Direct Inhibition of RetS Synthesis by RsmA Contributes to Homeostasis of the Pseudomonas aeruginosa Gac/Rsm Signaling System

ABSTRACT

The Gac/Rsm system is a global regulator of Pseudomonas aeruginosa gene expression. The primary effectors are RsmA and RsmF. Both are RNA-binding proteins that interact with target mRNAs to modulate protein synthesis. RsmA/RsmF recognize GGA sequences presented in the loop portion of stem-loop structures. For repressed targets, the GGA sites usually overlap the ribosome binding site (RBS) and RsmA/RsmF binding inhibits translation initiation. RsmA/RsmF activity is controlled by several small non-coding RNAs (sRNA) that sequester RsmA/RsmF from target mRNAs. The most important sequestering sRNAs are RsmY and RsmZ. Transcription of rsmY/rsmZ is directly controlled by the GacSA two-component regulatory system. GacSA activity is antagonized by RetS, a hybrid sensor kinase. In the absence of retS, rsmY/rsmZ transcription is derepressed and RsmA/RsmF are sequestered by RsmY/RsmZ. Gac/Rsm system homeostasis is tightly controlled by at least two mechanisms. First, direct binding of RsmA to the rsmA and rsmF mRNAs inhibits further synthesis of both proteins. Second, RsmA stimulates rsmY/rsmZ transcription through an undefined mechanism. In this study we demonstrate that RsmA stimulates rsmY/rsmZ transcription by directly inhibiting RetS synthesis. RetS protein levels are elevated 2.5-fold in an rsmA mutant. Epistasis experiments demonstrate that the rsmA requirement for rsmY/rsmZ transcription is entirely suppressed in an rsmA, retS double mutant. RsmA directly interacts with the retS mRNA and requires two distinct GGA sites, one of which overlaps the RBS. We propose a model wherein RsmA inhibits RetS synthesis to promote rsmY/rsmZ transcription and that this acts as a checkpoint to limit RsmA/RsmF availability.

IMPORTANCE The Pseudomonas aeruginosa Gac/Rsm system controls ∼500 genes and governs a critical lifestyle switch by inversely regulating factors that favor acute or chronic colonization. Control of gene expression by the Gac/Rsm system is mediated through RsmA and RsmF, small RNA-binding proteins that interact with target mRNAs to inhibit or promote protein synthesis and/or mRNA stability. RsmA/RsmF activity is governed by two small non-coding RNAs (RsmY and RsmZ) that sequester RsmA/RsmF from target mRNAs. The GacSA two-component regulatory system plays a pivotal role in the Gac/Rsm system by controlling rsmYZ transcription. This study provides insight into the control of homeostasis by demonstrating that RsmA directly targets and inhibits expression of RetS, an orphan sensor kinase critical for rsmYZ transcription.

INTRODUCTION

The Gram-negative bacterium Pseudomonas aeruginosa is an opportunistic pathogen frequently isolated from ventilated, immunocompromised, and cystic fibrosis (CF) patient populations (1). Ventilator-associated pneumonia and sepsis are considered acute infections and associated with high morbidity and mortality rates (1, 2). Acute P. aeruginosa infections are dependent upon the expression of a large set of virulence factors that incudes surface structures such as flagella, type IV pili, type II, and type III secretion systems, and secreted factors such as exotoxin A, pyocyanin, and proteases (3). While most P. aeruginosa infections are acute, one notable exception is the chronic airway infections common in CF. Individuals with CF are hypersusceptible to P. aeruginosa airway infections and have an infection duration that can last for decades with bacterial burdens reaching10⁹ CFU/ml sputum (4, 5). P. aeruginosa infections in CF are initiated by exposure to an environmental source of the organism. Adaptation to the host is critical for the successful colonization and persistence of P. aeruginosa. Host-associated signals drive adaptation through changes in gene expression. Those changes can result in more efficient utilization of the available energy and nutrient sources, host evasion, and persistence. Common P. aeruginosa adaptations to the CF environment include auxotrophy, polysaccharide production, biofilm growth, enhanced resistance properties, and reduced expression of some virulence factors that promote acute infection phenotypes (5–9).

Numerous regulatory pathways coordinate adoption of a chronic lifestyle. For example, the mucoid phenotype is common in chronic CF isolates (8). Mucoidy results from overproduction of alginate, a polysaccharide that promotes resistance to antimicrobials and host clearance and contributes to biofilm formation. (10, 11). Enhanced biofilm formation is a phenotype common in CF isolates. Although many regulatory systems control biofilm production, the signaling molecule cyclic-di-GMP (c-di-GMP) and the Gac/Rsm system play prominent roles (12–14). c-di-GMP is a nucleotide-based second messenger that controls cellular responses through direct interactions with riboswitches, transcription factors, and enzymes (15, 16). c-di-GMP is synthesized by diguanylate cyclases and degraded by phosphodiesterases (17). Several diguanylate cyclases are activated during biofilm growth where the concomitant increase in c-di-GMP promotes synthesis of the Pel and Psl polysaccharides, both components of the biofilm matrix (18, 19). The Gac/Rsm system controls gene expression at a post-transcriptional level through the action of two RNA binding proteins, RsmA and RsmF (also called RsmN) (20–22). RsmA/RsmF bind to a CANGGAYG consensus sequence (with the underlined GGA site being the most highly conserved) on target mRNAs and can have either positive or negative effects on gene expression (20, 21, 23). The GGA sequence often overlaps the ribosome binding site and is usually presented in the loop portion of a stem-loop structure (20, 24). RsmA/RsmF activity is controlled by several non-coding RNAs including RsmV, W, Y, and Z (25–28). Each of the RNAs has multiple RsmA/RsmF binding sites that function by sequestering RsmA/RsmF from target mRNAs. RsmA is the more dominant regulator and generally has a positive effect on genes associated with acute virulence, and an inhibitory effect on factors required for chronic colonization (29, 30). Thus, reduced RsmA availability promotes chronic virulence phenotypes, and this is accomplished through increased expression of the non-coding RNAs that sequester RsmA from target RNAs.

Regulation of RsmA activity by RsmY and RsmZ is understood best and is controlled by the GacSA two-component regulatory system (TCS) (25, 29). In response to poorly defined environmental signals, the GacA response regulator is phosphorylated by the GacS sensor kinase and activated GacA then binds to the P_rsmY and P_rsmZ promoter regions to stimulate transcription (25, 29, 31, 32). Two additional orphan hybrid sensor kinases (LadS and RetS) modulate GacS activity. LadS activates GacS in response to high environmental calcium levels, while RetS inhibits GacS phosphorylation (33–37). Deletion of gacA or gacS phenocopies a rsmYZ deletion mutant, resulting in an overabundance of RsmA (26, 34). Conversely, deletion of retS results in enhanced rsmYZ expression and a drastic reduction in RsmA availability (38).

In addition to environmental control of rsmYZ transcription, homeostatic mechanisms appear to maintain a defined window of RsmA and RsmF availability (25, 29, 39). Although the upper and lower boundaries of that window remain to be ascertained, both mechanisms are responsive to RsmA availability. First, RsmA directly binds to the leader region of the rsmA and rsmF mRNAs to inhibit translation (21). Because the relative binding affinity of RsmA for the rsmA and rsmF leader regions is significantly weaker than observed for RsmY and RsmZ (21), inhibition of RsmA and RsmF translation likely occurs at the upper boundary for RsmA availability. The second homeostasis mechanism involves RsmA-dependent stimulation of rsmY and rsmZ transcription through a mechanism that remains to be defined (25, 29, 39). In the present study we demonstrate that RsmA control of rsmYZ transcription occurs through inhibition of RetS synthesis.

RESULTS

RetS protein levels are elevated in an rsmA deletion mutant.

Previous studies established that rsmY and rsmZ transcription is directly activated by the GacA response regulator (26, 38) and that rsmA is required for maximal P_rsmY-lacZ and P_rsmZ-lacZ transcriptional reporter activity (39) (Fig. 1A). We first tested whether the rsmA defect could be corrected by GacA expressed from a plasmid. In the absence of either gacA or rsmA, P_rsmY-lacZ and P_rsmZ-lacZ reporter activities are very low (<300 Miller units) in cells carrying a vector control (pJN) (Fig. 1B). Complementation of the gacA mutant resulted in P_rsmY-lacZ and P_rsmZ-lacZ reporter activities that were much higher than observed in wt cells (Fig. 1A vs 1B). Likewise, expression of GacA in the rsmA mutant also resulted in high levels of reporter activity (Fig. 1B). The latter finding indicated that RsmA is not strictly required for rsmY and rsmZ transcription and suggested that a defect in GacA signaling might account for the lack of P_rsmY-lacZ and P_rsmZ-lacZ reporter activity in the rsmA mutant.

FIG 1.

RsmA inhibits RetS synthesis. (A–B) The indicated strains were transformed with either a vector control (pJN), an RsmA expression vector (pRsmA in panel A), or a GacA expression vector (pGacA in panel B). Cells were cultured in tryptic soy broth and assayed for P_rsmY-lacZ (open bars) or P_rsmZ-lacZ (gray bars) transcriptional reporter activity. The reported values in Miller units are the average of at least three independent experiments with the standard error, *, P < 0.01 when comparing wt and the rsmA mutant with the vector control to cells carrying the pRsmA complementation vector. (C) Whole cell extracts from the indicated strains carrying a Tn7 chromosomal integrant expressing VSV-tagged GacA or GacS, or native RetS from the chromosome, were immunoblotted using polyclonal serum to VSV or RetS. Cells lacking the Tn7 integrants (for GacA and GacS) or an retS mutant served as negative controls. The indicated band (#) in the RetS immunoblot is an unrelated, cross-reactive protein that served as a loading control.

In addition to the GacA response regulator, the Gac/Rsm signaling system includes the GacS sensor kinase, and the orphan sensor kinases LadS and RetS. LadS and RetS stimulate and inhibit GacS activity, respectively (33, 36, 40–42). We next tested whether RsmA influences the expression of GacA, GacS, LadS, and/or RetS. Because RsmA can have either positive or negative effects on gene expression, we predicted that RetS levels might be elevated in an rsmA mutant or that GacS, GacA, and/or LadS levels might be reduced. To detect GacA, GacS, and LadS, a VSV-epitope tag was fused to the carboxy-terminal end of each gene. The fusions were integrated at the Tn7 attachment site under the transcriptional control of their native promoters. RetS was detected using antibody raised against a soluble cytoplasmic portion of the protein. GacA and GacS levels were similar in wt cells and the rsmA mutant, while RetS expression increased ∼2.5 fold in the rsmA mutant (Fig. 1C and 2C). We were unable to detect VSV-tagged LadS in either wt cells or the rsmA mutant.

FIG 2.

The rsmA requirement for rsmY/rsmZ transcription is conditional. (A) Model for the Gac/Rsm system. The GacS sensor kinase phosphorylates GacA to activate transcription from the P_rsmY and P_rsmZ promoters. RsmY and RsmZ are small non-coding RNAs that sequester RsmA and RsmF. RsmA is an RNA binding that directly binds its own mRNA to inhibit RsmA synthesis. RsmA also inhibits RsmF synthesis. RetS and LadS are hybrid sensors that inhibit and stimulate the GacSA system, respectively. In this study we demonstrate that RsmA directly inhibits RetS synthesis. (B and D) The indicated strains were cultured in tryptic soy broth and assayed for P_rsmY-lacZ (panel B, open bars), P_rsmZ-lacZ (panel B, gray bars), or P_retS-lacZ (panel D) transcriptional reporter activity. Reporter activity is significantly reduced in the rsmA mutant relative to the wt strain for the data presented in panel B, *, P < 0.05. (C) Whole cell lysates fractions prepared from the indicated strains were analyzed by SDS-PAGE and immunobloted for RetS. The indicated bands (#) indicated served as loading controls. The band corresponding to RetS was quantified by densitometry from three independent blots and is reported as the fold change in the rsmA mutant relative to the wt strain, *, P < 0.01.

The RsmA requirement for rsmY and rsmZ transcription is conditional.

The observation of increased RetS expression in an rsmA mutant suggested that RsmA inhibits RetS synthesis to promote rsmY and rsmZ transcription (Fig. 1A). The first test of that hypothesis was an epistasis experiment using an ΔrsmA, retS double deletion mutant. Because RetS inhibits rsmYZ transcription by antagonizing the GacSA system, P_rsmY-lacZ and P_rsmZ-lacZ transcriptional reporter activities are significantly elevated in a retS mutant relative to the parental strain (Fig. 2B). Whereas rsmA is required for maximal P_rsmY-lacZ and P_rsmZ-lacZ reporter activity, the RsmA requirement was entirely suppressed in the ΔrsmA, retS background (Fig. 2B). The requirement for rsmA, therefore, is conditional and only required when retS is present. Our finding that RetS protein levels are elevated ∼2.5 fold in the rsmA mutant (Fig. 2C) is also consistent with the model shown in Fig. 2A To verify that the increase in RetS protein did not reflect increased P_retS promoter activity, we measured P_retS-lacZ transcriptional reporter activity in the parent, rsmA, retS, and rsmA,retS mutants and found no significant differences between the strains (Fig. 2D).

RsmA directly interacts with the retS leader region.

The data in Fig. 2 suggest that RsmA inhibits RetS synthesis. To determine whether the RsmA regulatory effect is direct, we performed electrophoretic mobility shift assays with purified RsmA and a radiolabeled retS RNA probe consisting of the 24 nt 5′ untranslated region (43) and 96 nt of the coding sequence. Incubation of the retS probe with a high concentration of RsmA (160 nM) generated a product with retarded mobility (Fig. 3A, lane 1 vs 2). The apparent equilibrium constant (K_eq) for RsmA determined through titration experiments was ∼50 nM (Fig. 3B), which is within the range typical of physiologically relevant RsmA targets (21, 44).

FIG 3.

RsmA directly interacts with the retS mRNA. (A) A wt retS RNA probe (100 pM), consisting of the first 120 nt of the mRNA, and mutant probes with GGA1-3 substitutions to CCU were radiolabeled, incubated with 160 nM purified RsmA as indicated, and examined by electrophoretic mobility shift assays and phosphorimaging. The positions of the unbound radiolabeled retS probe (Unbound retS*) and RsmA-bound retS* probe are indicated. (B) The wt retS probe (100 pM) was incubated with the indicated concentrations of RsmA and examined by electrophoretic mobility shift assays. (C) mFOLD RNA secondary structure prediction of the retS mRNA leader region and a portion of the coding sequence. Candidate GGA RsmA binding sites 1–3 are labeled blue, the retS GUG start codon is labeled red, and nucleotides corresponding to a putative RBS are underlined. (D) Competition experiments to assess the specificity of RsmA binding to the retS mRNA. The radiolabeled wt retS RNA probe (100 pM) was premixed with a 10- or 100-fold molar excess of unlabeled wt retS probe (lanes 2, 4-5) or GGA2 mutant probe (lanes 6–7), and then incubated with 320 nM RsmA as indicated. The positions of the unbound radiolabeled retS probe (Unbound retS*), RsmA-bound retS* probe, and a presumed radiolabeled retS* and unlabeled retS trans RNA-RNA species (retS*-retS) are indicated.

Canonical RsmA binding sites consist of a GGA sequence that is presented in the loop portion of a stem-loop structure. In directly repressed targets the RsmA binding site usually overlaps the RBS. The 5′ untranslated region of the retS mRNA contains one GGA sequence (designated GGA1) that overlaps the RBS (Fig. 3C). An mFOLD (45) prediction of the retS leader region and a portion of the coding sequence placed GGA1 in the loop portion of a stem-loop structure. Two additional GGA sequences (designated GGA2 and GGA3) are present in the early portion of the retS coding sequence (Fig. 3C). mFOLD predicts that GGA2 is also presented in the loop portion of a stem-loop structure, and that GGA3 is base-paired in a stem structure and likely inaccessible for RsmA binding. To determine which GGA sites contribute to RsmA binding, each sequence was changed to CCU. Although the GGA3 site was dispensable for binding, RsmA was unable bind when sites GGA1 or GGA2 were disrupted (Fig. 3A, lanes 3–8). The former finding was confirmed by titration experiments where the apparent equilibrium constant (K_eq) for RsmA binding to the GGA3 probe was similar to the wt probe (∼50 nM) (Fig. S1 in the supplemental material).

To examine the specificity of RsmA binding we performed competition experiments using unlabeled retS probe as a specific competitor and the unlabeled mutated GGA2 probe as nonspecific competitor. While unlabeled retS probe was an efficient competitor for RsmA-retS complex formation (Fig. 3D, lane 3 vs 4–5), the mutated GGA2 probe was unable to competitively inhibit binding (lane 3 vs 6–7). These data demonstrate that RsmA binds the retS probe with high specificity. We also observed a product with retarded mobility following incubation of the radiolabeled retS probe with an excess of unlabeled retS probe (Fig. 3D, lanes 2 and 5). This product likely reflects retS base pairing with itself in trans. The species was not observed with radiolabeled probe used at a relatively low concentration [100 pM (lane 1)] and only a small fraction of the radiolabeled probe formed the complex, even in the presence of a 100-fold molar excess of unlabeled retS probe. We conclude that the product is formed inefficiently and is probably not physiologically relevant.

The GGA2 site is required for RsmA control of RetS synthesis.

The data presented thus far demonstrate that RetS synthesis is elevated in an rsmA mutant and that RsmA directly binds to the retS mRNA. To verify that the regulatory effect of RsmA on RetS synthesis is direct, we examined RetS synthesis when RsmA binding to the retS mRNA is disrupted. This was accomplished by two independent approaches. First, the GGA2 site in the retS coding sequence was changed to CCU on the chromosome (Fig. 4A). The second approach involved swapping the entire 24 nt untranslated leader region of retS on the chromosome for the 34 nt leader region from pnp (Fig. 4A). The pnp leader region lacks GGA sequences. Although mutagenesis of the GGA1 site was another potential approach, the proximity of GGA1 to the RBS makes it challenging to disrupt RsmA binding without also disrupting ribosomal interactions. The effect of the GGA2 and pnp substitutions on RetS synthesis was examined by immunoblot analysis in wt and rsmA backgrounds. The GGA2 substitution resulted in lower levels of RetS when compared to wt cells (Fig. 4B, lane 1 vs 4), for reasons that are unclear, and the pnp substitution had elevated levels of RetS (lane 1 vs 6). The key point, however, is that while RetS levels are clearly elevated in the rsmA mutant relative to wt cells (lane 1 vs 3), both the GGA2 (lane 4 vs 5) and pnp (lane 6 vs 7) substitutions abolished RsmA-dependent control of RetS synthesis. Both observations are consistent with RsmA directly interacting with the retS mRNA to inhibit RetS synthesis.

FIG 4.

Direct binding of RsmA to the retS mRNA is required for regulation. (A) Sequence of the retS leader region and a portion of the coding sequence with the GGA1 and GGA2 sites important for RsmA binding shown in blue and the start codon in red. The GGA2 strain was constructed by replacing the GGA2 sequence with CCU and the pnp strain was made by replacing the entire untranslated retS leader with the pnp mRNA leader region (underlined), both substitutions were incorporated into the genome at the native chromosomal location. (B–E) The indicated strains were cultured in tryptic soy broth. Whole cell lysates fractions were analyzed by SDS-PAGE and immunobloted for RetS. The indicated bands (#) served as loading controls. Cells in panel C and D were assayed for P_rsmY-lacZ and P_rsmZ-lacZ transcriptional reporter activity. Cells in panel E were assayed for P_exsD-lacZ and P_tssA1’-‘lacZ reporter activity. The immunoblot blots are representative data from at least three experiments and the reported values in Miller units are the average of at least three experiments with the standard error, *, P < 0.01.

To determine the contribution of RsmA-dependent control of RetS synthesis to rsmY/Z transcriptional control we measured P_rsmY-lacZ and P_rsmZ-lacZ reporter activity. In the GGA2 background reporter activity was increased relative to wt cells but less than the levels observed for the ΔretS strain (Fig. 2B vs 4C). That finding correlates well to RetS levels in the GGA2 background, which fall between wt cells and the ΔretS mutant (Fig. 4B, lane 1 and 2 vs 4–5). Deletion of rsmA in the GGA2 background, however, had no significant effect on P_rsmY-lacZ and P_rsmZ-lacZ reporter activity (Fig. 4C). The latter findings are similar to the epistasis results in Fig. 2B, and suggests that the low level of RetS in the GGA2 mutant is sufficient to override the RsmA requirement observed in wt cells. In the pnp substitution mutant, RetS protein levels were elevated relative to wt cells, and P_rsmY-lacZ and P_rsmZ-lacZ reporter activities were significantly lower than seen for wt cells (Fig. 2B vs 4B-C). The observation that elevated levels of RetS result in reduced P_rsmY-lacZ and P_rsmZ-lacZ reporter activity is consistent with the model presented in Fig. 2A The activity of both reporters was even lower in the pnp, rsmA background suggesting that the effect of RsmA on RetS synthesis may only partially account for the regulatory effect that RsmA has on rsmY and rsmZ transcription.

RetS demonstrated an increasing gradient of expression in the GGA2, wt, rsmA, and pnp strains, with the latter showing the highest expression levels (Fig. 4B). This allowed us to directly test the hypothesis that the level of RetS controls P_rsmY-lacZ and P_rsmZ-lacZ reporter activity. Indeed the activity of both reporters was inversely correlated to RetS expression levels being maximal in the absence of retS followed in order by the GGA2, wt, pnp, and rsmA strains (Fig. 4D). Thus, relatively small changes in RetS expression have pronounced effects on P_rsmY-lacZ and P_rsmZ-lacZ reporter activity.

Two important virulence factors controlled by the Gac/Rsm system are the type III (T3SS) and type VI (T6SS) secretion systems. RsmA is required for T3SS gene expression (38, 46) and inhibits the T6SS (31). Our model predicts that as RetS levels increase: (i) RsmY/Z levels decrease, (ii) RsmA availability increases, (iii) expression of the T6SS is repressed, (iv) T3SS gene expression is stimulated. To test these predictions, T3SS (P_exsD-lacZ) (47) and T6SS (P_tssA1’-‘lacZ) (39) reporters were assayed in the retS, GGA2, wt, and pnp backgrounds. As expected, T3SS P_exsD-lacZ reporter activity was correlated with RetS expression levels while T6SS P_tssA1’-‘lacZ reporter activity was inversely correlated with RetS levels (Fig. 4E).

Control of RsmA activity by RsmY/Z maintains homeostasis of the Gac/Rsm system.

We next tested the hypothesis that feedback control of rsmYZ transcription by RsmA contributes to homeostasis of the Gac/Rsm system. An arabinose-inducible RsmA expression vector was introduced into rsmA and rsmA, rsmYZ (AYZ) mutants, and cells were assayed for expression of the P_rsmY-lacZ and P_rsmZ-lacZ transcriptional reporters. As expected, reporter activity was unresponsive to arabinose addition in the rsmA and rsmAYZ mutants carrying the vector control and demonstrated dose-dependent increases in both backgrounds when RsmA expression was induced with arabinose (Fig. 5A and B). The basal levels (no arabinose addition) of P_rsmY-lacZ and P_rsmZ-lacZ reporter activity were elevated when complemented for RsmA in the absence of rsmYZ (Fig. 5A and B, closed circles vs closed squares). The latter finding can be explained by the high levels of free RsmA, which in turn increase rsmY and rsmZ transcription. These data demonstrate that RsmYZ play a role in suppressing their own transcription.

FIG 5.

Roles of RsmA and RsmYZ in maintaining homeostasis of the Gac/Rsm system. (A–B) rsmA and rsmAYZ mutants transformed with either a vector control (pJN) or RsmA expression vector (pRmsA) were cultured in TSB with the indicated concentrations arabinose (to induce RsmA expression) and assayed for P_rsmY-lacZ (panel A) and P_rsmZ-lacZ (panel B) transcriptional reporter activity. The reported values in Miller units are the average of at least three experiments with the standard error, *, P < 0.01.

Increasing RsmA expression by arabinose titration revealed distinct kinetic differences in the activity patterns of the reporters. As described above, the primary rsmYZ effect on P_rsmY-lacZ reporter was at the level of basal activity (i.e., in the absence of arabinose). Aside from that difference, however, there was a similar dose-dependent increase in activity in both the absence and presence of rsmYZ (Fig. 5A, filled circles vs filled squares). At the highest level of RsmA expression (0.4% arabinose), the rsmYZ effect on reporter activity was negated. The expression pattern for the P_rsmZ-lacZ reporter differed in that the presence of rsmYZ appeared to more effectively suppress activity when RsmA was induced by arabinose titration (Fig. 5B, filled circles vs filled squares). Even at the highest level of RsmA expression (0.4% arabinose), reporter activity still did not fully reach that observed in the absence of rsmYZ.

Role of RsmF in control of rsmY and rsmZ transcription.

We also observed that P_rsmY-lacZ and P_rsmZ-lacZ reporter activity was significantly elevated in an rsmAYZ mutant when compared to the rsmA mutant (Fig. 5A and B, open circles vs open squares). We next tested whether that activity was attributable to RsmF, an RsmA paralog that generally plays a smaller role in the Gac/Rsm system (21). Although deletion of rsmA results in a significant decrease in P_rsmY-lacZ and P_rsmZ-lacZ reporter activity, deletion of rsmF had no significant effect and resembled the wt strain (Fig. 6A and B). This is consistent with prior studies showing that deletion of rsmF alone has negligible effects on RsmA-controlled phenotypes (eg., biofilm formation and T6SS gene expression) (21). With regard to the latter phenotypes, deletion of both rsmA and rsmF results in a synergistic increase in biofilm formation and T6SS gene expression that significantly exceeds that observed in the rsmA mutant (21). Deletion of rsmA and rsmF, however, had no significant effect on P_rsmY-lacZ and P_rsmZ-lacZ reporter activity when compared to the single rsmA mutant (Fig. 6A and B). Although RsmF activity is best detected in mutants lacking both rsmA and rsmYZ, there was no significant difference in reporter activity between the rsmA, rsmAF, rsmAYZ, or rsmAFYZ strains (Fig. 6A and B).

FIG 6.

Contribution of RsmF to rsmY and rsmZ transcription. (A-B) The indicated strains were cultured in TSB and assayed for P_rsmY-lacZ (panel A) and P_rsmZ-lacZ (panel B) transcriptional reporter activity. (C) An rsmAF mutant transformed with either a vector control (pJN), an RsmA expression vector (pRmsA), or an RsmF (pRsmF) expression vector was cultured in TSB with 0.2% arabinose (to induce RsmA/F expression) and assayed for P_rsmY-lacZ and P_rsmZ-lacZ transcriptional reporter activity. The reported values in Miller units are the average of at least three experiments with the standard error, *, P < 0.01.

The final question we addressed is whether RsmF activity is independent from RsmA. Expression of either RsmA or RsmF from a plasmid complemented an rsmAF mutant for P_rsmY-lacZ and P_rsmZ-lacZ reporter activity (Fig. 6C), demonstrating that RsmF activity can function independently of RsmA. Although RsmF activity is evident under these conditions, it appears to be reliant upon RsmF overexpression and could be an artifact. We conclude that RsmF can influence rsmYZ transcription in the absence of competition with RsmA and the sequestering RsmY and RsmZ sRNAs. The effect of RsmF on P_rsmY-lacZ and P_rsmZ-lacZ reporter activity, however, is modest under the conditions tested in this study and RsmA plays a much larger role in maintaining homeostasis of the Gac/Rsm system through control of rsmY and rsmZ transcription.

DISCUSSION

The widespread CsrA/RsmA family of RNA-binding proteins control gene expression at the post-transcriptional level. Regulation occurs on a global scale with targets including central metabolism, stress responses, and virulence. The well-studied E. coli CsrA system is tightly maintained through multiple positive and negative feedback loops (48). CsrA activity is controlled by two small non-coding sRNAs (CsrB and CsrC) that sequester CsrA from target mRNAs. Transcription of csrB and csrC is controlled by the BarA/UvrY TCS (49), which is homologous to the P. aeruginosa GacSA TCS (50). Similar to the RsmA requirement for rsmYZ expression in P. aeruginosa, CsrA is required for maximal csrBC expression in E. coli (51, 52). The mechanism of control by CsrA is multifaceted and involves positive effects on uvrY transcription and translation, and stimulation of BarA kinase activity. Collectively, those effects lead to increased levels of UvrY/BarA activity and csrBC transcription (48). CsrA availability, therefore, is indirectly linked to csrBC transcription and serves as one mechanism to prevent runaway expression of the CsrA regulon. An additional level of negative feedback control involves CsrA directly targeting the csrA mRNA and inhibiting its own synthesis (53).

Comparable regulatory mechanisms appear to serve a similar purpose in the P. aeruginosa Gac/Rsm system. RsmA directly targets the rsmA and rsmF mRNAs to inhibit synthesis of both proteins (21) and increased RsmA availability leads to elevated levels of rsmY/Z transcription (Fig. 5). Rather than directly targeting and stimulating GacSA TCS activity, however, we provide evidence that RsmA indirectly stimulates GacSA signaling through its inhibitory effect on RetS synthesis. RetS is an orphan sensor kinase that reduces rsmYZ expression by preventing GacS phosphorylation. P_rsmY-lacZ and P_rsmZ-lacZ reporter activity is inversely correlated to RetS expression levels (Fig. 4D). When compared to wt cells, RetS levels increase 2.5 fold in the rsmA mutant (Fig. 2C) and P_rsmY-lacZ and P_rsmZ-lacZ reporter activities decrease ∼40- and 10-fold, respectively (Fig. 4D). This seemingly modest change in RetS levels relative to the much larger change in reporter activity suggests that GacSA signaling is exquisitely sensitive to the effects of RetS.

Although rsmA is required for rsmYZ transcription in an otherwise wt background, epistasis experiments revealed that the rsmA requirement is entirely suppressed in the absence of retS (Fig. 2B). Our findings that purified RsmA binds the retS mRNA (Fig. 3A) and that disruption of that interaction abolishes RsmA-dependent restriction of RetS synthesis (Fig. 4B) is consistent with RetS being a direct regulatory target of RsmA. Our in vitro binding data are corroborated by a recent ChIPPAR-seq study showing that RsmA targets the nascent retS mRNA in vivo (54). Binding and mutagenesis studies show that the GGA1 and GGA2 sites on the retS mRNA are essential for RsmA binding (Fig. 3A). Interpretation of those findings is somewhat complicated by the RsmA requirement for both primary sequence determinants and secondary structure. Further studies would be required to determine whether those substitutions disrupt the RsmA binding site per se and/or alter the RNA secondary structure. mFOLD predictions of the GGA1 and GGA2 substitutions, however, indicate that neither alters the secondary structure of the RNA. Our findings that RsmA inhibits RetS synthesis and that GGA1 overlaps the RBS are supportive of a model wherein direct binding of RsmA to GGA1 inhibits translation initiation. The requirement for additional RsmA interactions with GGA2 likely reflects the dimeric nature of RsmA in solution and indicates that stable complex formation and/or inhibition of translation initiation is dependent upon both interactions.

The relative contribution of RsmA-dependent inhibition of RetS synthesis was examined by directly comparing reporter activity in wt cells and the pnp strain in the presence and absence of rsmA. P_rsmY-lacZ reporter activity was 42 times higher in wt cells relative to the rsmA mutant but only 11-fold higher in the pnp backgrounds, where RetS expression is no longer controlled by RsmA (Fig. S2). P_rsmZ-lacZ reporter demonstrated a similar pattern being 11 times higher in wt cells relative to the rsmA mutant but only 2.5 times higher in the pnp backgrounds. Control of RetS synthesis may not account for the entire RsmA effect. In the pnp strain RetS levels are higher than wt cells (Fig. 4B), and P_rsmY-lacZ and P_rsmZ-lacZ reporter activity is reduced (Fig. 4C). Deletion of rsmA in the pnp mutant results in a further reduction in reporter activity. There are two possible interpretations of this finding. The first is that a modest increase in RetS expression (below our detection limit) in the rsmA mutant accounts for reduced reporter activity. The second possibility is that another RsmA-controlled target contributes to rsmY and rsmZ promoter activity. Given that it is more common for RsmA to repress protein synthesis, as opposed to activate, and that GacA and GacS are required for rsmY/Z transcription (26), we considered GacA and GacS as potential targets. RsmA-dependent inhibition of GacA and/or GacS synthesis would reduce rsmY and rsmZ promoter activity. The expression of each, however, was similar in the wt and rsmA backgrounds (Fig. 1C). It remains possible that GacSA phosphorylation status and/or activities are influenced by RsmA. Another candidate was LadS. RsmA-dependent stimulation of LadS synthesis would promote rsmY and rsmZ promoter activity. Our inability to detect LadS prevented us from testing the effect of RsmA on LadS synthesis. While it is possible that another RsmA-regulated target participates in control of rsmY and rsmZ promoter activity, the epistasis findings with the rsmA, retS double mutant suggest that the effect is mediated largely, if not entirely, through RsmA effects on RetS.

Elevated levels of free RsmA lead to activation of rsmY/Z transcription, sequestration of RsmA by RsmY and RsmZ, and homeostatic control of the Gac/Rsm system. The contribution of RsmF to this process is less clear. Although P_rsmY-lacZ and P_rsmZ-lacZ reporter activity is essentially absent in an rsmA mutant, there was no significant difference in activity between wt cells and the rsmF mutant (Fig. 6A and B). Overexpression of RsmF in an rsmAF background stimulated high levels of P_rsmY-lacZ and P_rsmZ-lacZ reporter activity (Fig. 6C) indicating that RsmF has the potential to impact rsmY/Z transcription. That potential, however, does not appear to be significantly realized. Comparison of the P_rsmY-lacZ and P_rsmZ-lacZ reporters in the rsmA, rsmAF, rsmAYZ, and rsmAFYZ mutants revealed no significant differences in activity (Fig. 6A and B), and paradoxically both reporters were elevated in the rsmAF mutant when compared to the rsmA mutant. Although that comparison was not statistically significant it was the largest difference between those four strains and raises the possibility that RsmF has a more complex role, with potential for both positive and negative effects on rsmY/Z transcription. The effect of RsmF may be more prominent under different environmental conditions. We have observed one condition where the contribution of RsmF is more apparent. P_rsmY-lacZ and P_rsmZ-lacZ reporter activity were significantly elevated in an rsmAYZ mutant when compared to an rsmA mutant, but only when the strains were transformed with a vector control (Fig. 5A and B, no arabinose addition). The increases in reporter activity were entirely suppressed in an rsmAFYZ mutant (Fig. S3A in the supplemental material) when lacking the vector control, demonstrating a clear role for RsmF. We have noted this vector effect in the past and it appears to be directly related to vector dependent stimulation of P_rsmY-lacZ and P_rsmZ-lacZ reporter activity (Fig. S3B). While an explanation for the vector-dependent effect is lacking, the simplest explanation would involve enhancement of RsmF expression and/or activity since the effect is largely RsmF-dependent.

The regulatory output of the Gac/Rsm system is controlled by RsmA availability. Studies examining effects of the Gac/Rsm system have generally been limited to the use of strains with either low (eg., rsmA mutants) or high (eg., rsmYZ mutants) levels of RsmA availability. Such mutants have proven effective in defining the regulon but represent extreme ends of the RsmA availability scale that are probably never achieved under normal physiological conditions. In reality, RsmA availability appears to be tightly controlled within a narrower window. This is accomplished through feedback control of rsmY/rsmZ transcription by RsmA and direct inhibition of its own synthesis. Potential exists for additional layers of regulation including roles for the other sRNAs that sequester RsmA (i.e., RsmV and RsmW) (27, 28), RNA binding proteins that effect RsmY/Z stability (e.g., Hfq) (55, 56), and factors that influence rsmY or rsmZ transcription (e.g., HptB) (57). Integration of these additional factors into a comprehensive model will be necessary to fully appreciate the extent and significance of homeostatic control by the Gac/Rsm system.

MATERIALS AND METHODS

Strain and plasmid construction.

The strains and plasmids, cloning details, and primer sequences are provided in Tables S1–S3 in the supplemental material. Routine cloning was performed with E. coli DH5alpha cultured in LB-Lennox medium with tetracycline (12 μg/ml) or gentamicin (15 μg/ml) as required. The gacA allelic exchange vector was generated using the primer pairs listed in Tables S2 and S3 and cloned by isothermal assembly into pEXG2Tc (58, 59). The ΔgacA and ΔrsmAΔretS double mutants were generated by allelic exchange and sacB-mediated resolution as previously described (60).

β-galactosidase assays and immunoblots.

PA103 strains were cultured at 37°C overnight in LB containing 80 μg/ml gentamicin as required. The next day, strains were diluted to an absorbance (A₆₀₀) of 0.1 in tryptic soy broth (TSB) supplemented with 100 mM monosodium glutamate and 1% glycerol, and gentamicin (80 μg/ml) as required. Arabinose was added, as indicated in the figure legends, to induce RsmA or RsmF expression from the P_BAD promoter. Cultures were incubated at 37°C and harvested when the A₆₀₀ reached 1.0. β-galactosidase activity was performed as previously described with the substrates ortho-nitrophenyl-galactopyranoside (47) or chlorophenol red-β-d-galactopyranoside (CPRG) (61). CPRG activity was determined by measuring product formation at 578 nM as follows: CPRG units = [A₅₇₈/culture A₆₀₀/time (min)/culture volume (ml)] × 1,000.

Whole-cell lysate samples were prepared by pelleting 1.25 ml (A₆₀₀ = 1.0) of cells, suspending the pellet in 0.25 ml SDS-PAGE sample buffer, sonicating for 10 s, and heating for 5 min at 37°C. Samples were analyzed by SDS-PAGE followed by immunoblotting using RetS antibodies raised against the RetS signaling kinase sensory domain (62) and processed using enhanced chemiluminescent fluorescence detection reagents (Thermo Scientific, Rockford, IL).

RsmA and RsmF purification.

E. coli Tuner (DE3) expressing histidine-tagged RsmA or RsmF were cultured overnight in LB with ampicillin (200 μg/ml). Cells were diluted into a fresh 2-liter culture to an A₆₀₀ of 0.1 in LB with ampicillin (200 μg/ml). At an A₆₀₀ of 0.5, isopropyl-β-D-1-thiogalactopyranoside (IPTG) was added at 1 mM to induce protein production. Once the culture reached an A₆₀₀ of 3.0, cells were harvested by centrifugation and suspended in nickel nitrilotriacetic acid (Ni-NTA) binding buffer (20 mM Tris-HCL [pH 7.9], 300 mM NaCl, 5 mM imidazole) supplemented with 1 complete protease inhibitor cocktail tablet (Roche). Cells were lysed by passage through a microfluidizer at 17,000 lb/in² (Microfluidics, Westwood, MA). Cell lysates were cleared by centrifugation and immediately loaded onto a 1-ml HiTrap His column (GE Healthcare Life Sciences) and eluted with binding buffer containing 300 mM imidazole. The peak elution fractions were pooled and dialyzed against Ni-NTA binding buffer containing 5 mM dithiothreitol (DTT) for 4 h at 4°C. The resulting protein was flash-frozen in 1-ml aliquots. Two 1-ml frozen aliquots of RsmA or RsmF were buffer-exchanged by diluting to 10 ml with Ni-NTA binding buffer and then concentrated to 1 ml using an Amicon Ultra centrifugal filter (nominal molecular weight limit, 10,000 Da) and repeated five times. The concentrated RsmA or RsmF samples were then combined and exposed to 0.15 ml of preequilibrated Talon beads (Clontech) for 10 min, rocking at 4°C. Beads were then washed with 5 ml of Ni-NTA binding buffer three times and eluted with 10 ml of binding buffer containing 0.5 M imidazole. The eluted RsmA/RsmF were concentrated again and buffer-exchanged three times to storage buffer (20 mM Tris-HCl [pH 7.9], 300 mM NaCl, 1 mM DTT). Protein concentrations were determined using the Bradford assay.

Electrophoretic mobility shift assays.

DNA templates from PA103 genomic DNA or gBlocks (IDT) encoding the 5′ UTR and a portion of the ORF of various genes with and without mutations (details provided in Table S2 in the supplemental material) were PCR amplified by a primer with the T7 promoter at the 5′ end. PA103 UTR regions were identified after comparison to PA14 (44). RNA was generated in vitro using the MEGAshortscript™ T7 kit (Invitrogen) and end-labeled with [γ-³²]-ATP (Perkin Elmer) as previously described (21). Unincorporated nucleotides were removed using NucAway Spin Columns (Invitrogen). RNA in 20% 10x binding buffer (100 mM Tris-Cl pH [7.5], 100 mM MgCl₂, 1000 mM KCl, 50% glycerol) was heated to 96°C and allowed to cool to room temperature. Purified RsmA and RsmF were incubated with the probes in 1X binding buffer, 0.25% bovine serum albumen (New England Biolabs), 3.25 ng/μl total yeast tRNA (Life Technologies), 10 mM DTT, 5% glycerol, 2X folding buffer (50 mM HEPES [pH 7.2], 200 mM potassium acetate 0.6 mM magnesium acetate, 0.2% Triton), and 0.1 units RNaseOut (Invitrogen) for 30 min at 37°C and then mixed with 4 μl of gel loading buffer (50% glycerol, 0.05% xylene cyanol) and immediately subjected to electrophoresis on 7.5% (wt/vol) native polyacrylamide glycine gels (10 mM Tric-HCl [pH 7,5], 380 mM glycine, and 1 mM EDTA) at 4°C. Imaging was performed using the Sapphire Biomolecular Imager (Azure Biosystems).

Statistical analyses.

Data analyses comparing two groups were statistically analyzed using two-tailed unpaired t tests. One-way analysis of variance and Tukey’s posttest determined the statistical significance of three or more groups. Statistical analyses were performed using Prism 7 (GraphPad Software version, Inc. La Jolla, CA).