|Home | About | Journals | Submit | Contact Us | Français|
In Escherichia coli, the adaptor protein SprE (RssB) controls the stability of the alternate sigma factor RpoS (σ38 and σS). When nutrients are abundant, SprE binds RpoS and delivers it to ClpXP for degradation, but when carbon sources are depleted, this process is inhibited. It also has been noted that overproduction of SprE is toxic. Here we show that null mutations in pcnB, encoding poly(A) polymerase I (PAP I), and in hfq, encoding the RNA chaperone Hfq, suppress this toxicity. Since PAP I, in conjunction with Hfq, is responsible for targeting RNAs, including mRNAs, for degradation by adding poly(A) tails onto their 3′ ends, these data indicate that SprE helps modulate the polyadenylation pathway in E. coli. Indeed, in exponentially growing cells, sprE deletion mutants exhibit significantly reduced levels of polyadenylation and increased stability of specific mRNAs, similar to what is observed in a PAP I-deficient strain. In stationary phase, we show that SprE changes the intracellular localization of PAP I. Taken together, we propose that SprE plays a multifunctional role in controlling the transcriptome, regulating what is made via its effects on RpoS, and modulating what is degraded via its effects on polyadenylation and turnover of specific mRNAs.
Natural microbial environments often require bacteria to rapidly adapt to a variety of stresses such as nutrient starvation. Escherichia coli responds to starvation by entering a protective, metabolically less active state, known as stationary phase (20). Cells in stationary phase can survive for weeks without essential nutrients and rapidly resume growth once environmental conditions improve and nutrients become available again. The transition from exponential growth phase to stationary phase is accompanied by multiple changes that affect all cellular processes. For example, there is an increase in proteolysis, accompanied by a decrease in overall transcription and translation (15, 20). Despite the global decrease in these processes, the expression of a subset of stress-responsive genes is induced as the cells enter stationary phase. These genes are controlled by the stationary phase sigma factor RpoS (σ38 and σS) (15, 50). The RpoS regulon consists of genes that specify proteins that combat multiple stresses and catalyze the synthesis of storage compounds, such as glycogen (15, 25).
Because of the drastic changes that occur upon entry into stationary phase, RpoS is one of the most highly regulated proteins in E. coli. It is controlled at all levels: transcription, translation, protein stability, and activity (15). During exponential growth, when nutrients are abundant, RpoS levels are kept low by the action of the adaptor protein SprE (RssB) and the ClpXP protease. After SprE binds to RpoS, the complex interacts with the ATP-dependent chaperone ClpX. RpoS is subsequently unfolded by ClpX and degraded by the ClpP protease, releasing SprE, which is recycled (19, 34, 39, 52). Upon entry into stationary phase, this process is inhibited, leading to the accumulation of RpoS and the transcription of its target genes (22, 51).
The N-terminal domain of SprE shares strong sequence homology with the receiver domain of the response regulator family of proteins (5). In a typical two domain response regulator, regulation of activity occurs through phosphorylation at a conserved aspartate (D58 in SprE) in the N-terminal domain. However, a mutant form of SprE, in which the conserved aspartate residue was changed to alanine, responds correctly to starvation for various essential nutrients, including carbon. Thus, phosphorylation does not regulate SprE activity in response to starvation (37). The C-terminal output domain of SprE shares no sequence homology with any known family of proteins, making it a novel response regulator.
It has recently been discovered that SprE is targeted by a set of anti-adaptor proteins, exemplified by IraP, that bind to SprE and inhibit its activity under certain stress conditions. Each anti-adaptor protein appears to be specific for a particular stress, but not all stresses necessarily elicit the production of an anti-adaptor protein. For example, no SprE-specific anti-adaptor protein has been identified during carbon starvation. If SprE indeed senses this particular type of starvation, the mechanism is not understood (7, 8, 12).
SprE levels are growth phase regulated. There are low levels of SprE during exponential growth, while entry into stationary phase increases protein levels by two- to threefold (41). SprE acts to modulate RpoS primarily during exponential phase; therefore, it is not obvious why SprE levels increase at a time when it no longer functions in this capacity. One possible explanation is that increasing SprE levels in stationary phase is necessary to prime the cells for the rapid destruction of RpoS when cells encounter nutrients and return to exponential growth (41). Another hypothesis is that SprE has a second function in stationary-phase cells that requires higher levels of the protein. Interestingly, it has been noted that overproduction of SprE causes a significant growth defect in exponentially growing cells (34, 43). The expected phenotype of the overproduction of SprE is lower RpoS levels, even in stationary phase. However, rpoS null mutants do not exhibit the growth defect seen with SprE overproduction; thus, SprE-mediated degradation of RpoS cannot account for this observed toxicity.
Here we have exploited the growth defect associated with SprE overproduction to uncover an additional role for this protein in polyadenylation and the control of mRNA stability. Specifically, in exponentially growing cells, the absence of SprE results in a significant reduction in poly(A) levels and a consequent increase in the half-lives of specific mRNAs, which have previously been shown to be dependent on polyadenylation for their decay. In stationary-phase cells lacking SprE, the intracellular location of poly(A) polymerase I (PAP I) is altered, suggesting a multifunctional role for SprE in the control of the transcriptome.
All strains are listed in Table Table1.1. Standard microbial techniques were used for strain constructions (44). Luria broth and M63 liquid medium and agar were prepared as described previously (44) and were supplemented with the appropriate antibiotics as needed. Antibiotic concentrations used were as follows: 125 μg/ml ampicillin (Ap), 25 μg/ml tetracycline, 20 μg/ml chloramphenicol, 50 μg/ml kanamycin, and 20 μg/ml spectinomycin. Bacteria were grown at 37°C with aeration, and growth was monitored by measuring the optical density at 600 nm (OD600). Plasmids were induced with 0.2% arabinose or 200 ng/ml anhydrotetracycline (ATC), where applicable.
For construction of pZS*11sprE+, colony PCR was performed on MC4100 cells with primers that amplify the sprE open reading frame (ORF). The primers 5sprEEcoRI (5′-AGAAACCGAATTCATTAAAGAGGAGAAAGGTACCGCATGACGCAGCCATTGGTCGG) and 3sprEXbaI (5′-GGCTCTAGACTCAGCTAATTAAGCTCATTCTGCAGACAACATCAAGCGC) were used for cloning. The forward primer contained a nonnative ribosome binding site upstream of the sprE start codon. After digestion with EcoRI and XbaI (sites underlined), the resulting PCR product was ligated into the EcoRI and XbaI sites of pZS*11 (23). This plasmid (three or four copies/cell) placed the sprE gene under the control of the tetracycline promoter (23). The tetracycline repressor was supplied in trans to repress sprE expression.
For the construction of pBADsprE+, colony PCR was performed using 5sprEEcoRI and 3sprEpBADHindIII (5′-TCCAAGCTTTGCTCATTCTGCAGACAACATCAAGCGC). After digestion with EcoRI and HindIII (site underlined), the PCR product was ligated into the EcoRI and HindIII sites on pBAD18. This plasmid (15 to 30 copies/cell) placed the sprE gene under the control of the arabinose-inducible promoter PBAD (13). The ppcnB-GFP plasmid was constructed as follows. The primers 5pcnBStuI (5′-TGGCGGAGGCCTCAGCGTCGAGCAAATCCTTCAG) and 3pcnBNheIns (5′-GGATCTGCTAGCTGCTGCTGCTGCGGTACCCTCACGACGTGGT) were used to amplify 140 bp upstream of pcnB, which contained the native pcnB promoters as well as the entire pcnB ORF. The cloned pcnB fragment replaced the stop codon with three alanine codons to serve as a linker to an in-frame translational fusion, with green fluorescent protein (GFP) at the C terminus. The resulting PCR product was digested and ligated into the StuI and NheI restriction sites on pCMW1, which was kindly provided by Chris Waters (49). This plasmid expressed the PAP I-GFP fusion protein under the control of the native pcnB promoters. The plasmids pBMK28 [pcnB-His6 Cmr] and pDK24 (rnb Apr) have been described previously (32, 33). All oligonucleotides were synthesized by Integrated DNA Technologies. Each construct was verified by DNA sequence analysis by Genewiz, Inc. (South Plainfield, NJ).
Random PCR mutagenesis was performed using the GeneMorph random mutagenesis kit (Stratagene), as per manufacturer's instructions. Briefly, 500 ng of the plasmid pZS*11sprE+ was subjected to 30 rounds of mutagenic PCR in order to achieve a mutation rate of 0 to 3 mutations/kb. The primers 5sprEEcoRI and 3sprEXbaI were used for the PCR, so that we could clone the mutagenized sprE products as described above. The pool of mutagenized plasmids was transformed into VC30, which contained the rpoS750′-′lacZ reporter fusion. This fusion is subject to the same regulation as full-length RpoS (45). Colonies were screened for color on LB agar containing 125 μg/ml Ap plus 80 μg/ml X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) after incubation overnight at 37°C. Colonies that appeared to have a different color than that of the parent (VC30), indicating changes in RpoS regulation, were analyzed further. Plasmids were miniprepped by following the manufacturer's instructions (Qiagen), and candidates were sequenced by Genewiz, Inc.
Strains containing pBADsprE+ were streaked on LB agar containing 125 μg/ml Ap plus 0.2% arabinose. Faster growing colonies were selected from the slow-growing parental lawn and were purified. Plasmid DNA was miniprepped as described above and was retransformed into the parent strain to check for plasmid linkage (44). All plasmid-linked mutations were discarded. Chromosomal mutations were mapped by using a library of random transposon mutations, according to standard procedures (44).
Growth was monitored every 20 min for 3 to 4 h by measuring the OD600. Doubling times were calculated using data points obtained during mid-exponential growth. At least three independent experiments were performed for each strain analyzed.
Plasmids were mutagenized using the GeneTailor site-directed mutagenesis system (Invitrogen), according to the manufacturer's instructions. Plasmid DNA was methylated at cytosine residues prior to PCR. The D58A mutation was engineered in trans into the plasmid pZS*11sprE60 (see Results). The primers sprED58Afor (5′-CTCCAGACCTGATGATATGTGCTATCGCGATG) and sprED58Arev (5′-ACATATCATCAGGTCTGGAGTGAAACCTCC) were used to introduce a GAT→GCT transversion (underlined). Mutagenized plasmids were transformed into cells that contained the McrBC endonuclease to degrade the methylated, nonmutated parental strands. Plasmids were miniprepped as described above and were sequenced by Genewiz, Inc.
Cells were grown to an OD600 of 0.1, followed by induction of plasmid expression for 1 h, where applicable. At mid-exponential phase (OD600 of 0.4 to 0.5), the cells were pelleted and were immediately resuspended with sodium dodecyl sulfate (SDS) sample buffer. Samples were boiled for 10 min, and then equal volumes were loaded onto a 12% polyacrylamide gel (44). Following electrophoresis, proteins were transferred to a Protran nitrocellulose membrane (Whatman) and were probed with either a 1:6,000 dilution of RpoS antibody or a 1:4,000 dilution of SprE antibody (laboratory stocks). Donkey anti-rabbit immunoglobulin G-horseradish peroxidase conjugate (GE Healthcare) was used as a secondary antibody at a dilution of 1:6,000. Bands were visualized using the ECL antibody detection kit (GE Healthcare) and HyBlot CL film (Denville Scientific Inc.). Quantification of bands was carried out using the ImageJ program (1).
One milliliter of overnight cultures or mid-exponential-phase cultures was pelleted and washed twice with 1 ml of M63 liquid medium. Overnight cultures were resuspended in 1 ml of M63, while mid-exponential-phase cultures were resuspended in 200 μl M63 in order to concentrate the cells. From the resuspended cells, 5 μl was spotted onto a 1% agarose pad (in M63). Images were obtained employing a QImaging Rolera-XR camera on a Nikon 90i microscope equipped with a Nikon PlanApo 1.4/100× Oil Ph3 phase objective, using the NIS-Elements software package. At least two independent experiments were performed per strain, and at least 150 cells were counted for determination of localization percentages.
The total RNA from exponentially growing cultures was isolated using the Catrimide method, as described previously (26). All cultures were grown to Klett 50 (no. 42 green filter) in LB at 37°C with shaking. Total RNA from the stationary-phase cultures was isolated after 16 to 18 h of growth in LB at 37°C with shaking using an SDS lysis method, as described previously (26). The strains containing pZS*11sprE60 were grown in the presence of 200 ng/ml of ATC. The RNA-DNA dot blots used to determine total intracellular poly(A) levels were carried out as previously described (26). Total RNA for half-life determinations was isolated at various time points after addition of rifampin (rifampicin) to stop the transcription initiation, as described previously (26). The half-lives of full-length transcripts were determined by Northern analysis by separating the transcripts either by 7% acrylamide/8 M urea or by 1.5% agarose-glyoxal gels (26). Half-lives were determined using linear regression analysis.
It has previously been reported that overexpression of SprE causes significant growth defects (34, 43). To determine the levels of SprE required to cause toxicity, we initially examined a strain carrying an rssA2::cam chromosomal allele, which contains a transposon inserted upstream of the sprE gene, such that expression is driven by the constitutive chloramphenicol promoter (41). This allele resulted in a fourfold increase in SprE levels relative to those of the wild type (Fig. (Fig.1A,1A, lanes 1 and 3; Table Table2),2), with no observable effect on growth rate (Table (Table2).2). Next, the sprE ORF was cloned under the control of the ATC-inducible PLtetO-1 promoter into a low-copy-number (three or four copies/cell) vector, pZS*11. Induction from this vector led to a fivefold increase above wild-type SprE levels (Fig. (Fig.1A,1A, lane 4; Table Table2).2). Strains containing pZS*11sprE+ had no apparent growth defect on solid medium containing ATC. However, this plasmid did confer a slight growth defect in liquid medium under inducing conditions (40.6 ± 4.5 min of doubling time [Table [Table2])2]) compared to growth of the parent strain carrying the vector control (33.8 ± 4.3 min of doubling time [Table [Table2]).2]). To further increase SprE levels, the pBADsprE+ plasmid was constructed, which led to a >20-fold overproduction of SprE when induced (Fig. (Fig.1A,1A, lane 5). Overproduction of SprE at this level caused a pronounced toxicity, as evidenced by poor growth and the appearance of multiple suppressors growing out of the parental lawn on solid medium. The generation time of the strain harboring pBADsprE+ was >60 min (data not shown).
The overproduction of a given protein can be toxic to cells due to a particular protein's activity or the formation of toxic aggregates. To help distinguish these two possibilities, we mutagenized pZS*11sprE+, as described in Materials and Methods, since this plasmid did not confer a significant growth defect (Table (Table2).2). We reasoned that if toxicity resulted from an activity of SprE, we could isolate mutants that showed enhanced activity.
We identified two mutants that conferred toxicity without changing the intracellular level of the SprE protein (Fig. (Fig.1B,1B, compare lanes 1 and 2). Sequencing of these mutants revealed two different alleles, both of which carried multiple mutations, as follows: sprE(L29F, A97T) and sprE(L29F, V120I, L124F). Reconstruction experiments using site-directed mutagenesis demonstrated that each of the multiple mutations was required for the mutant phenotype (data not shown). Since both alleles were phenotypically indistinguishable, sprE(L29F, V120I, L124F), referred to as sprE60 for simplicity, was used for all studies described below.
The SprE60 protein caused a growth defect similar to that seen with pBADsprE+ (61.8 ± 8.6 min [Table [Table2]).2]). However, in either case, it should be noted that this growth rate may actually be an underrepresentation of the true growth rate of these mutants. Occasionally, mutant cultures had doubling times similar to those of the wild-type control, indicating that a faster-growing suppressor took over the culture at an early time point, particularly since a doubling time in excess of 100 min was also observed once (data not shown).
Expression of sprE from the pBAD plasmid resulted in the production of ~20 times greater SprE levels than those of the wild type and a generation time of >60 min. Expression of the sprE60 allele from the PLtetO-1 promoter on pZS*11 was much lower, yielding protein levels five to six times greater than those of the wild type. Nevertheless, expression of the two plasmid constructs resulted in comparably poor growth with similar generation times. These results suggested that the point mutations associated with the sprE60 allele may have increased the inherent toxicity associated with high-level overproduction of wild-type SprE. Although not conclusive, these results supported the possibility that it is an activity of SprE that causes toxicity.
SprE plays an important role in controlling the intracellular level of the RpoS sigma factor. However, strains lacking rpoS, sprE, or clpXP, all players in the regulation of RpoS protein levels, are all viable, and they exhibit no apparent growth defects. In order to determine if any member of the RpoS pathway was involved with the toxicity associated with the overproduction of SprE, chromosomal null alleles of rpoS, sprE, and clpXP were introduced into the strains carrying either pZS*11sprE60 or pBADsprE+. Upon induction of expression of either of the sprE alleles, none of these null alleles relieved the toxicity associated with either of the plasmids (data not shown). These results indicate that toxicity does not depend upon RpoS levels, per se, as the rpoS and clpXP mutations affect the levels of RpoS in a completely opposite fashion. These results also demonstrate that the toxicity we observed was not related to SprE's known function in the regulated proteolysis of RpoS.
As noted above, the N-terminal domain of SprE shares homology with the response regulator family of proteins and contains a conserved aspartate that serves as the phosphoacceptor site (6). Surprisingly, mutation of this site, D58A, resulted in an altered SprE protein that is still functional (37), contrary to most other response regulators (16). In order to determine if SprE phosphorylation was associated with toxicity, we cloned the sprED58A allele into pBAD18. Overproduction of the nonphosphorylatable SprE protein still caused toxicity (data not shown), suggesting that overproduction of the wild-type SprE protein does not lead to a titration of phosphate from an essential kinase(s). The D58A mutation was also engineered by site-directed mutagenesis into pZS*11sprE60 to create pZS*11sprE60D58A. Under these circumstances, there was a partial restoration of the doubling time, from 61.8 ± 8.6 min to 48.0 ± 5.0 min (Table (Table2).2). We conclude that phosphorylation is not required for the toxicity phenotype per se, but it may enhance it.
In order to gain insight into the cause of the toxicity, we isolated and characterized suppressors of the slow-growth phenotype. Specifically, strains carrying pBADsprE+ were plated on LB agar containing 0.2% arabinose, and faster-growing colonies that grew above the background lawn were purified. Plasmid DNA was isolated from the faster-growing variants to check for plasmid-linked mutations, which were discarded. Standard genetic techniques were used to map one of the chromosomal suppressors, as described in Materials and Methods. DNA sequence analysis showed that this suppressor was a null allele of pcnB, which encodes PAP I, the enzyme that adds poly(A) tails onto the 3′ ends of full-length mRNAs and decay products, facilitating their degradation (21, 27, 36).
PAP I is also involved in the copy number control of ColE1-based plasmids through the polyadenylation of the antisense RNA (RNA I), which is involved in replication control (14). In the absence of PAP I, plasmid copy number is significantly decreased, possibly explaining why this mutation suppressed the toxicity of pBADsprE+, a plasmid containing a ColE1 origin of DNA replication (13). If a decrease in plasmid copy number were the primary reason for the suppression of the slow-growth phenotype, then inactivation of PAP I would not be expected to alter the toxicity caused by pZS*11sprE60, since this plasmid contains the pSC101 origin, which is not affected by polyadenylation (24). However, a ΔpcnB allele also suppressed the toxicity associated with pZS*11sprE60 (Table (Table2).2). In fact, the presence of the ΔpcnB allele led to a doubling time that was only slightly longer than that of the wild-type control (Table (Table2).2). Moreover, the absence of PAP I did not significantly affect the levels of the SprE60 protein (Fig. (Fig.1B,1B, compare lanes 2 and 3; Table Table2).2). These results suggest that inactivation of PAP I does not relieve the toxicity by lowering gene dosage or the levels of the sprE mRNA. In fact, in a ΔpcnB strain, there was a slight increase in SprE levels (Table (Table2).2). The absence of PAP I also suppressed the slight growth defect associated with the presence of pZS*11sprE+, restoring the growth rate to 32.6 ± 0.9 min, which was equivalent to that of the strain carrying the empty vector control pZS*11 (Table (Table2).2). These results provide evidence that the toxicity associated with SprE60 and the overproduction of wild-type SprE are indeed the same.
PAP I requires the RNA chaperone Hfq to polyadenylate mRNAs containing 3′ Rho-independent transcription terminators (33). Since PAP I forms a complex with Hfq (33), we hypothesized that an Δhfq null allele might also suppress the toxicity phenotype. A strain carrying pZS*11sprE60 and a Δhfq allele had a doubling time of 38.6 ± 2.8 min, which was very similar to that of the parent containing pZS*11sprE+ and was a result comparable to that observed with a ΔpcnB allele (Table (Table2).2). Furthermore, inactivation of Hfq did not affect SprE60 protein levels (Fig. (Fig.1B,1B, compare lanes 2 and 5). Interestingly, inactivation of Hfq in an otherwise wild-type genetic background led to a twofold increase in SprE levels (Table (Table2).2). We conclude that the toxicity phenotype caused by high levels of SprE or by SprE60 requires the presence of both PAP I and Hfq.
Since polyadenylation has been shown to affect the decay of a significant number of E. coli mRNAs (21, 28, 33, 36), we tested whether any of the ribonucleases involved in mRNA decay also interacted with SprE. Accordingly, the plasmid pZS*11sprE60 was transformed into strains carrying one of the following loss-of-function mutations: rne-1 (RNase E), rncΔ38 (RNase III), rnb-296 (RNase II), pnpΔ683 (polynucleotide phosphorylase [PNPase]), or rnr::kan (RNase R). Transformants were easily obtained in the rne-1, rncΔ38, rnb-296, and rnr::kan strains, and the phenotypic properties of these transformants were comparable to those observed for a wild-type control carrying pZS*11sprE60 (data not shown).
In contrast, very slow-growing transformants were obtained with the pnpΔ683 strain. These transformants were extremely unstable, throwing off faster-growing variants, many of which had lost the plasmid. It is thus likely that a pnpΔ683 pZS*11sprE60 strain is a synthetic lethal combination. Indeed, we were unable to construct this double mutant by generalized transduction. Since it is known that poly(A) levels increase significantly in a pnp null mutant (33), we hypothesized that SprE60 might also cause an increase in poly(A) levels such that a synergistic interaction occurred in the pnpΔ683 pZS*11sprE60 strain, leading to cell death. The idea that SprE60 (or high levels of wild-type SprE) increases poly(A) levels also explains the suppression of the toxicity phenotype by pcnB and hfq null alleles, since these alleles lead to reduced polyadenylation (33).
It has also been observed that PAP I overproduction leads to increased levels of poly(A) and cell death over time (27). This cell death is suppressed by co-overproduction of RNase II, an exonuclease known to degrade poly(A) tails (32). If SprE60 toxicity were also the result of increased poly(A) levels, increasing levels of RNase II might suppress the toxicity. Accordingly, we transformed a compatible plasmid containing rnb under the control of its native promoter into strains carrying pZS*11sprE+ and pZS*11sprE60. Overproduction of RNase II in combination with pZS*11sprE+ resulted in a doubling time of 47.9 ± 4.6 min, and in combination with pZS*11sprE60, the doubling time was nearly identical at 47.7 ± 3.8 min (Table (Table2).2). We conclude that the toxicity caused by SprE60 was at least partially suppressed by increasing the levels of RNase II.
Since inactivation of PAP I suppressed the growth defect associated with overproduction of SprE60, we hypothesized that SprE played some role in the regulation of intracellular levels of poly(A). To test this hypothesis directly, we measured poly(A) levels in exponentially growing cells using RNA-DNA dot blots. As seen in Fig. Fig.2,2, the presence of pZS*11sprE60 led to a twofold increase in poly(A) levels compared to those of the wild-type control. When the same experiment was carried out in either a ΔpcnB pZS*11sprE60 or hfq-1 pZS*11sprE60 strain, poly(A) levels were significantly lower than those observed in the wild-type control. These data correlated with the shorter generation times observed in these strains, relative to the sprE+ pZS*11sprE60 strain (Table (Table22).
Since overproduction of the SprE60 protein led to a significant increase in total poly(A) levels (Fig. (Fig.2),2), we tested to determine if wild-type SprE was involved in the regulation of intracellular levels of polyadenylation. In agreement with previously reported results in the MG1655 genetic background (27, 33), poly(A) levels were reduced 70% and 60% in ΔpcnB and hfq-1 strains, respectively (Fig. (Fig.2),2), while deletion of the sprE locus caused a 40 to 50% reduction in poly(A) levels in exponentially growing cells (Fig. (Fig.2).2). In contrast, poly(A) levels were unchanged in a clpX null strain (Fig. (Fig.2),2), which suggests that the RpoS degradation function of SprE is separate and distinct from the polyadenylation phenotype.
The data presented above shows that deletion of the SprE protein leads to a significant reduction in the intracellular poly(A) levels, suggesting a role of SprE in the regulation of the polyadenylation pathway. Since it has been shown that the half-lives of specific mRNAs are increased in the absence of polyadenylation (21, 33, 36), we predicted that for those mRNAs whose decay is stimulated by polyadenylation, we would see comparable half-lives in ΔpcnB and ΔsprE mutants. Accordingly, we used Northern blot analysis to determine the half-lives of six specific mRNAs. As shown in Table Table3,3, the half-lives of five out of six of the tested mRNAs were increased in both the ΔpcnB and ΔsprE strains. The single exception, rplY, had identical half-lives in all genetic backgrounds. This result was consistent with earlier observations that the half-life of rplY is not affected by the level of polyadenylation (29). Further support for the direct role of SprE in the polyadenylation and degradation of mRNAs was derived from the decay patterns observed for at least one of the six transcripts. As shown in Fig. Fig.3,3, specific breakdown products (arrows) of the yfiA mRNA transcript accumulated in both the ΔpcnB and ΔsprE strains but were barely detectable in the wild-type control. In fact, relative to the amount of full-length yfiA mRNA in this analysis, the amount of the decay intermediate (Fig. (Fig.3,3, upper arrow) increased from <0.1% at 0 min up to ~10% at 32 min in the wild-type strain. In contrast, it increased from 0.2% to 18% and 0.2% to 36% in the ΔpcnB and ΔsprE strains, respectively. These results support a role for SprE in mRNA stability.
It has been shown that intracellular levels of polyadenylation can be affected by changing the level of PAP I (27), altering the levels of either RNase II or PNPase (32), or inactivating Hfq, a component of the polyadenylation complex (33). Accordingly, we used Western blot analysis to measure these protein levels in wild-type and ΔsprE strains. PAP I levels were identical, within experimental error, in both strains (Fig. (Fig.4A).4A). Similar results were obtained with PNPase (data not shown). In contrast, there was a small but consistent 30 to 40% increase in the level of RNase II (Fig. (Fig.4B).4B). Clearly the decreased polyadenylation in strains lacking SprE is not due to decreases in the levels of either PAP I or PNPase. Although it cannot be ruled out that the decreased polyadenylation results from the small increase in RNase II observed in the absence of SprE, we think this is unlikely.
During the course of this work, we noticed that a strain carrying the rssA2 allele exhibited an extended lag phase when exiting stationary phase. Specifically, after diluting an overnight stationary-phase culture of wild-type E. coli by 100-fold into fresh medium, there was an approximate 20-min delay before growth resumed. In contrast, rssA2 strains did not resume growth until 40 min after dilution (Fig. (Fig.5).5). In both the wild-type and rssA2 strains, the numbers of CFU were identical at time zero, indicating that a decreased number of viable cells was not the underlying cause of the delay. Since deleting either rpoS or sprE had no effect on the lag time compared to that of a wild-type control (data not shown), we hypothesized that the defect in the exit from stationary phase might be related to altered polyadenylation in stationary-phase cells, resulting from increased levels of the SprE protein. If this were the case, a ΔpcnB allele would be expected to suppress this defect. As shown in Fig. Fig.5,5, this was in fact the case.
If changes in polyadenylation resulted in a defect in the exit from stationary phase, then overproduction of PAP I from a plasmid should produce a similar phenotype. When pBMK28, a plasmid carrying a pcnB-His6 allele under the control of an IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible promoter (33), was introduced into a wild-type strain, even the uninduced transformants showed a 40-min delay in the exit from stationary phase, identical to what was observed with the rssA2 mutant (Fig. (Fig.5).5). Strikingly, in the absence of SprE, strains harboring pBMK28 did not show this defect (Fig. (Fig.5).5). Note that the uninduced levels of PAP I-His6 produced under these conditions are not high enough to cause the lethality associated with high-level overproduction of PAP I (27). We further showed that in a strain that constitutively produces LacIq (23), the exit defect associated with pBMK28 was also suppressed (data not shown). Overproduction of both PAP I and SprE simultaneously resulted in additive effects, where growth had not resumed, even after 60-min postdilution (Fig. (Fig.5).5). Results presented here suggest that proper control of polyadenylation is important for the exit from stationary phase.
It has been previously reported that the first 24 amino acids of PAP I are sufficient for its localization to the inner membrane (17). We hypothesized that SprE might affect intracellular levels of polyadenylation by altering the cellular localization of PAP I. To test this directly, we constructed a PAP I-GFP fusion protein. The fusion protein was able to complement the plasmid-copy-number defect associated with ΔpcnB strains (data not shown) and resulted in a sixfold increase in levels of polyadenylation (data not shown), demonstrating that the PAP I-GFP protein is functional.
In a wild-type background during exponential growth, PAP I-GFP localized to the inner membrane in 96.8% of the cells examined (Fig. (Fig.6A),6A), as demonstrated by a ring of fluorescence at the cell periphery. It should be noted that in many cases, in addition to rings, there appeared to be localized foci of PAP I-GFP at the membrane. During stationary phase, PAP I-GFP was no longer membrane localized but formed distinct foci in the cytoplasm in 94.4% of cells (Fig. (Fig.6B).6B). The spatial differences in PAP I localization may be important for regulating PAP I activity. In sprE, rpoS, and rssA2 strains, the PAP I-GFP localization in exponential phase was similar to that observed in the wild type; around 95% cells showed membrane localization (Fig. (Fig.6A).6A). The rssA2 and rpoS strains appeared to be similar to the wild type in stationary phase as well. However, PAP I-GFP was still localized to the membrane in 64.9% of cells in the sprE null strain in stationary phase (Fig. (Fig.6B).6B). This suggests that under these conditions, SprE is required for efficient release of PAP I into the cytoplasm. Thus, SprE affects not only the activity of PAP I in exponentially growing cells but also its cellular location in stationary-phase cells.
It has been reported that the overproduction of wild-type SprE is toxic to cells (34, 43). We have exploited the growth defect caused by SprE overproduction to uncover a novel function of SprE. Our data indicate that in exponentially growing cells, SprE stimulates polyadenylation, thereby affecting the decay of specific mRNA species. Indeed, the SprE60 mutant protein increases levels of polyadenylation by at least twofold, and poly(A) levels are reduced by twofold in the absence of SprE (Fig. (Fig.2).2). Thus, SprE has two distinct functions, both of which affect the transcriptome. SprE controls what is made by regulating the stability of RpoS, and it also controls which mRNAs are destroyed by stimulating polyadenylation.
While the exact changes in the polyadenylation profile of RNA substrates in the presence of SprE60, or in the absence of SprE, are not yet known, we assume that these changes reflect changes in the activity of PAP I. In support of this idea, removal of the SprE protein led to changes in mRNA half-lives that were almost identical to that obtained upon inactivation of PAP I (Table (Table3).3). However, not all of the decay intermediates observed in the absence of PAP I were present in the SprE mutant (Fig. (Fig.3).3). For example, out of the two yfiA decay intermediates that accumulated in the ΔpcnB strain, only the larger species was stable in the ΔsprE mutant (Fig. (Fig.3).3). This could indicate that the effect of SprE on mRNA decay is dependent on PAP I-mediated polyadenylation as well as some other PAP I-independent mechanism. Furthermore, we cannot rule out that the strong effect of both sprE and pcnB mutations on the ompA mRNA half-life (Table (Table3)3) is due to a combination of effects related to other factors such as growth rate, Hfq modulation, and the action of small RNA regulators that affect RNase E cleavage of this transcript (35, 40, 47, 48). Nonetheless, our data strongly suggest that SprE and PAP I function together during exponential phase to regulate polyadenylation and mRNA stability.
Interestingly, connections between SprE and PAP I have been previously reported. Sarkar et al. (43) identified SprE as a multicopy suppressor of the pcnB plasmid-copy-number defect. PAP I controls the levels of an antisense RNA (RNA I) that inhibits an RNA replication primer of ColE1-type plasmids (14). In the absence of pcnB, RNA I is stabilized, which prevents plasmid replication. Overproduction of SprE suppresses this defect. Given our results, it seems likely that the overproduction of SprE suppressed the pcnB defect either by stimulating PAP I-independent polyadenylation or by stimulating the degradation of the nonpolyadenylated RNA I species. One possibility is that SprE stimulates the other known polyadenylation enzyme PNPase, which accounts for the residual polyadenylation in a pcnB null strain in exponential-phase cells (31). While it is true that we did not observe a significant increase in PNPase-dependent polyadenylation by SprE60 in the ΔpcnB background (Fig. (Fig.2),2), this result may simply be due to the fact that the heteropolymeric tails generated by PNPase (32, 33) cannot be efficiently detected with RNA-DNA dot blots using the oligo(dT)20 probes we employed.
Santos et al. (42) noted that in the absence of PAP I, RpoS levels were reduced by threefold in stationary-phase cells. They also demonstrated that the effect seen was posttranscriptional and specifically affected protein stability. They hypothesized that PAP I controls the activity of SprE, and that is the cause for reduced RpoS levels. Taking into account our findings, it is plausible that the reduction of RpoS levels the authors reported was due to the lack of SprE's second function. Perhaps when there is no PAP I, higher levels of SprE are available to promote RpoS degradation. Indeed, the rssA2::cam allele, which increases SprE levels by fourfold, decreases RpoS levels in all growth phases (39). What the authors perceived as PAP I regulation of SprE may actually be a titration of SprE into the polyadenylation pathway.
PAP I and SprE clearly function together during exponential phase, and we believe this is true during stationary phase as well. Polyadenylation and mRNA stability in stationary phase have not been extensively studied and remain poorly understood. However, we believe that SprE must play some role in regulating polyadenylation in stationary phase, since the localization of PAP I changes in a SprE-dependent manner when growth ceases. Moreover, this localization defect is associated with clear phenotypic changes.
In an otherwise wild-type background, PAP I-GFP was localized to the inner membrane during exponential phase and was released from the membrane as cells entered stationary phase (Fig. (Fig.6).6). This release of PAP I-GFP from the membrane was SprE dependent; in cells lacking SprE, PAP I-GFP remained membrane associated. These SprE-dependent changes in the localization of PAP I-GFP provide evidence for a functional connection between PAP I and SprE during stationary phase. It should be noted that the hybrid gene producing PAP I-GFP was carried on a multicopy plasmid, and thus, the fusion protein is overproduced compared to the wild-type endogenous PAP I. This overproduction facilitated visualization of the fluorescent molecule. We assume, but have not proven, that this change in PAP I-GFP localization reflects what happens with endogenous PAP I. If this is true, then it seems likely that during exponential phase, efficient polyadenylation takes place at the inner membrane. It is also possible that the entire mRNA degradation process is compartmentalized. Indeed, it has been recently reported that the mRNA degradosome is anchored to the inner membrane by the N-terminal portion of RNase E (18).
It is important to note that the cells overproducing PAP I-GFP (Fig. (Fig.6)6) exhibit a defect in the exit from stationary phase that is similar to that caused by overproducing PAP I-His6 (Fig. (Fig.5).5). We believe that this exit defect is caused by the elevated activity of PAP I, because removal of SprE suppresses this exit defect. We have shown that removal of SprE decreases PAP I activity in exponential-phase cells (Fig. (Fig.2),2), and we assume its removal has similar effects on PAP I-GFP and PAP I-His6 in stationary phase as well. Strikingly, in reciprocal fashion, increased production of SprE also causes a defect in the exit from stationary phase, and removal of PAP I suppresses this defect. We assume that overproduction of SprE in stationary-phase cells also increases PAP I activity. Indeed, SprE60 increases polyadenylation in exponential-phase cells (Fig. (Fig.2).2). However, we do not know if the elevated activity of PAP I that occurs in stationary-phase cells under conditions of either PAP I or SprE overproduction is due to changes in PAP I localization, activity, or both. Nonetheless, this reciprocal suppression provides additional evidence for a functional connection between SprE and PAP I.
As noted above, little is known about polyadenylation and mRNA stability in stationary-phase cells, and we do not currently understand why increasing PAP I activity, by increasing production of either PAP I or SprE, would cause a stationary-phase exit defect. It seems likely that this exit defect results from alterations in stationary-phase polyadenylation. It is possible that altered polyadenylation results in the depletion of an essential mRNA or noncoding RNA, or that the tRNA pool is damaged and therefore depleted. In wild-type cells, a reduction of PAP I activity during stationary phase may help change the polyadenylation profile of mRNAs and therefore greatly change which ones are degraded. The RNAs that are degraded during exponential phase may be quite different from those that are degraded during stationary phase. In fact, the stationary-phase-specific mRNAs generally contain polynucleotide tails (9), which are most probably generated by PNPase (32), and the tails generated by PNPase are poor substitutes for poly(A) tails (33).
Upon entry into stationary phase, SprE levels are increased by two- to threefold (41). It is somewhat paradoxical that SprE levels would increase when the protein is inactive with regard to RpoS degradation. One proposed explanation is that increasing SprE levels in stationary phase is necessary to promote the rapid destruction of RpoS when cells encounter nutrients and return to exponential phase. This explanation seems unlikely because cells that only mildly overproduced SprE exhibited a defect in the exit from stationary phase. Alternatively, we suggest that SprE levels increase because the protein has an additional function in stationary phase, the regulation of polyadenylation. While we currently have no evidence for a direct interaction between SprE and PAP I, SprE clearly plays a role in the PAP I-pathway in exponential phase, and we have documented numerous genetic interactions between sprE and pcnB that affect both exponential- and stationary-phase phenotypes. Further studies are required to elucidate the exact mechanism by which SprE contributes to the polyadenylation pathway.
This work was supported by grants from the National Institute of General Medical Sciences (grant GM65216) to T.J.S. and (grant GM57220) to S.R.K.
Published ahead of print on 18 September 2009.