cAMP were independent of parental cell growth rate and the selection procedure. These findings led to the prediction that glpFK-activating IS5 insertions can also occur in wild-type (Crp⁺) cells under conditions that limit cAMP production. Accordingly, we found that IS5 insertion into the activating site in wild-type cells is elevated in the presence of glycerol and a non-metabolizable sugar analogue that lowers cytoplasmic cAMP concentrations. The resultant IS5 insertion mutants arising in this minimal medium become dominant constituents of the population after prolonged periods of growth. The results show that DNA binding transcription factors can reversibly mask a favored transposon target site, rendering a hot spot for insertion less favored. Such mechanisms could have evolved by natural selection to overcome environmental adversity.

Introduction

Wild type E. coli cells can grow on glycerol as a sole carbon source, but cells lacking the cAMP receptor protein (Crp) cannot (Lin, 1976, Won et al., 2009, Fic et al., 2009). In a previous communication (Zhang and Saier, 2009a), we showed that a Δcrp strain could mutate to rapid glycerol utilization due to insertion of the small transposon, Insertion Sequence 5 (IS5) (Sousa et al., 2013). To cause activation, IS5 hops into a single site, in a single orientation, upstream of the glpFK operon promoter. The presence of IS5 at this site activates the glpFK promoter so that it becomes stronger than that in wild type cells (Zhang and Saier, 2009b). The glpFK-activating insertional event occurred at high frequency in the presence of glycerol, but not in the presence of glucose or another carbon source. Glycerol increased insertion of IS5 at this specific site but not in other operons (Zhang and Saier, 2009a, Zhang and Saier, 2011). Glycerol-promoted IS5 insertion into the glpFK-activating site proved to be regulated by binding of the glycerol repressor, GlpR, to its four adjacent glpFK operators, O1, O2, O3 and O4 in the glpFK control region. However, it became clear that the effect of GlpR-binding on IS5 insertion was not mediated by increased expression of glpFK, or by increased growth, since binding to O1 primarily controlled IS5 insertion without a significant impact on transcription, while binding to O4 primarily controlled transcription (Zhang and Saier, 2009a). Moreover, insertion could be shown to occur independently of the selection procedure (Zhang and Saier, 2009a). Thus, the inhibition of IS5 insertion into the upstream activating site is a newly recognized function of GlpR that is distinct from the previously recognized function of repressing glpFK transcription (Zhang and Saier, 2011). Finally, we demonstrated that IS5 can precisely excise, showing that its insertion can be considered to be fully reversible (Zhang et al. 2010).

In this communication, we first report that in ΔcyaA Crp⁺ cells, which lack the cyclic AMP biosynthetic enzyme, adenylate cyclase, Cya (Gancedo, 2013), IS5-mediated glpFK activation occurs in a manner strictly analogous to that observed in Δcrp cells. We further show that addition of cAMP to the growth medium, known to increase the cytoplasmic cAMP concentration (Saier et al., 1982), greatly suppresses IS5 insertion specifically at this site. This effect occurred independently of GlpR, but it depended exclusively on Crp and the two adjacent Crp binding sites (CrpI and CrpII) that overlap the two GlpR binding sites, O2 and O3, in the glpFK control region (Zhang and Saier, 2009a, Weissenborn et al., 1992). It thus became clear that the conditions that predispose the glpFK operon to activation by IS5 in wild type cells were (i) the presence of glycerol, and (ii) the presence of an environmental agent that lowers cytoplasmic cAMP levels.

Non-metabolizable glucose analogues and other sugar substrates of the phosphoenolpyruvate (PEP):sugar phosphotransferase system (PTS) are among the compounds known to lower cellular cAMP concentrations by inhibiting adenylate cyclase (Feucht and Saier, 1980). These sugar analogues include 2-deoxy-D-glucose (2DG) and methyl-α-D-glucoside (αMG) (Saier et al., 1982). Here we show that incubation of wild type E. coli cells in glycerol media together with 2DG or αMG promotes glpFK-activating IS5 insertional events. Our results are consistent with a scenario in which environment-sensitive transcription factors such as GlpR and Crp reversibly mask transposition target sites so as to suppress or promote IS5 insertional activation of genes, depending on conditions. We discuss these results in the context of the current understanding of mutagenic mechanisms that are proposed to be active in the absence of appreciable growth.

Materials and Methods

Bacterial strains and growth conditions

Strains and DNA oligonucleotides used in this study are described in Supplementary Tables S1 and S2, respectively. The cyaA deletion mutant was generated from the parental strain (E. coli K-12 strain BW25113) using the method of (Datsenko and Wanner, 2000). Briefly, a kanamycin resistance gene (km), flanked by the FLP recognition site (FRT) was amplified from the template plasmid pKD4 using mutation oligos cyaA1-P1 and cyaA2-P2 (Supplementary Table S2), each of which is composed of a ~20 bp region at the 3′ end that is complementary to the FRT-flanking km sequence, and a ~50 bp region at the 5′ end that is homologous to cyaA. The PCR products were gel purified, treated with DpnI, and then electroporated into BW25113 cells expressing the lamada-Red proteins encoded by plasmid pKD46. The pKD46 plasmid, which carries a temperature-sensitive origin of replication, was removed by growing the mutant cells overnight at 40 °C. The Km^r mutants were verified for the replacement of the target gene by the FRT-flanking km gene by PCR. The km gene was subsequently eliminated (leaving an 85-bp FRT sequence) using plasmid pCP20 that bears the FLP recombinase. The cyaA glpR double mutant was constructed by transferring a km insertional mutation of the cyaA gene into the glpR deletion mutant background (Zhang and Saier, 2009a) using P1 transduction.

To fuse the chloramphenicol-resistance gene (cat) with the glpFK operon, downstream of glpK in the chromosome, the plasmid pKD13-cat made previously (Zhang and Saier, 2009a), was used. In this plasmid, the cat gene is located upstream of a FRT-flanking km gene (Datsenko and Wanner, 2000). The cat structural gene with its own ribosome binding site (RBS), together with the downstream km gene, was amplified from pKD13-cat using primers glpFKcat1-P1 and glpFKcat2-P2 (Supplementary Table S2). The PCR products were electroporated into wild type, ΔcyaA and ΔcyaA ΔglpR cells to replace the 85-bp downstream region between the 8^th nucleotide and the 94^th nucleotide relative to the glpK stop codon in the chromosome. After electroporation, the cells were selected on LB + Km agar plates. The Km^r colonies were verified for the substitution of the 85 bp glpK/glp intergenic region by PCR and subsequent DNA sequencing. In the resultant strains (named BW_cat, ΔcyaA_cat and ΔcyaA ΔglpR_cat, respectively), glpF, glpK and cat form a single operon with its expression solely under the control of the glpFK promoter (PglpFK).

Strains were cultured in LB, NB or minimal M9 media with various carbon sources at 37°C or 30°C. When appropriate, kanamycin (Km; 25 μg/ml), ampicillin (Ap; 100 μg/ml), or chloramphenicol (Cm; 20–60 μg/ml) was added to the media.

Mutations of chromosomal Crp operators

To modify the chromosomal Crp binding sites in the control region of the glpFK operon, the previously made plasmid pKD13-PglpFK (Zhang and Saier, 2009a), was used. In this plasmid, PglpFK and the FRT-flanking km gene were oriented in opposite directions. Using the quick-change site-directed mutagenesis kit (Agilent) and oligos PglpFK_CrpI&II-F and PglpFK_CrpI&II-R (Supplementary Table S2), both Crp operators (CrpI and CrpII) in the glpFK control region, contained within pKD13-PglpFK, were mutated by changing tatgacgaggcacacacattttaagt (−69 to −44 relative to +1 of PglpFK) to gacagcgaggcatctgcattttaatc (substitutions are underlined). The substitutions were confirmed by sequencing. Using the resultant plasmid, pKD13-PglpFK_O_Crp, as template, the region containing the km gene and PglpFK_O_CrpI&II was PCR amplified using the primers PglpFK_CrpI&II-P1 and PglpFK_CrpI&II-P2 (Supplementary Table S2). The PCR products were integrated into the ΔcyaA_cat and ΔcyaA ΔglpR_cat mutant chromosome to replace the wild type PglpFK. The nucleotide substitutions in both CrpI and CrpII operators were confirmed by sequencing. The km gene was removed, and the resultant strains were named ΔcyaA O_Crp_cat and ΔcyaA ΔglpR O_crp_cat, respectively (Supplementary Table S1).

Glp⁺ mutation assay using a ΔcyaA mutant strain

Using the ΔcyaA deletion mutant, mutation to Glp⁺ was first measured on minimal M9 + 0.2% glycerol agar plates as described previously (Zhang and Saier, 2009a). Briefly, cells from an overnight LB culture were washed and inoculated onto plates (~10⁸ cells/plate). The plates were then incubated in a 30 °C incubator and were examined daily for the appearance of Glp⁺ colonies with each colony representing an independently arising Glp⁺ mutation. On these glycerol minimal agar plates, any colonies appearing by day 2 were considered to be from Glp⁺ cells initially present when applied to the plates. They were therefore subtracted from the subsequent measurements. The total numbers of Glp⁻ cells were determined as described by (Cairns and Foster, 1991). The Glp⁺ mutations were determined by counting the Glp⁺ colonies that appeared on the original agar plates. The frequencies of Glp⁺ mutations on glycerol M9 plates were determined by dividing the numbers of Glp⁺ colonies by the total Glp⁻ populations. To determine if any of the Glp⁺ colonies arose from Glp⁺ cells initially plated, the ΔcyaA cells, together with small numbers of ΔcyaA Glp⁺ cells, were plated onto the same M9 + 0.2% glycerol plates. The plates were incubated and examined as above.

To determine the effect of cAMP on the frequency of IS5 insertion into the glpFK activating site, strain ΔcyaA_cat (in which glpF, glpK and cat are fused in a single operon, see Supplementary Table S1) was used. This strain is sensitive to Cm at 8 μg/ml while the same strain with the IS5 insertion (ΔcyaA Glp⁺_cat) is resistant to Cm at 20 μg/ml. Preliminary experiments showed that all ΔcyaA_cat cells resistant to Cm at 20 μg/ml were due to IS5 insertion in front of PglpFK. To determine the effect of cAMP on IS5 insertion, an 8-h old culture from a single ΔcyaA_cat colony was diluted 1000 × into 5 ml LB ± cAMP (0 to 5 mM) contained in 30 ml glass tubes (2.5 cm × 20 cm). The tubes were shaken at 250 rpm in a 30 °C water bath shaker. After 15 h, the cells were washed 1x (to remove residual cAMP) with carbon source-free M9 salts, serially diluted, and applied onto LB + glucose agar plates and LB + glucose + Cm agar plates. The plates were incubated at 37 °C for 15 to 18 h. Total populations and Glp⁺ populations were determined based on numbers of colonies on LB + glucose plates and on LB + glucose + Cm plates, respectively. The frequencies of Glp⁺ mutation were determined by the ratios of Cm^r populations to total populations.

To see if there is any difference in growth rate between cyaA Glp⁻ and cyaA Glp⁺ cells, these two strains (ΔcyaA:km_cat and ΔcyaA_cat Glp⁺) were cultured in LB without or with cAMP (0 to 1 mM). Briefly, overnight LB cultures from single colonies were diluted into fresh 5 ml of LB ± cAMP (final OD₆₀₀ = 0.005). The tubes were shaken (250 rpm) at 30 °C. Samples were collected every 45 minutes to 1 hour, and their optical cell densities (OD₆₀₀) were measured using a Bio-Rad SmartSpect 3000 machine.

To compare their competitive abilities, the Glp⁻ strain, ΔcyaA:km_cat [in which a Km^r gene subsitutes for cyaA, so that it is resistant to kanamycin (Km^r) but sensitive to chloramphenicol (Cm^s)] and the Glp⁺ strain, ΔcyaA_cat Glp⁺ (that is Cm^r but Km^s), were mixed in a 1:1 ratio in the same tubes. The tubes were agitated (250 rpm) at 30 °C. Samples were taken after 5, 10 and 15 hours of growth and serially diluted. The dilutions were applied onto LB + Km plates for determination of Glp⁻ cells and LB + Cm plates for determination of Glp⁺ cells.

To examine if the presence of glpFK has an effect on IS insertions, the cat gene was substituted for the glpFK operon. In the resultant strain, the cat gene is directly under the control of the glpFK promoter (Zhang and Saier, 2009a). This construct was moved to a ΔcyaA ΔglpR background by P1 transduction, yielding strain ΔcyaA ΔglpR_cat2 (see Table S1). The same methods as above were used to determine cAMP effects on IS5 insertion.

To determine if cAMP affects IS5 insertion into another chromosomal site, we chose to analyze mutants resistant to Furazolidone (FZD) using ΔcyaA_cat cells. First, step I mutants were isolated by spreading the cells onto nutrient broth (NB) agar plates with a low concentration (1 μg/ml) of FZD. The cells of a step I mutant were then applied onto the agar plates with higher concentrations (5–7.5 μg/ml) of FZD. The plates were incubated at 30 °C for 36 h or more, and colonies obtained were examined for the presence of IS elements in the nfsB gene by PCR using oligos nfsB-ver-F and nfsB-ver-R (Supplementary Table S2). Among those mutants carrying IS elements (such as IS1, IS2 and IS5), IS5 insertional mutants were determined by two rounds of PCR, using oligos IS5-ver-F / nfsB-ver-R and nfsB-ver-F / IS5-ver-F, respectively. The ratio of IS5 mutants to total IS insertion mutants was calculated by dividing total IS mutant numbers by IS5 mutant numbers.

To establish the effects of mutations in the Crp operators in the glpFK control region on the appearance of Glp⁺ IS5 insertional mutations, the ΔcyaA O_Crp_cat cells with mutations in operators, CrpI and CrpII, were examined for the appearance of Glp⁺ mutations in LB media with or without cAMP as described above. To determine the effect of loss of glpR on Glp⁺ mutation frequency, the ΔcyaA ΔglpR double mutant was examined for Glp⁺ mutations in liquid LB ± cAMP (0.1 mM) as compared to the single ΔcyaA mutant as described above.

To determine the effect of glpR overexpression on the appearance of Glp⁺ mutations, the glpR structural gene was amplified from the wild type genomic DNA using primers glpR-KpnI and glpR-BamHI (Supplementary Table S2). The PCR products were digested with KpnI and BamHI, gel purified, and then ligated to the same sites of pZA31 (Lutz and Bujard, 1997), yielding pZA31-glpR, in which glpR is driven by a synthetic tet promoter (Ptet). The same plasmid carrying a random fragment (RF), pZA31-RF, served as a control (Levine et al., 2007). To repress Ptet activity, the constitutively expressed tetR cassette, located at the attB site, was transferred to ΔcyaA_cat carrying pZA31-glpR from BW-RI (Levine et al., 2007) by P1 transduction. Therefore, the expression of glpR could be induced using a tetracycline analog, chlorotetracycline (cTc). The resultant strain containing the tetR repressor, and pZA31-glpR was tested for Glp⁺ mutations in LB ± 0.1mM cAMP as described above. To induce expression of glpR in pZA31-glpR, cTc (250 ng/ml) was added to the medium.

To further demonstrate cAMP inhibitory effects on the appearance of IS5 insertional mutants, and the competitive abilities of the mutants under low cAMP conditions, we performed a short-term experiment by transferring cultures to new media at various intervals. To do this, an LB culture (10 μl) of ΔcyaA_cat was used to inoculate M9 + glycerol ± cAMP (0 to 1 mM) media. Before the first transfer, when all cells were in the original test tube, samples were removed and analyzed about every 12 h. The cultures were serially diluted onto LB + glucose plates for total population determination and onto LB + glucose + Cm plates for Glp⁺ (IS5 insertion) mutant population determination. After about 2 or 2.5 days, the cultures were 1000x diluted into new tubes with the same media. Then at one-day intervals, the cultures were 1000x diluted into fresh media. For each transfer, the total cells and the Glp⁺ mutant populations were determined as stated above.

IS5 insertional mutation assay using a wild type genetic background

2-deoxyglucose (2DG) is a non-metabolizable glucose analog that reduces the level of cytoplasmic cAMP level when added to the media (Saier, 1989). Preliminary experiments showed that E. coli cells are sensitive to this compound at 0.1%. To determine if IS5 insertion upstream of PglpFK occurred in a wild type background, BW_cat cells (Supplementary Table S1) were tested for IS5 insertion on M9 + glycerol (0.2%) + 2DG (0.12%) ± Cm (60 μg/ml) agar plates. The plates were incubated at 30 °C, and the colonies were examined for the presence of IS5 in the upstream glpFK operon control region by PCR followed by gel electrophoresis.

Long term evolutionary experiments

At least two types of mutations arose when BW_cat cells were incubated on M9 + glycerol + 2DG ± Cm agar plates, the IS5 insertional mutation and a non-IS5 mutation of unknown nature. To determine if the IS5 insertional mutants are more competitive than the non-IS5 insertional mutants, 10 μl of a fresh LB culture of BW_cat was used to inoculate 5 ml of M9 + glycerol (0.2%) + 2DG (0.12%) ± Cm (60 μg/ml) in 30 ml glass tubes. The tubes were incubated with shaking (250 rpm) in a 30 °C water bath shaker. After 2.5 days of incubation, the cultures (i.e., mutant cells resistant to 2DG and Cm) were 1000x diluted into new tubes with the same media. Every two days the mutants were 1000x diluted into new media of the same composition. For analysis of the cell populations after each transfer, the mutant cultures were serially diluted using carbon source-free M9 salts, and the 10⁵-fold and 10⁶-fold dilutions were applied onto M9 + glycerol + 2DG ± Cm agar plates before incubation at 37 °C. After 2 days, 100 colonies from each transfer were subjected to PCR and subsequent gel electrophoresis analyses to determine the percentages of IS5 insertion mutants to the total mutants. Note that the parental cells do not grow under these conditions. This experiment was also conducted without Cm to show that the beneficial consequences of IS5 insertion were not related to the cat gene fusion.

Chromosomal lacZ fusions and β-galactosidase assays

Using pKD13-PglpFK (Zhang and Saier, 2009a) and pKD13-PglpFK_O_Crp (Supplementary Table S1) as templates, PglpFK (−204 to +66 relative to the transcriptional start site) and PglpFK_O_Crp plus their upstream FRT-flanked km^r gene were amplified using oligos PglpFKz-P1 and PglpFKz-P2 (Supplementary Table S2). Using the method of (Datsenko and Wanner, 2000), the promoters plus the upstream km^r gene (km^r:PglpFK or km^r:PglpFK_O_Crp) were integrated into the chromosome to replace the lacI gene and the native lac promoter (including the 5′ UTR of lacZ) of MG1655, deleted for lacY (Klumpp et al., 2009). This chromosomal replacement was confirmed by PCR and subsequent DNA sequencing analysis. The resultant strains are deleted for both lacI and lacY, but they carry the lacZ gene that is expressed under the control of PglpFK or PglpFK_O_Crp. Both constructs were transferred into ΔcyaA and ΔcyaA ΔglpR strains by P1 transduction, yielding ΔcyaA_PglpFK-lacZ, ΔcyaA_PglpFK_O_Crp-lacZ, ΔcyaAΔglpR _PglpFK-lacZ, and ΔcyaAΔglpR _PglpFK_O_{Crp_}lacZ (Supplementary Table S1).

For β-galactosidase assays, strains were cultured in liquid LB media ± 1 mM cAMP at 30 °C. When cultures entered the exponential phase, samples were collected for measurement of β-galactosidase activities as described by (Miller, 1972).

Results

IS5 insertional activation of the glpFK operon occurs in ΔcyaA cells

We previously demonstrated that Δcrp mutant cells of E. coli could regain the ability to utilize glycerol by IS5-mediated insertional mutations that specifically occurred at a single site, upstream of the glpFK promoter, preferentially in the presence of glycerol (Zhang and Saier, 2009a). In this study, we first used ΔcyaA mutant cells lacking the cAMP biosynthetic enzyme, adenylate cyclase (Cya). Like the Δcrp mutant cells, the ΔcyaA mutant cells cannot utilize glycerol as the sole carbon source for growth. However, after a prolonged incubation on M9 + glycerol agar plates, Glp⁺ mutants appeared (Figure 1). These mutants were IS5 insertional mutants carrying IS5 in the same position and orientation upstream of the glpFK promoter as those isolated previously from Δcrp cells. The time course (Figure 1) for their appearance and the properties of these double mutants were indistinguishable from those isolated previously (Zhang and Saier, 2009a). Among at least 100 independent Glp⁺ mutants analyzed, no other types of mutants arose under the conditions used. There was an approximately two-day delay before mutant colonies appeared, and these mutants clearly arose during incubation on the plates, since when small numbers (e.g. 11 and 23) of identical Δcya Glp⁺ insertional mutants were added to the ΔcyaA cells prior to plating, they gave rise to colonies within a shorter time period (Figure 1).

Figure 1. Glp⁺ mutations in ΔcyaA cells on M9 + glycerol agar plates.

ΔcyaA cells (~10⁸) from a fresh LB culture were spread on M9 + glycerol (0.2%) agar plates. The plates were incubated at 30°C and examined for the appearance of Glp⁺ colonies (each colony represents an IS5 insertional mutation) at 24h intervals. The mutation frequencies were determined as in Zhang and Saier (2009a). ◆ = no Glp⁺ cells added before initially plating; ■ and = 11 and 23 Glp⁺ (IS5 insertional mutant) cells were included before plating.

IS5 activation of the glpFK operon in ΔcyaA cells is suppressed by exogenous cAMP

Exogenous cAMP can enter cells to increase the cytoplasmic concentration of this nucleotide (Saier et al., 1982). Figure 2A shows the effect of increasing concentrations of external cAMP on the glpFK-specific IS5 insertional frequency in ΔcyaA_cat cells (see Materials and Methods and Supplementary Table S1). At a concentration of 10 μM, exogenous cAMP had only a slight inhibitory effect on IS5 insertion, but at 100 μM, cAMP inhibited over 90%, whereas at 1 mM, cAMP essentially abolished IS5 insertion (Figure 2A).

Figure 2. Effects of cAMP on (A) IS5 insertion upstream of the glpFK regulatory region in ΔcyaA cells, and (B) in the nfsB gene as a control.

In A, a fresh LB culture from a single ΔcyaA-cat (see Supplementary Table S1) colony was diluted 1000 × into 5 ml of LB ± cAMP (0 to 5 mM) in 30 ml glass tubes (2.5 cm × 20 cm). The tubes were shaken at 250 rpm in a 30 °C water bath shaker. After ~15 h, the cells were washed 1x (to remove residual cAMP) with a carbon source-free M9 salts solution, serially diluted, and applied onto LB + glucose agar plates (for total population determination) and LB + glucose + Cm (20 μg/ml) agar plates (for IS5 insertional mutant population determination). In B, the cells (~2x 10⁸) of a step I furazolidone (FZD) resistant (FZD^r) mutant strain isolated from a ΔcyaA-cat strain, were applied to nutrient broth (NB) agar plates, with FZD (5 or 7.5 μg/ml) ± cAMP (1mM). The plates were incubated at 30 °C for 36 h before being examined for the appearance of FZD^r mutants. Among these FZD^r mutants, IS5 and other IS insertional mutants in the nfsB gene were determined and quantitated by PCR (see Materials and Methods). In all cases, the proportions of IS5 mutants (~60%) was the same.

To determine whether the decrease in IS5 insertion frequency due to the presence of cAMP was specific to the glpFK promoter, we analyzed IS5 insertion at the nfsB gene in ΔcyaA cells. Mutational inactivation of nfsB confers resistance to furazolidone (FZD) (Whiteway et al., 1998) and a substantial fraction of inactivating mutations (20%) are due to IS insertions. Of these, 60% are due to IS5. As shown in Figure 2B, cAMP did not influence the frequency of either total insertional events (grey bars) or of IS5 insertional events (black bars) among FZD-resistant mutants within experimental error.

Control experiments that eliminate preferential growth and selection as causes of increased IS5 insertion

We have conducted control experiments to eliminate the possibility that the changes in IS5 insertion observed in response to cAMP were due to differentiated growth or to the selection procedure used. The results are presented in Figure 3 and Figure S1. We grew the cyaA Glp⁻ and cyaA Glp⁺ (IS5 insertion) cells separately (Figure 3A) or together in a 1:1 ratio (Figure 3B). It can be seen that the two strains grew in LB medium at essentially the same rate when the exogenous cAMP concentration was 0, 0.1 or 1.0 mM (Figure 3A). The duration of the experiment was 15 h (11 generations), the duration of the experiments reported in Figure 2 and Figures 4–6 (see below). Similarly, when cells were first mixed in a 1:1 ratio and then grown for the same period of time, the ratio remained constant within 10%, the error of the measurements (Figure 3B). This was true regardless of whether cAMP was absent or present at strongly inhibitory concentrations (see Figure 2).

Figure 3. Growth rates and competition assays for ΔcyaA Glp⁻ and ΔcyaA Glp⁺ cells.

Strains ΔcyaA:km_cat (Glp⁻, Km^r and Cm^s ) and ΔcyaA_cat Glp⁺ (Glp⁺, Km^s and Cm^r ) (see Table S1) and LB ± cAMP (0 to 1mM) were used for these assays. In A, these two strains were cultured in separate tubes. The tubes were shaken at 250 rpm at 30°C. Samples were taken every 45 minutes to 1 hour, and their optical cell densities (OD₆₀₀) were measured. ■, ▲ and ● = ΔcyaA:km_cat. □, △ and ○ = ΔcyaA_cat Glp⁺. In B, the two strains were mixed at a 1:1 ratio in the same tube. After 5, 10 and 15 hours of growth with shaking at 30°C, samples were taken to determine the populations of each strain. The cell samples were serially diluted and applied onto LB + Km plates for ΔcyaA:km_cat Glp⁻ population determination and LB + Cm plates for ΔcyaA_cat Glp⁺ population determination. ● = no cAMP; ■ = 1mM cAMP.

Figure 4. Effect of Crp binding site mutations on (A) IS5 insertion upstream of the glpFK promoter region and (B) the promoter activity in ΔcyaA and ΔcyaA ΔglpR cells.

In both A and B, cells were grown in LB medium ± cAMP. The concentrations of cAMP are indicated below the x-axes. O_crp indicates the mutated Crp binding sites (CrpI and CrpII) in the upstream regulatory region of the glpFK operon to prevent Crp binding. In A, strains used carry the cat gene fused to and downstream of glpFK. In B, the activities of the glpFK promoter (PglpFK) and the same promoter (PglpFK_O_crp) mutated in CrpI and CrpII were measured using the LacZ reporter in both ΔcyaA cells and ΔcyaA ΔglpR cells. See Supplementary Table S1 for the detailed strain information.

Figure 6. IS5 insertion upstream of the glpFK control region in wild type cells.

Wild type cells carrying the glpFK_cat (chloroamphenicol acetyl transferase) fusion (i.e., BW_cat, see Methods section) were plated onto M9 + glycerol (0.2%) + 2DG (0.12%) + chloramphenicol (Cm) (60 μg/ml) agar plates. The plates were incubated at 30 °C. After 5 days of incubation, colonies were examined for the presence of IS5 upstream of the glpFK control region by PCR. In both the upper and the lower panels, only one type of IS5 insertion was obtained. A DNA marker showing four bands with known sizes (4 kb, 2 kb, 1 kb and 0.5 kb from above) is indicated in each panel.

In a separate experiment, we used an isogeneic strain in which the glpFK promoter was fused to a cat (chloramphenicol acetyl transferase) gene with both the glpF and glpK genes deleted. The experiment (Figure S1) was conducted without and with 0.1 mM cAMP in a ΔcyaA ΔglpR double mutant. There can not be selection on glycerol with such strains because glycerol utilization is absolutely dependent on GlpK. It can be seen that 0.1 mM cAMP strongly inhibited the frequency of IS5 insertion. Thus, these experiments separate mutation rate from any possible effect of glycerol utilization, growth, or selection. These results confirm our earlier report (Zhang and Saier, 2009a) demonstrating that IS5 insertion into the glpFK promoter activating site was dependent on GlpR in a crp genetic background under conditions where the glpFK promoter was fused directly to the cat gene in the absence of GlpF and GlpK function.

The effect of cAMP on the IS5 insertion frequency requires the Crp binding sites in the glpFK promoter region

To determine whether the inhibition of IS5 insertion upstream of the glpFK promoter by cAMP is due to the binding of the cAMP-Crp complex to the two adjacent Crp binding sites (CrpI and CrpII), present in the glpFK promoter region, we analyzed the consequences of point mutations within these binding sites. These mutations essentially abolished the inhibitory effect of Crp on IS5 insertion (Figure 4A). Although these Crp operator mutations eliminated binding of the cAMP-Crp complex, they did not change the glpFK promoter strength in the cyaA deletion background, and the promoter activity was still under the control of GlpR (Figure 4B). Thus, it can be concluded that inhibition by cAMP-Crp of IS5 insertion into the glpFK activating site is due to the binding of the cAMP-Crp complex solely to these two operators present in the glpFK promoter region.

Crp and GlpR independently affect IS5 insertion upstream of the glpFK promoter

To determine if GlpR plays a role in the inhibitory effect of Crp on IS5 insertion, we deleted the glpR gene in the ΔcyaA background, yielding a ΔcyaA ΔglpR double mutant (Supplementary Table S1). Higher IS5 insertional frequencies were observed in the ΔcyaA ΔglpR cells than in the ΔcyaA cells when grown in LB (no glycerol added) (Figure 5A). However, in the absence of GlpR, cAMP still exerted its inhibiting effect, presumably by binding to Crp, which then bound to its two glpFK operon binding sites (compare column 2 and column 4 in Figure 5A). Comparable inhibition was observed regardless of the presence of glycerol or GlpR. It was therefore concluded that regulation of IS5 insertion by the cAMP-Crp complex occurs independently of glycerol and GlpR.

Figure 5. Effects of glpR deletion (A) and glpR overexpression (B) on IS5 insertion in the control region of the glpFK operon in ΔcyaA cells grown in LB ± cAMP.

In A, the exogenous cAMP concentration was either 0 (left) or 0.1 mM (right). In B, vector: no glpR present (control); GlpR indicates that the glpR gene is present downstream of the tet promoter, Ptet, in an expression vector. cTc = chloro-tetracycline (an inducer of the tet promoter) at a concentration of 0 (left) or 250 ng/ml (right). glpR expression is thus under the control of the cTc induced promoter, Ptet, in the expression vector. ΔcyaA cells that constitutively produce TetR (repressing Ptet) were used in these experiments. Note the scale difference for Figures 4A and 4B. All strains carry the cat gene fused to and downstream of the glpFK genes.

When GlpR was over-produced in the absence of cAMP, GlpR still exerted its strong inhibitory effect (Figure 5B). In the left panel, the glpR gene was expressed at an extremely low level, and GlpR exerted only a minimal effect because chlorotetracycline (cTc), the inducer, was not present. When the cTc concentration was high (250 ng/ml), glpR was expressed at a high level, and the rate of IS5 insertion into the glpFK upstream site was greatly reduced (see right panel of Figure 5B). The same experiments were conducted in the presence of cAMP (0.1 mM). The IS5 insertion frequency decreased, while overexpression of GlpR further inhibited IS5 insertion (data not shown). These results suggest that GlpR and the cAMP-Crp complex exert their effects on IS5 insertion independently of each other. These experiments also provide evidence that GlpR and the cAMP-Crp complex can bind to the glpFK control region simultaneously.

IS5 insertional activation of the glpFK operon occurs in a wild type background in the presence of 2-deoxyglucose

Since a reduction in the cytoplasmic cAMP concentration promotes IS5 insertion in the glpFK activating site, we sought to determine whether environmental conditions that lead to a reduction in cAMP concentrations elevate IS5 insertion into the glpFK promoter. Non-metabolizable glucose analogues, such as 2-deoxyglucose (2DG) and α-methylglucoside (αMG) (He and Liu, 2002; Holst and Williamson, 2004; Kumar et al., 2013; Moller, 2010; Tantanarat et al., 1996) are known to lower cytoplasmic cAMP levels by inhibiting adenylate cyclase activity (Gabor et al., 2011, Gershanovich, 2003, Saier et al., 1996, Vastermark and Saier, 2014). These analogues also strongly inhibit growth on glycerol, at least in part due to inhibition of both cytoplasmic cAMP production by adenylate cyclase and of cytoplasmic glycerol-3-phosphate (substrate/inducer) production by glycerol kinase (Kuroda et al., 2001, Peterkofsky et al., 2001, Schlegel et al., 2002, Saier and Reizer, 1994). We therefore asked if we could isolate IS5 insertional mutants in a wild-type background on minimal M9 agar plates containing glycerol, inhibitory concentrations of 2DG or αMG and chloramphenicol. The results for αMG proved to be same as for 2DG, and consequently, only those obtained with 2DG are presented here.

In these experiments, a glpFK_cat (chloramphenicol acetyl transferase) fusion (strain, BW_ cat) was used to measure IS5 insertion (see Supplementary Table S1 and Materials and Methods). Regardless of the inhibitory glucose analogue used, analogue-resistant mutants (which were also Glp⁺) could be isolated in a wild type E. coli genetic background (Jones-Mortimer and Kornberg, 1980, Kornberg et al., 2000, Rephaeli and Saier, 1980). As shown in Figure 6, PCR analyses of colonies that appeared after a five-day incubation at 30°C revealed two types of Glp⁺ mutants: (i) a majority that did not have an insertion in the glpFK promoter (dubbed “non-IS5 mutants” here), and (ii) a minority (<10%; “IS5 mutants”) that had IS5 inserted into the glpFK promoter-activating site.

Sequencing revealed that the non-IS5 insertion mutants did not have genetic alterations in the glpFK operon, or in the fruR (cra) gene [which appears to regulate crp gene expression (Zhang et al., 2014)]. Unlike IS5 insertional mutants, the non-IS5 mutants showed a pleiotropic phenotype in addition to their increased growth in a glycerol + 2DG medium, such as poor utilization of sorbitol (D-glucitol) and succinate (Table 1) which are utilized efficiently only when high cytoplasmic concentrations of the cAMP-Crp complex are available (Zhang et al., 2014).

Table 1.

Growth of a wild type strain, an IS5 insertional mutant and a non-IS5 mutant on minimal M9 agar plates supplemented with various carbon sources¹,².

Strain	Glycerol	Sorbitol	Succinate	Xylose	Mannitol
Wild type	++	++	++	++	++
IS5 mutant	+++	++	++	++	++
Non-IS5 mutant	++	+	+	++	++

¹Overnight LB cultures from single colonies were washed once with a carbon source-free M9 salts solution. The pellets were resuspended in the same M9 solution, and cells were inoculated onto the agar plates by streaking using a sterile transfer loop. The plates were incubated at 30 °C for 36 h before being examined for colony sizes.

²+, ++ and +++ denote the relative sizes of the colonies. All values are relative to growth of the wild type strain (++) on agar plates containing a single carbon source as indicated. Thus, +++ indicates increased colony size while + indicates decreased size.

To demonstrate that the Glp⁺ phenotype of IS5 mutants results solely from the IS5 insertional event, we carried out P1 transduction experiments to determine if the phenotypes of 2DG^r, Cm^r and Glp⁺, could be transferred together into another E. coli genetic background. In these experiments, we used two “wild type” strains, BW25113 and BW_ cat (Supplementary Table S1) (with or without the glpFK_cat fusion) as recipients. For both recipient strains, transductants were obtained using the IS5 insertional mutant (expressing the fused glpFK_cat operon) as donor, when plated on M9 + glycerol + 2DG ± Cm agar plates. The regulatory regions of 23 independently isolated transductants were amplified by PCR, and all were found to carry the IS5 element. DNA sequencing showed that IS5 was located in the same position and in the same orientation as described previously (Zhang and Saier, 2009a, Zhang and Saier, 2009b). These results showed that in these mutants, IS5 insertion is necessary and sufficient to give rise to the 2DG^r Glp⁺ phenotype.

Short-term and long-term evolution experiments to evaluate if IS5-mediated activation of glpFK operon expression could have evolved in wild type cells

Initially, we conducted short-term evolutionary experiments using ΔcyaA_cat cells in which a chloramphenicol-resistance gene was transcriptionally fused to the glpFK operon such that glpFK activation by IS5 insertion led simultaneously to a Glp⁺ and a Cm^r phenotype (see Materials and Methods). These short-term experiments (several transfers) were conducted by incubating ΔcyaA cells in the presence of glycerol alone (Figure 7A), glycerol plus 0.1 mM cAMP (Figure 7B), and glycerol plus 1.0 mM cAMP (Figure 7C). In the absence of cAMP, IS5 insertional mutants appeared as the only species after 37 h of incubation with shaking. After several subsequent transfers, virtually 100% of the cells contained IS5 in the glpFK-activating site (Figure 7A). When cAMP was added at 0.1 mM, IS5 insertional mutants again appeared as the only species after a 61 h incubation with shaking. After the first transfer, virtually all cells contained IS5 (Figure 7B). When exogenous cAMP was added at 1 mM, the appearance of these mutants was strongly inhibited. As a result, after the 6th transfer, only about 20% of the population were IS5 insertional mutants (Figure 7C). These observations are consistent with the conclusion that the cAMP-Crp complex inhibits IS5 insertion upstream of the glpFK operon.

Figure 7. Percentages of IS5 mutant populations vs the total populations during growth in M9 + glycerol ± cAMP over time using strain ΔcyaA_cat.

For the first period [up to 61 h (A and B) or 45 h (C)], time is expressed in hours. Thereafter (vertical dotted line), time is expressed in numbers of transfers where each transfer occurred about once per day. Figures A–C show the effects of increasing cAMP concentrations: A, 0; B, 0.1 mM; C, 1 mM.

In order to conduct long-term evolutionary experiments in a wild type genetic background, we initially used glpFK_cat cells (i.e., BW_cat) in which a chloramphenicol-resistance (cat) gene was transcriptionally fused to the glpFK operon such that glpFK activation by IS5 insertion led simultaneously to a Glp⁺ and Cm^r phenotype (see Materials and Methods). Prolonged incubation of glpFK_cat wild type cells under conditions analogous to those described above for the short-term ΔcyaA experiments revealed that IS5 mutants became an appreciable fraction of the population after several generations in minimal glycerol medium with 2-deoxyglucose, with (Figure 8A) or without (Figure 8B) chloramphenicol. When chloramphenicol was present, IS5 insertional mutants first appeared after six transfers, and then continued to accumulate during the remainder of the experiment (17 transfers), after which about 25% of the cells bore the IS5 insertion (Figure 8A). When chloramphenicol was absent, it took a little longer; insertional mutants became appreciable following the eighth transfer, and accounted for over 20% of the cells after 20 transfers (Figure 8B). In both experiments, the percentages were still increasing when the time course was terminated. These results show that glpFK activation by IS5 is advantageous when wild type cells are exposed to glycerol in the presence of a sugar analogue inhibitor of adenylate cyclase such as 2-deoxyglucose (Novotny et al., 1985). The process could, therefore, have evolved by natural selection.

Figure 8. Percentages of IS5 insertional mutant populations versus total mutant populations over time using the “wild type” strain carrying the chromosomal glpFK_cat fusion medium.

Cells were incubated with shaking (250 rpm) at 30 °C in 5 ml of M9 + glycerol (0.2%) + 2DG (0.12%) medium plus (A) or minus (B) chloramphenicol (Cm) (60 μg/ml) in glass tubes (20 mm × 200 mm). After 2.5 days of incubation, 5 μl culture samples were transferred to 5 ml of the same media (1^st transfer) and incubated with shaking at 30 °C. Later, every two days, 5 μl cultures were transferred to new media and grown under the same conditions. For every transfer, the cultures were diluted with carbon source-free M9 salt solution and plated onto M9 + glycerol + 2DG + Cm agar plates. After 2 days of incubation at 30 °C, 100 colonies were examined by PCR for the presence of IS5 in the glpFK control region (See Figure 6).

Discussion

Primary features of the IS5-glpFK experimental system

In wild type E. coli cells, glycerol utilization requires expression of genes that are poorly expressed in the presence of glucose or a glucose analogue. The first operon in the glycerol utilization pathway, glpFK, codes for the glycerol uptake facilitator (GlpF) and glycerol kinase (GlpK) that phosphorylates glycerol to glycerol-3-phosphate (G3P). G3P, the inducer of the glp regulon, is then oxidized by GlpD to dihydroxyacetone phosphate, an intermediate of glycolysis. The glpFK operon is repressed by the binding of the GlpR repressor to its four binding sites (operators) in the promoter region, and is activated by the binding of the cAMP-Crp complex to its two binding sites that partially overlap the GlpR binding sites (Zhang and Saier, 2009a; Weissenborn et al., 1992). G3P binds to and inactivates GlpR, which, however, is insufficient to activate the weak native glpFK promoter, the activation of which requires the binding of the cAMP-Crp complex to the two binding sites (CrpI and CrpII) in the promoter. Thus, cells lacking the crp (Crp) or cyaA (biosynthesis of cAMP) gene are Glp⁻ because the glpFK operon cannot be activated.

Prolonged incubation of Δcrp (Glp⁻) cells or ΔcyaA (Glp⁻) cells on glycerol minimal plates results in the continuous generation of Glp⁺ colonies. Sequence analysis of several hundred independent Glp⁺ colonies revealed that in each and every case, IS5 is found to be inserted upstream of the glpFK promoter at the same specific location and in the same orientation. Subsequent work established several key features of this glpFK-activating insertional mutation as reviewed elsewhere (Zhang et al., 2013; Saier and Zhang, 2014, Saier and Zhang, 2015), and we have interpreted the results to mean that the IS5-glpFK system represents an example of a novel mutagenic mechanism.

Formation and selection of mutations under stressful conditions

Whether specialized mutagenic mechanisms are activated in response to environmental conditions has been the subject of active investigation for nearly three decades. Much of the work addressing this question is based on the so-called Foster-Cairns E. coli experimental system [recent review: (Williams and Foster, 2012)]. This system is based on the E. coli strain FC40/F’lac which cannot utilize lactose because a frameshift mutation prevents expression of the gene (lacZ) for β-galactosidase, an enzyme required for lactose utilization. However, when FC40 cells are subjected to prolonged incubation on minimal lactose plates, on which Lac⁻ cells cannot grow, Lac⁺ colonies appear, starting on day 3, and accumulate linearly over several subsequent days. Foster and coworkers provided evidence that these “adaptive” or “stress-induced” mutations are dependent on DinB (DNA Polymerase IV) and homologous recombination functions, and are due to error-prone DNA synthesis associated with the processing of recombination intermediates (Williams and Foster, 2012). Rosenberg and coworkers proposed an alternative scenario, suggesting that adaptive mutations arose in a hypermutable subset of cells in which starvation stress activated a specialized error-prone double-stranded DNA break repair pathway (Rosenberg et al., 2012, Rosenberg et al., 1998).

Without disputing the experimental observations made by proponents of adaptive mutagenesis, Roth and coworkers proposed that the Lac⁺ colonies in the Foster-Carins system arose solely from mutation-capture by selection, a mechanism to be contrasted with induction of a novel mutagenic mechanism [for an updated summary of this view, see: (Maisnier-Patin and Roth, 2015)]. In this “capture” model, all Lac⁺ revertants arise from a pre-existing subset of cells in which the F plasmid is replicated without concomitant cell division. However, Stumpf et al. (2007) have provided concrete evidence that gene amplification cannot account for adaptive mutation in the E. coli lac system. Both the “adaptive mutagenesis” and “selection-capture” hypotheses, although originally proposed to explain mutations in the Foster-Carins system, are extendable to other experimental systems with minor modifications.

The IS5-glpFK experimental system, when Glp⁻ cells are subjected to prolonged incubation on minimal glycerol medium on which they cannot grow, Glp⁺ colonies appear after day 3. However, because the mutation is due to transposon (IS5) insertion at a single specific location and orientation, there is no need for the participation of either cellular DNA polymerases or homologous recombination. Not surprisingly, loss of the recA gene has little effect on IS5-glpFK mutation (Zhang and Saier, 2009a).

“Directed Transposition” vs. “Selected Transposition” models

In our “directed transposition” model, we propose that the DNA sequence upstream of the glpFK promoter is a conditional hotspot (Z. Humayun, Z. Zhang & M. Saier, unpublished results) for IS5 insertion that is somehow masked by the vicinal binding of one or more DNA binding proteins. In Δcrp cells, GlpR is known to bind near the IS5 target sequence to suppress IS5 insertion (Zhang and Saier, 2009a) and in this work, Crp has now been shown to also suppress IS5 insertion. In this model, passive diffusion of glycerol, or entry via GlpF (Richey and Lin, 1972), followed by its phosphorylation to glycerol-3-phosphate (G3P) by leaky expression of GlpK leads to unmasking of the IS5 target site.

Under the plausible “mutation-capture” hypothesis (Maisnier-Patin and Roth, 2015), a subset of the plated Glp⁻ cells amplify the glpFK region and are selectively subjected to the “positive feedback loop” in which availability of small amounts of energy from leaky expression of GlpFK leads to further amplification of the glpFK region. Either the accumulation of G3P, which inactivates GlpR, or simple titration of the limited amount of the GlpR protein by the amplified excess of glpFK DNA, or both, may lead to the unmasking of the IS5 insertion site. It is important to note, that regardless of the mechanism(s) by which binding of GlpR and the cyclic AMP-Crp complex inhibits IS5 insertion, the system appears to exhibit the characteristics defined as “directed mutation” (Cairns et al., 1988).

IS5-glpFK mutation as a “last resort” option for survival

Regardless of the mechanism considered above, the discovery that GlpR occupation of O₁ (Zhang and Saier 2009a) and cAMP-Crp occupation of CrpI and CrpII (this communication) inhibit IS5 insertion suggests that the IS5 insertion target is unmasked in wild-type cells only under a specific set of environmental circumstances. Wild type cells preferentially use glucose and repress the pathways required for the utilization of other carbon sources as long as glucose continues to be available in the medium. This repression (confusingly named “catabolite repression”) is accomplished by blocking cAMP synthesis, and therefore, the formation of the cAMP-Crp complex. Since the cAMP-Crp complex is a required positive transcriptional activator for genes necessary for utilizing alternative sugars under normal conditions of glucose depletion, the cAMP-Crp complex is expected to bind to and activate most operons subject to catabolite repression, including the glpFK operon. Thus, insertional activation in the IS5-glpFK system can be seen as a “last resort” survival mechanism that is activated only when glycerol is present, but utilization is blocked due to an environmental condition that interferes with the production of the cAMP-Crp complex, such as occurs in the presence of a non-utilizable glucose analogue, as shown here. We note that numerous toxic and non-metabolizable sugar analogues are synthesized by microorganisms, plants, fungi and man. They include deoxy sugars such as 2-deoxyglucose, methylated sugars such as 3- and 6-O methylglucose, fluoro-sugars, and a variety of α- and β-glycosides such as methyl α-glucoside (He and Liu, 2002, Holst and Williamson, 2004, Kumar et al., 2013, Moller, 2010, Saier and Ballou, 1968, Tantanarat et al., 2012, Xi et al., 2014).

Other IS-mediated stress-induced mutations

In addition to the IS5-glpFK system, a number of reports describe beneficial transposition events that appear to relieve a variety of environmental stress conditions. Target operons include the flhDC operon encoding the flagellar master regulator, FlhDC (Wang and Wood, 2011), zinc inducible zinc resistance (Vandecraen et al., 2016), β-glucoside utilization (Schnetz and Rak, 1992), growth on propandiol involving the fuc regulon (Zhang et al., 2010), and resistance to toxic nitro-substituted compounds (Whiteway et al., 1998). In an early review article, Hall summarized several situations where transposon insertions activate cryptic operons apparently at elevated frequencies under conditions of starvation (Hall, 1999).

The example of IS5 insertion that occurs at higher frequencies when beneficial, giving rise to activation of the E. coli flagellar master regulator operon, flhDC, than when neutral or detrimental (Wang and Wood, 2011), seems particularly relevant to the proposal of directed mutagenesis. Insertional activation of flhDC substantially enhances bacterial migration through semisolid agar media, and this phenotype is beneficial under adverse conditions such as nutrient depletion or a need to escape a toxic substance. Under conditions where swarming is not beneficial or permitted (i.e., in liquid or on solid media, respectively), insertional activation of the flhDC operon occurs at greatly reduced frequencies. Thus, insertion of IS5 upstream of the flhDC operon is another example of a mutation whose frequency is elevated under conditions where the mutation is advantageous. We have now confirmed and extended the observations of Wang and Wood, 2011 in our laboratory (Zhang et al., 2013; Kukita et al., manuscript submitted for publication). Few mechanistic molecular details are available for these experimental systems, but they have the potential of testing the generality of hypotheses that seek to explain the relationships between stress and transposon-mediated mutations that can relieve such stresses. Further work will be required to define the details of these processes.

Highlights.

1. IS5 inserts into the upstream regulatory region of the glpFK operon to activate its promoter.

2. This process is negatively regulated by GlpR as shown previously.

3. We here show that the binding of cAMP-Crp to its binding sites in the same operon negatively controls IS5 insertion into the activating site.

4. The effects of GlpR and Crp are independent of each other.

5. Control experiments show that insertion is independent of the selection procedure.

6. Wild type cells are subject to this mutagenic mechanism when glycerol is present and cAMP levels are low.

7. We believe this process evolved via natural selection.