The Pseudomonas aeruginosa CrcZ RNA interferes with Hfq-mediated riboregulation

The RNA chaperone Hfq regulates virulence and metabolism in the opportunistic pathogen Pseudomonas aeruginosa. During carbon catabolite repression (CCR) Hfq together with the catabolite repression control protein Crc can act as a translational repressor of catabolic genes. Upon relief of CCR, the level of the Hfq-titrating RNA CrcZ is increasing, which in turn abrogates Hfq-mediated translational repression. As the interdependence of Hfq-mediated and RNA based control mechanisms is poorly understood, we explored the possibility whether the regulatory RNA CrcZ can interfere with riboregulation. We first substantiate that the P. aeruginosa Hfq is proficient and required for riboregulation of the transcriptional activator gene antR by the small RNA PrrF1-2. Our studies further revealed that CrcZ can interfere with PrrF1-2/Hfq-mediated regulation of antR. The competition for Hfq can be rationalized by the higher affinity of Hfq for CrcZ than for antR mRNA.


Introduction
Numerous studies have been performed during the last decade to decipher the function and structure of the RNA chaperone Hfq. Most studies were conducted in E. coli, and it is now well established that Hfq fulfills several functions in post-transcriptional regulation. It can stabilize small regulatory RNAs (sRNAs) and facilitate annealing between sRNAs and their target mRNAs. The latter mode of action may result either in translational repression accompanied by degradation of both RNAs or in translational activation and stabilization of the mRNA. In addition, it may stimulate polyadenylation of mRNAs, which can trigger 3´to 5´directional decay [1]. Moreover, there is accumulating evidence that Hfq can act per se as a translational repressor of mRNAs [2][3][4].
To fulfil its role in riboregulation, the E. coli Hfq hexamer (Hfq Ec ) has dedicated RNA binding surfaces, preferably binding uridine-rich stretches of sRNAs around the central pore of the proximal surface [5][6][7][8][9], and A-rich sequences, which are predominantly present around the ribosome binding sites of E. coli mRNAs [10], on the distal surface [11,12]. In agreement, amino acid (aa) exchanges in K56 located at the proximal site and Y25 located at the distal site abolished binding to poly(U) and poly(A)-tracts, respectively [7]. A third RNA binding site has been identified on the lateral rim and consists of conserved basic residues [13,14]. This basic patch binds to RNA with low sequence specificity, and appears to contribute to the annealing function of Hfq Ec [14,15]. The C-terminus of Hfq Ec may provide a fourth interaction site for RNA. An Hfq Ec variant, comprising only the conserved core (65 N-terminal aa) was non-functional in hfq-autoregulation and riboregulation [16]. Moreover, biophysical experiments supported an interaction of hfq mRNA with the C-terminus [12,17]. In a model [8] devised to explain the role of Hfq in riboregulation it is envisioned that the mRNA binding surfaces of the Hfq-hexamer serve to transiently increase the local concentration of two RNA substrates, whereas the ability of Hfq to stochastically facilitate base-pairing is ascribed to its inherent capacity to induce conformational changes in RNAs [14,18,19].
In contrast to E. coli Hfq, the P. aeruginosa (PAO1) Hfq (Hfq Pae ) lacks an extended C-terminus, but contains the conserved residues of the proximal and distal binding sites as well as a basic patch at the lateral rim [20]. In accordance, several reports have indicated that Hfq Pae can stabilize sRNAs [21][22][23] as well as larger protein-binding RNAs [24]. In addition, recent in vitro [15] and in vivo assays [21,22] indicated that Hfq Pae is proficient in canonical riboregulation, i.e. in sRNA-mRNA annealing.
In addition to riboregulation, our recent studies provided evidence that Hfq Pae acts as the principle post-transcriptional regulator of carbon catabolite repression (CCR) in P. aeruginosa by direct binding to target mRNAs [3]. CCR ensures that the utilization of less preferred carbon sources is impeded until the preferred one is consumed [25]. During growth on succinate (CCR) Hfq was shown to be required for translational silencing of several PAO1 catabolic genes, which was attributed to an interaction of the distal face of Hfq Pae with A-rich sequences within or adjacent to ribosome binding sites (RBS) [3]. Upon relief of CCR, e.g. after exhaustion of succinate and resumed growth on mannitol, the levels of the Hfq-binding RNA CrcZ increase, leading to sequestration of Hfq; this in turn abrogates Hfq-mediated translational repression of the respective catabolic mRNAs [26]. Thus, analogously to the CsrA/B/C system in E. coli and the RsmA/Y/Z system in Pseudomonas spp. [27], CrcZ acts as a RNA sponge for Hfq, and thus can cross-regulate catabolic genes [3].
In Enterobacteriaceae several examples are known, where regulatory sRNAs are titrated by competing endogenous RNAs. Base-pairing of sRNAs with mRNA, sRNA or even tRNA fragments other than their primary targets can interfere with regulation of the latter or reduce transcriptional noise resulting from production of sRNAs in unstressed cells [28]. As mentioned above, one alternative crosstalk is the competition of dedicated protein binding RNAs for regulatory proteins. Hierarchical binding of mRNAs can likewise exert competition for CsrA, which is exemplified by the control of fimbrial gene expression in S. enterica [29]. In E. coli, ectopic overexpression of a number of sRNAs was shown to compete with endogenous sRNAs for binding to Hfq Ec , which interfered with sRNA-mediated post-transcriptional regulation [30]. Similarly, over-expression of mRNA targets in the absence of the cognate regulatory sRNA was shown to diminish sRNA-mediated regulation of other target mRNAs [31]. These studies taken together with the observation that overproduction of Hfq Ec could compensate for some of the regulatory defects caused by over-production of sRNAs [31] indicated that Hfq Ec can be limiting for sRNA function.
As the interdependence of Hfq-mediated and RNA based control mechanisms is poorly understood in P. aeruginosa, we explored here the possibility whether the regulatory RNA CrcZ can interfere with riboregulation mediated by the sRNAs PrrF1-2. The PAO1 PrrF1-2 sRNAs are encoded in tandem, share 95% sequence identity, and are functional orthologues of the E. coli sRNA RyhB [32]. They are transcriptionally controlled by Fur, induced upon iron depletion, and implicated in post-transcriptional regulation of genes encoding functions involved in iron metabolism [32]. In addition, Oglesby et al. [33] showed that under iron limiting conditions the sRNAs PrrF1-2 reduce the levels of antR mRNA, encoding a transcriptional activator of the antABC operon, which is required for anthranilate degradation. It was hypothesized that this involves riboregulation by PrrF1-2 of antR mRNA [33].
In this study we first substantiated that Hfq Pae is proficient in and required together with the PrrF1-2 sRNAs for riboregulation of antR mRNA. Next, we addressed the question whether the regulatory PAO1 RNA CrcZ can interfere with riboregulation of antR by PrrF1-2.
To this end, we show that the increased synthesis of CrcZ is paralleled by de-repression of PrrF1-2-mediated regulation of antR. In contrast, ectopic overexpression of both crcZ and hfq resulted again in repression of antR, suggesting that Hfq can be limiting for PrrF1-2-mediated regulation of antR. We further provide evidence that the competition for Hfq can be rationalized by the higher affinity of Hfq for CrcZ than for antR mRNA.

Bacterial strains, plasmids and growth conditions
The strains and plasmids used in this study are listed in S1 Table. Unless indicated otherwise, the cultures were grown at 37˚C in BSM minimal medium [26] supplemented with 40 mM succinate. If required, E. coli was grown in the presence of 100 μg ml -1 ampicillin, 25 μg ml -1 tetracycline or 15 μg ml -1 gentamicin. PAO1 was grown in the presence of 250 μg ml -1 carbenicillin, 100 μg ml -1 tetracycline or 50 μg ml -1 gentamicin, respectively. The construction of plasmids used in this study is described in S1 Text.

β-galactosidase assays
The β-galactosidase activities were determined as described by Miller [34]. The cells were permeabilized with 5% toluene.

Protein purification
The Hfq Pae , Hfq PaeY25DFlag and Hfq PaeK56A proteins were produced in the hfq deficient E. coli strain AM111F 0 harboring the plasmids pHfq Pae , pHfq PaeY25DFlag and pHfq PaeK56A , respectively. The protein purifications were performed as described by Beich-Frandsen et al. [17]. Hfq Y25DFlag was used as it showed increased stability when compared to the untagged version.

Western-blot analyses
Equal amounts of total proteins were separated on 12% SDS-polyacrylamide gels and electroblotted to a nitrocellulose membrane. The blots were blocked with 5% dry milk in TBS buffer, and then probed with rabbit anti-Hfq (Pineda), rabbit anti-Flag (Roth) or rabbit anti-S1 (control) antibodies. The antibody-antigen complexes were visualized with alkaline-phosphatase conjugated secondary antibodies (Sigma) using the chromogenic substrates nitro blue tetrazolium chloride (NBT) and 5-Bromo-4-chloro-3-indolyl phosphate (BCIP).

Results and discussion
Hfq Pae functions in riboregulation: It stabilizes PrrF1-2 sRNAs and stimulates base-pairing between the sRNA PrrF2 and antR mRNA Previous studies revealed that PrrF1-2 binds to Hfq [37] of P. aeruginosa and that the antR levels were increased in the absence of Hfq [24] and of PrrF1-2 [33] , respectively. Taken with the in silico prediction of complementarity between PrrF1-2 RNAs and the translation initiation site of antR mRNA [33], these observations prompted us to adopt the PrrF1-2/antR entity to substantiate that Hfq Pae executes canonical riboregulation, i.e. facilitating base-pairing between a sRNA and its mRNA target.
We first used transcriptional and translational reporter genes to test the influence of Hfq Pae on PrrF1-2 mediated repression of antR. The strains were grown in BSM medium supplemented with 40 mM succinate and 2 μM FeSO 4 , which resulted in PrrF1-2 expression (Fig 1A, upper panel, atop lane 1). When compared with the wild-type strain no significant difference in the β-galactosidase activity conferred by the transcriptional antR-lacZ fusion gene was observed in the PAO1Δhfq strain (S1 Fig). At variance, despite the presence of PrrF1-2, the absence of Hfq Pae resulted in de-repression of antR::lacZ translational efficiency (Fig 1A, lane  2). Interestingly, the absence of PrrF1-2 but presence of Hfq in strain PAO1ΔprrF1-2 did not abolish repression to the same extend as seen in the absence of Hfq alone (Fi.g 1A, lane 3). Given that Hfq Pae has been shown to act as a translational repressor on several mRNAs [3], we speculate that Hfq Pae per se represses antR::lacZ translation, albeit less efficiently than in combination with PrrF1-2. Although Hfq Pae is also required for stabilization of PrrF1-2 (S2 Fig), additional ectopic expression of prrF2 did not result in antR::lacZ repression in the absence of Hfq Pae (Fig 1A, lane 5), indicating that the RNA chaperone is required for PrrF1-2/antR annealing.
As anticipated for canonical riboregulation [1], we next asked whether the proximal and distal interaction sites of Hfq Pae are required for binding of PrrF2 sRNA and antR mRNA, respectively. As shown in S3A Fig, the Hfq PaeY25D variant, which is deficient in distal site binding of A-rich sequences [7], did not bind to the antR (-95-+67) mRNA fragment encompassing the translational initiation region (TIR). Likewise, the Hfq PaeK56A mutant protein, which is defective in binding to U-rich sequences of sRNAs [7] was deficient in binding to PrrF2 RNA (S3B Fig). These EMSA assays are in agreement with the observation that ectopic expression of the hfq PaeY25D and hfq PaeK56A mutant alleles did not lead to repression of antR::lacZ translation (Fig 1B, lanes 3 and 4).
Vice versa, labelling of antR (-95-+67) mRNA confirmed that base pairing with PrrF2 is inefficient in the absence of Hfq Pae (S5B Fig, lanes 2-5 and lanes 7-10). In summary, these studies substantiated that Hfq Pae -like Hfq Ec -can perform canonical riboregulation, and thus analysis. 5S rRNA served as a loading control. The Hfq and S1 (loading control) levels were determined by western-blotting.
https://doi.org/10.1371/journal.pone.0180887.g002 The regulatory RNA CrcZ competes for Hfq and interferes with antR translation As both antR (-95-+67) (S3A Fig) and CrcZ [3] bind to the distal face of Hfq, we next tested whether ectopic overexpression of crcZ interferes with PrrF1-2-mediated regulation of antR. As shown in Fig 2 (lane 2), elevated levels of CrcZ increased the translational efficiency of the antR::lacZ gene approximately 1.7-fold, indicating that CrcZ can interfere with PrrF1-2/antR CrcZ interferes with Hfq-mediated riboregulation riboregulation. In agreement, overexpression of crcZ and hfq resulted in a similar repression of antR::lacZ as under non-induced conditions (Fig 2, lanes 1 and 3). This suggested that Hfq was limiting for PrrF1-2 function in the experiment shown in Fig 2 (lane 2). To rationalize the competition for Hfq Pae by CrcZ, we next determined the affinity of Hfq Pae for CrcZ and antR (-95-+67) using microscale thermophoresis. The K d was determined with~7.4 nM and~37.3 nM for CrcZ and antR (-95-+67) RNA, respectively (Fig 3). The low K d of Hfq Pae for CrcZ can be reconciled with six A-rich stretches in CrcZ to which Hfq Pae can bind with its distal binding site(s) [3]. Hence, it seems reasonable to assume that CrcZ can efficiently compete for Hfq Pae as long as the affinity of the distal binding surface of Hfq Pae for any given (m)RNA is lower than for CrcZ. This result is in agreement with the recent finding that CrcZ was shown to compete for the distal site of Hfq Pae with amiE mRNA to which Hfq binds with a K d of~67.0 nM [3].
Next, we asked whether this interference is also apparent under conditions that are closer to physiology. Given that the CrcZ levels increase in poor carbon sources, e.g. mannitol [26], the strain PAO1(pTLantR), harboring the plasmid borne antR::lacZ translational reporter gene, and the strain PAO1(pTCantR2), harboring the plasmid borne antR-lacZ transcriptional reporter gene, were grown in BSM medium containing either succinate (BSM-S) or mannitol (BSM-M). As shown in Fig 4, when compared with growth in BSM-S, the Hfq levels were unaltered and the CrcZ levels were increased approximately 2.7-fold in BSM-M. Under these conditions, the translational efficiency of the antR::lacZ gene in strain PAO1 was de-repressed during growth in BSM-M when compared with growth in BSM-S (Fig 4). It should be noted that de-repression occurred despite the PrrF1-2 levels were-for unknown reasonsincreased. Therefore, we concluded that this observation is attributable to the higher levels of CrcZ observed in BSM-M, i.e. to titration of Hfq by CrcZ.