Regulation of Hfq by the RNA CrcZ in Pseudomonas aeruginosa Carbon Catabolite Repression

Carbon Catabolite repression (CCR) allows a fast adaptation of Bacteria to changing nutrient supplies. The Pseudomonas aeruginosa (PAO1) catabolite repression control protein (Crc) was deemed to act as a translational regulator, repressing functions involved in uptake and utilization of carbon sources. However, Crc of PAO1 was recently shown to be devoid of RNA binding activity. In this study the RNA chaperone Hfq was identified as the principle post-transcriptional regulator of CCR in PAO1. Hfq is shown to bind to A-rich sequences within the ribosome binding site of the model mRNA amiE, and to repress translation in vitro and in vivo. We further report that Crc plays an unknown ancillary role, as full-fledged repression of amiE and other CCR-regulated mRNAs in vivo required its presence. Moreover, we show that the regulatory RNA CrcZ, transcription of which is augmented when CCR is alleviated, binds to Hfq with high affinity. This study on CCR in PAO1 revealed a novel concept for Hfq function, wherein the regulatory RNA CrcZ acts as a decoy to abrogate Hfq-mediated translational repression of catabolic genes and thus highlights the central role of RNA based regulation in CCR of PAO1.


Introduction
The opportunistic human pathogen Pseudomonas aeruginosa causes acute as well as chronic infections in immunocompromised individuals. Moreover, airway epithelia of patients suffering from cystic fibrosis are frequently colonized by the pathogen [1]. P. aeruginosa is a metabolically versatile organism with the ability to utilize numerous carbon sources, which allows the bacterium to thrive in different environments such as soil, marine habitats as well as on/in different organisms [2].
In Bacteria, the uptake and utilization of carbon compounds is controlled in a hierarchical manner by a mechanism known as carbon catabolite repression (CCR). Generally speaking, CCR prevents the utilization of less preferred carbon sources until the preferred one is consumed. In Escherichia coli CCR prevents the expression of catabolic genes, the transcription of which requires the transcriptional activator CRP (cyclic AMP receptor protein) in conjunction with cAMP, whereas in Bacillus subtilis CCR is mediated by the transcriptional repressor CcpA (catabolite control protein A). In both organisms CCR is regulated by a signal transduction pathway inherent to the phosphoenolpyruvatecarbohydrate phosphotransferase system [3].
In most studied Pseudomonas spp. the presence of organic acids (for example succinate) results in CCR, which leads to repression of catabolic genes required for the consumption of other carbon sources. During CCR catabolic genes were deemed to be downregulated by the translational repressor Crc (catabolite repression control protein) [4]. It has been suggested that Crc binds to CArich motifs within or adjacent to ribosome binding sites (RBS) of multiple target mRNAs, and thereby prevents their translation [5][6][7]. Upon relief of CCR, the regulatory RNAs, CrcZ in PAO1 [7], CrcZ/CrcY in P. putida [8] and CrcZ/CrcX in P. syringae [9] were proposed to bind to and to counteract Crc by trapping the protein. This hypothesis was in line with the observation that the CrcZ levels increase in the presence of poor carbon sources, and that they are reduced in the presence of a preferred carbon source [7]. However, our recent structural and biochemical studies challenged the role of Crc as a direct translational repressor of genes governed by CCR in PAO1. Recombinant Crc purified to homogeneity did neither bind to amiE mRNA, encoding aliphatic amidase, nor to CrcZ RNA [10,11]. Rather, the previously reported RNA binding activity of His-tagged Crc purified by nickel affinity chromatography [6,7] was attributed to a contamination of the Crc-His preparations with the RNA chaperone Hfq [10,11].
In Enterobacteriaceae Hfq is pivotal for riboregulation [12,13], which results on the one hand from binding to and protection of sRNAs from nucleolytic decay [14], and on the other hand from accelerating base-pairing between sRNAs and their target mRNAs [15][16][17]. E. coli Hfq hexamers have dedicated RNA binding sites, preferably binding uridine-rich stretches of sRNAs around the central pore of the proximal surface [18,19] and A-rich sequences on the distal surface [20]. In addition, the lateral surface of the hexamer can as well contribute to sRNA binding [21]. The dedicated sRNA and mRNA binding surfaces on either site of the Hfq-hexamer may serve to transiently increase the local concentration of two RNA substrates. Moreover, the inherent capacity of Hfq to induce conformational changes in RNAs together with the observed structural flexibility of RNA ligands bound to Hfq could stochastically facilitate base-pairing [22,23].
Although many sRNA candidates have been identified in PAO1 [24][25][26][27], the function of only a few has been revealed. The sRNAs PhrS [28] and PrrF [29] have been shown and inferred, respectively, to act by base-pairing with target mRNAs, whereas the protein binding RNAs RsmY and RsmZ are known to antagonize the function of the translational regulator RsmA [30]. PAO1 Hfq was shown to stabilize the protein binding RNA RsmY [31,32] and to affect expression of some sRNAs including PhrS [25]. In PAO1, Hfq acts as a pleiotropic regulator, impacting on growth, virulence, motility, and quorum sensing [31,33]. A transcriptome analysis of a PAO1hfq-strain revealed that ,15% of all genes were de-regulated. These included a number of genes encoding proteins involved in carbon compound catabolism, which were up-regulated in the absence of Hfq [31].
Here, we studied the impact of Hfq on CCR in PAO1. In vivo and in vitro studies revealed that Hfq acts as a translational repressor of several catabolic genes. Moreover, we present evidence that the regulatory RNA CrcZ binds to and sequesters Hfq, which in turn results in translation of Hfq-regulated mRNAs. Hence, this study revealed a novel mechanistic twist on posttranscriptional regulation by Hfq and highlights its role in regulating the central metabolism in P. aeruginosa.

Hfq represses catabolic genes at the post-transcriptional level
Several observations suggested a link between Hfq and CCR in PAO1. A comparative transcriptome analysis of a PAO1 wt and a PAO1hfq-strain disclosed transcripts encoding functions related to carbon compound and amino-acid catabolism that were upregulated in the absence of Hfq ( [31]; Table S1). Many of these transcripts comprise A-rich stretches (Table S1) within or adjacent to the RBS, which could serve as recognition motifs for the distal poly-A binding site of Hfq [20,34]. Mutations within these A-rich stretches in certain mRNAs abrogated CCR [7,35].
Crc was deemed to act as a translational regulator in PAO1 CCR [5][6][7]. However, Crc was recently shown to be deficient in binding to CrcZ and to the CCR regulated amiE mRNA [10]. In fact, these studies identified Hfq as a contaminant RNA binding activity in Crc preparations [10,11]. This observation prompted us to test (i) whether Hfq serves as the principle post-transcriptional regulator of CCR in PAO1, and (ii) whether the regulatory RNA CrcZ, displaying several A-rich motifs [7], might abrogate Hfqmediated regulation by sequestering Hfq.
We first revisited post-transcriptional regulation of amiE mRNA, encoding an aliphatic amidase. Strain PAO1(pTCamiE) and strain PAO1(pME9655) harboring a plasmid borne transcriptional amiE-lacZ and a translational amiE::lacZ fusion, respectively, were grown in BSM medium either in the presence of succinate/acetamide (CCR) or mannitol/acetamide (no CCR). Acetamide was added to induce transcription of the chimeric amiE genes, i.e. to mimic CCR. As shown in Figure S1A (left panel), the b-galactosidase activities conferred by the transcriptional fusion was comparable in either medium, i.e. in the presence and absence of CCR. However, when compared with growth in the presence of mannitol/acetamide (no CCR), amiE::lacZ translation was repressed in strain PAO1(pME9655) ( Figure S1A; right panel) during cultivation in the presence of succinate/acetamide, i.e. when CCR was in place.
We next tested whether amiE mRNA and CrcZ RNA associate with Hfq upon induction of amiE transcription during CCR. Strain PAO1hfq-harboring plasmid pMMBhfq Flag (encodes flag-tagged Hfq) and the control plasmid pMMB67HE, respectively, was grown to an OD 600 of 1.5 in BSM medium supplemented with succinate. In addition, acetamide was added to induce transcription of the amiE gene, i.e. to mimic CCR. Then, cell lysates were prepared and Hfq-associated RNAs were co-immunoprecipitated (CoIP) with Hfq-specific antibodies. As revealed by RT-PCR both, amiE and CrcZ, were found in complex with Hfq ( Figure S2). In contrast RsmZ, which does not bind to Hfq [7], could not be detected among the RNAs that co-immunoprecipitated with Hfq ( Figure S2). Taken together, these initial studies validated amiE as a model mRNA to scrutinize the hypothesized role of Hfq in CCR.
To obtain first hints whether Hfq is involved in posttranscriptional regulation of amiE during CCR, the strains PAO1, PAO1hfq-, PAO1Dcrc and PAO1hfq-Dcrc were transformed with plasmid pTCamiE harboring the transcriptional amiE-lacZ fusion and with plasmid pME9655 harboring the translational amiE::lacZ fusion, respectively. The strains were grown in BSM medium in the presence of succinate to establish CCR, and in the presence of acetamide to induce amiE-lacZ/amiE::lacZ transcription. The b-galactosidase activity conferred by the transcriptional amiE-lacZ fusion was comparable in the presence and absence of Hfq and/or Crc in either strain ( Figure S1B). In contrast, the bgalactosidase activity conferred by the translational amiE::lacZ fusion differed in strains PAO1, PAO1hfq-, PAO1Dcrc and PAO1hfq-Dcrc when grown in BSM medium containing succinate and acetamide (CCR). In contrast to PAO1, amiE::lacZ translation was greatly increased in the PAO1hfq-strain and in the PAO1hfq-Dcrc double mutant, respectively ( Figure 1A). The absence of Crc resulted as well in marked amiE::lacZ translation, albeit at a lower level when compared with the hfqmutant or the hfq-Dcrc double mutant, suggesting that Hfq exerts a more pronounced negative effect on amiE translation during CCR than Crc. To verify these experiments, we tested whether Hfq likewise affects translation of the estA and the phzM genes, which are also known to be regulated by CCR [7,35,36]. The impact of Hfq was again monitored using translational lacZ reporter gene fusions. The results obtained mirrored those obtained with the amiE::lacZ reporter gene. Their translation was repressed during growth of PAO1 in BSM medium containing succinate, whereas in the absence of Hfq, Crc or both, the synthesis of the encoded fusion proteins increased (Figures 1B

Author Summary
Carbon assimilation in Bacteria is governed by a mechanism known as carbon catabolite repression (CCR). In contrast to several other bacterial clades CCR in Pseudomonas species appears to be primarily regulated at the post-transcriptional level. In this study, we have identified the RNA chaperone Hfq as the principle post-transcriptional regulator of CCR in P. aeruginosa (PAO1). Hfq is shown to act as a translational regulator and to prevent ribosome loading through binding to A-rich sequences within the ribosome binding site of mRNAs, which encode enzymes involved in carbon utilization. It has been previously shown that the synthesis of the RNA CrcZ is augmented in the presence of non-preferred carbon sources. Here, we show that the CrcZ RNA binds to and sequesters Hfq, which in turn abrogates Hfq-mediated translational repression of mRNAs, the encoded functions of which are required for the breakdown of non-preferred carbon sources. This novel mechanistic twist on Hfq function not only highlights the central role of RNA based regulation in CCR of PAO1 but also broadens the view of Hfq-mediated post-transcriptional mechanisms. The error bars represent standard deviations from three independent experiments. The corresponding Crc Flag levels were determined by western-blot analysis using anti-Flag antibodies in strains PAO1Dcrc(pMMBcrc Flag ; pME9655), PAO1hfq-Dcrc(pMMBcrc Flag ; pME9655) and PAO1hfq-Dcrc(pMMB67HE; pME9655). Immunodetection of ribosomal protein S2 (loading control) was performed as described in Materials and Methods. (E) At an OD 600 of 1.0 IPTG (1 mM final concentration) was added to strains PAO1hfq-(pMMBhfq Flag ; pME9655) (blue bar), PAO1hfq-Dcrc(pMMBhfq Flag ; pME9655) (green bar) and PAO1hfq-Dcrc(pMMB67HE; pME9655) (purple bar), respectively. The bars represent the b-galactosidase values conferred by the plasmid pME9655 encoded translational amiE::lacZ fusion 3 h after hfq Flag induction. The error bars represent standard deviations from three independent experiments. The corresponding Hfq Flag levels were determined by western-blot analysis using anti-Flag antibodies in strains PAO1hfq-(pMMBhfq Flag ; pME9655), PAO1hfq-Dcrc(pMMBhfq Flag ; pME9655) and PAO1hfq-Dcrc(pMMB67HE; pME9655) (top panel). Immunodetection of ribosomal protein S2 (loading control) was performed as described in Materials and Methods. doi:10.1371/journal.pgen.1004440.g001 and C). In contrast, the expression of the heterologous lacZ gene (variation control), was comparable in strains PAO1, PAO1hfq-, PAO1Dcrc and PAO1hfq-Dcrc ( Figure S1C). Taken together, these initial studies supported our hypothesis that Hfq is involved in post-transcriptional regulation of CCR regulated genes.
Hfq acts as the principle post-transcriptional regulator and Crc has an auxiliary function in CCR As shown in Figure 1A-C, both, Crc and Hfq, were required for full repression of all three CCR regulated genes. Although translation occurred in the absence of Crc, a pronounced increase in translation only required the absence of Hfq, i.e. the observed de-repression was comparable in the hfqstrain and in the hfq-Dcrc double mutant. We interpreted this as showing that Hfq acts as the principal translational repressor, whereas Crc seemed to act as an auxiliary factor, somehow amplifying the negative regulation exerted by Hfq. To further test this hypothesis, amiE::lacZ translation was monitored in the PAO1hfq-Dcrc double mutant complemented with a plasmid borne crc Flag and hfq Flag gene, respectively. In consideration that crc could impinge on hfq expression and vice versa, the plasmid pMMBcrc Flag and pMMBhfq Flag borne crc Flag and hfq Flag genes, respectively, were equipped with the same expression signals, i.e. their expression was controlled by the P tac promoter and identical translation initiation signals. The different strains used in this experiment were grown in BSM medium supplemented with succinate and acetamide (CCR). At an OD 600 of 1.0, IPTG was added to induce ectopic expression of the crc Flag and hfq Flag genes, respectively. Three hours thereafter, Figure 2. Hfq binding to the RBS of amiE mRNA revealed by enzymatic probing. Lanes 1-4: sequencing reactions. Lane 5, primer extension (PE). Lanes 6-8, RNase A (A) cleavage in the absence and in the presence of increasing amounts of Hfq-hexamer (Hfq 6 ). Lanes 9-11, RNase T1 (T1) cleavage in the absence and in the presence of increasing amounts of Hfq 6 . The nucleotides indicated on the left of the autoradiograph and in the RNA sequence (below) are numbered with regard to the A (+1) of the amiE start codon. Arrowheads and arrows denote RNase A and RNase T1 cleavage, respectively. Black and grey symbols indicate strong and weak cleavage, respectively. Red circles indicate protection from RNase cleavage by Hfq. The region protected by Hfq is indicated by a red bar on the right of the autoradiograph. Only the relevant part of the autoradiograph is shown. The A-rich stretch from nt position 226 to 28, which can be potentially accommodated in the six distal binding pockets of Hfq, are indicated in red in the amiE RNA sequence. The Shine and Dalgarno sequence and the start codon of amiE are underlined. doi:10.1371/journal.pgen.1004440.g002 amiE::lacZ translation was monitored by determination of the bgalactosidase activities and the Crc-Flag and Hfq-Flag levels were determined by quantitative western-blot analysis.
Next, amiE::lacZ translation was monitored in strains PAO1hfq-(pMMBhfq Flag , pME9655) and PAO1hfq-Dcrc(pMMBhfq Flag ; pME9655) after growth in BSM succinate/acetamide medium (CCR) and 3 h after ectopic expression of the hfq Flag gene. The absence of Crc in the double mutant strain PAO1hfq-Dcrc(pMMBhfq Flag ; pME9655) ( Figure 1D; green bar) resulted in an increased de-repression of amiE::lacZ translation when compared with strain PAO1hfq-(pMMBhfq Flag ;, pME9655) ( Figure 1D; blue bar) despite comparable levels of Hfq Flag in both strains. Taken together, these experiments showed that Crc only impacts on amiE::lacZ translation in the presence of Hfq. Hence, they corroborate the idea that Crc does not act per se as a translational regulator but functions as an ancillary factor in Hfq-mediated repression of target genes.
Hfq binds to the translation initiation region of amiE mRNA and prevents in vitro ribosome binding and translation Next, we tested whether Hfq directly represses translation of amiE mRNA by binding to the translation initiation region (TIR).
First, a filter binding assay was performed with purified PAO1 Hfq and an amiE mRNA fragment encompassing nucleotides (nt) from position 2134 to +20 with regard to the A (+1) of the start codon. This experiment revealed that Hfq binds to amiE 2134-+20 with a K d of ,67.061.4 nM ( Figure S3). Next, the Hfq binding site(s) were mapped on amiE RNA. Enzymatic probing was performed with riboendonucleases T1 (G-specific cleavage) and A (C/Uspecific cleavage) in the absence and presence of Hfq. As shown in Figure 2, Hfq protected the segment of amiE RNA extending from G 228 to U 25 . This region includes the Shine and Dalgarno sequence (SD) sequence of amiE mRNA. Thus, Hfq binding to this region would readily explain the observed translational repression of amiE::lacZ mRNA ( Figure 1A, D, E).
To further test whether Hfq acts as a translational repressor of amiE mRNA, the PURExpress system was employed. The in vitro translation system was programmed with amiE Flag mRNA, encoding a Flag-tagged amidase. As shown in Figure S4, translation of amiE Flag mRNA was already impeded at a 1:1 molar ratio of Hfq to mRNA. As the in vitro system is reconstituted from purified components of the translation machinery of E. coli, no additional PAO1 component was apparently required for repression.
To further demonstrate that Hfq directly interferes with ribosome binding, a toeprinting assay was performed with amiE 2134-+76 mRNA in the presence and absence of Hfq. Briefly, in the presence of tRNA fMet , 30S ribosomes form a stable ternary complex at the RBS of mRNAs, which can be visualized by inhibition of cDNA synthesis primed downstream of the start codon [37]. A toeprint signal usually occurs at position +15 to +17 with regard to the A (+1) of the start codon. As shown in Figure 3, lane 7, a toeprint signal at the amiE RBS was observed in the absence of Hfq, whereas the addition of Hfq to amiE 2134-+76 , which is in accordance with earlier observations that translational repressors do not always provide a roadblock for cDNA synthesis by reverse transcriptase under these conditions [38]. Taken together, these in vitro studies strongly supported the idea that Hfq acts as a translational repressor that prevents ribosome loading on amiE mRNA.
The distal site of Hfq is required for regulation of amiE mRNA Link et al. [20] reported a crystal structure of E. coli Hfq in complex with poly(A 15 ), wherein the poly(A) tract is bound to the distal face using tripartite binding motifs. They consist of an adenosine specific site (A-site), a purine nucleotide selectivity site (R-site) and a sequence-non-discriminating E-site. The aminoacids involved in building the A-R-N motifs are fully conserved in the Hfq protein of PAO1 [34]. As the hexamer Hfq could accommodate the entire A-rich stretch from nt 226 to 28 of amiE mRNA ( Figure 2) in the six binding pockets, we next tested whether the distal binding site of Hfq is required and sufficient for translational repression of amiE mRNA. We therefore engineered the PAO1 hfq variants hfq Y25DFlag and hfq K56AFlag as the corresponding E. coli mutant proteins were shown to be deficient in binding to polyA-and polyU-tracts [18], respectively. In contrast to the PAO1 hfq Flag gene and the PAO1 hfq K56AFlag allele, ectopic expression of the hfq Y25DFlag allele did not result in repression of amiE::lacZ translation, albeit all three proteins, Hfq Flag , Hfq Y25DFlag and Hfq K56AFlag , were present at comparable levels ( Figure 4A). Basically the same results were obtained when translation of the estA::lacZ ( Figure 4B) and phzM::lacZ ( Figure 4C) fusion genes was monitored in the presence of Hfq Flag , Hfq Y25DFlag and Hfq K56AFlag , respectively. Hence, translational repression of these reporter genes apparently required an intact distal polyA binding site of Hfq.
To verify these in vivo data, the PAO1 Hfq variants Hfq Y25D and Hfq K56A were purified, and binding to amiE 2134-+20 mRNA was assessed using electrophoretic mobility shift assays (EMSA). As shown in Figure S5A, while PAO1 Hfq and Hfq K56A bound to the mRNA fragment, the Hfq Y25D protein failed to bind. With increasing concentrations of either Hfq or Hfq K56A two shifted bands were observed ( Figure S5A), suggesting that Hfq binds at two sites of the amiE 2134-+20 fragment. This observation can be explained by the Hfq binding site mapped between nucleotides G 228 to U 25 ( Figure 2) and by a probable second binding site, comprising an A-rich region (nucleotides 292 to 271), which was inferred from an RNomics approach after CoIP with Hfq-specific antibodies [25]. Taken the in vivo and in vitro studies together, these experiments strongly suggested that Hfq binds with its distal face to the RBS of amiE, and by inference most likely also to estA and phzM mRNA.
To further corroborate the idea that Hfq directly represses amiE translation the experiment was also performed in the heterologous E. coli hfqstrain JW4130. The strain was transformed with the control plasmid pME4510 and derivatives thereof harboring the PAO1 hfq Flag gene, the PAO1 hfq Y25DFlag allele and the PAO1 hfq K56AFlag allele, respectively. In addition, these strains were transformed with plasmid pME9658, wherein the amiL terminator preceding the amiE gene was deleted. As shown in Figure S5B, the experimental results paralleled that performed in PAO1. The translation of the amiE::lacZ gene was repressed in the presence of Hfq Flag and Hfq K56AFlag , whereas the reporter gene was translated in the presence of Hfq Y25DFlag . Thus, the presence of Hfq was apparently necessary and sufficient for repression of amiE translation in E. coli.

The regulatory RNA CrcZ titrates Hfq in vitro
The regulatory RNA CrcZ is present at lower levels during CCR, when compared to conditions when CCR is not in place, e.g. in BSM medium containing mannitol as the sole carbon source [7]. The 407 nt long CrcZ RNA contains six A-rich stretches ( Figure S6A) to which Hfq can potentially bind with its distal surface, i.e. with the same binding surface as it binds to the RBS of amiE ( Figure S5A). Binding of Hfq to the first 151 nt of CrcZ, containing 3 A-rich stretches ( Figure S6A) was confirmed using EMSA assays ( Figure S6B). Three shifted bands could be discerned ( Figure S6B, lane 2), which would be consistent with one Hfq-hexamer binding to either A-rich stretch. As anticipated, the PAO1 Hfq Y25D mutant protein was defective in binding to CrcZ 151 , whereas the Hfq K56A variant bound to the CrcZ fragment like native Hfq. Thus, CrcZ binds to the distal face of Hfq like amiE mRNA, and therefore has the potential to titrate Hfq, which in turn would explain the observed increase in translation of amiE::lacZ mRNA in BSM mannitol medium, i.e. in the absence of CCR (see Figure S1A).
In the next set of experiments we therefore asked whether CrcZ can abrogate Hfq-mediated translational repression in vitro. First, a mobility shift assay was performed with radioactively labeled amiE 2134-+20 RNA in the presence of unlabelled specific CrcZ 151 and non-specific RsmZ competitor RNAs, respectively. As shown in Figure 5A, CrcZ 151 competed with amiE 2134-+20 for binding to Hfq ( Figure 5A, lanes 5 and 6), whereas the addition of RsmZ RNA, which does not bind to Hfq [7], did not result in a downshift of labeled amiE 2134-+20 RNA ( Figure 5A, lane 7). When compared with the addition of unlabelled amiE 2134-+20 RNA, the addition of an equimolar concentration of CrcZ 151 resulted already in a significant downshift, i.e. loss of Hfq binding to amiE 2134-+20 . Hence, CrcZ 151 acted as a better competitor for Hfq than amiE 2134-+20 , which is anticipated as the used CrcZ 151 has three Hfq binding sites, whereas the amiE fragment has only two.
Next, amiE Flag mRNA was translated in vitro in the presence of Hfq as well as in the presence of Hfq and CrcZ mRNA. As shown before ( Figure S4), Hfq inhibited translation of amiE Flag , ( Figure 5B, lanes 2-4). In contrast, the mRNA was translated in the presence of CrcZ ( Figure 5B, lanes 5 and 6), whereas the Hfqmediated translational repression was not relieved in the presence of the non-specific competitor RNA RsmZ ( Figure 5B,  lane 7).
Similarly, the addition of CrcZ RNA to the Hfq-amiE complex, under conditions that inhibited translation initiation complex formation, resulted in reappearance of a toeprint signal (Figure 3, lanes 10 and 11). Thus, CrcZ counteracted the function of Hfq. Taken together, these in vitro experiments indicated that CrcZ can titrate Hfq, and that it can prevent it from binding to the amiE RBS.

CrcZ-mediated regulation of catabolic genes in vivo depends on Hfq
In our model titration of Hfq by CrcZ would result from increased CrcZ levels [7] in the presence of non-preferred carbon sources, which in turn would lead to de-repression of Hfqregulated genes. Hence, it would be anticipated that ectopic overexpression of CrcZ would cause de-repression of Hfq-regulated genes even during CCR, i.e. under conditions when the CrcZ levels are low [7]. With this line of reasoning, we next asked whether ectopic expression of crcZ during CCR results in increased expression of the estA gene, which was assessed by monitoring the esterase activity. To avoid any interference of Crc, the experiments were performed in a PAO1Dcrc strain. The control strain PAO1Dcrc(pMMB67HE) and the crcZ over-expressing strain PAO1Dcrc(pMMBcrcZ) were grown in BSM medium supplemented with 40 mM succinate to an OD 600 of 1. Then, crcZ expression was induced and samples were taken 60 minutes thereafter. Esterase was produced even without induction of crcZ, which can be reconciled with the absence of Crc (see Figure 1B) and by endogenous background expression of crcZ in strain PAO1Dcrc(pMMB67HE) ( Figure 6A, upper panel). Nevertheless, ectopic over-expression of crcZ ( Figure 6A, upper panel) resulted in Similarly, over-expression of crcZ in PAO1 resulted in derepression of amiE::lacZ translation during CCR in BSM succinate/acetamide medium ( Figure S7A), which is in agreement with our model wherein CrcZ titrates Hfq. In contrast to strain PAO1(pMMBcrcZ;pME9655) ( Figure S7A), ectopic expression of crcZ did not affect translation of amiE::lacZ in the hfqstrain PAO1hfq-(pMMBcrcZ;pME9655) ( Figure S7B), clearly showing that CrcZ-mediated regulation requires Hfq.
The model would further specify that a deletion of the crcZ gene should increase repression of amiE::lacZ during CCR. As shown in Figure 6B, when compared with strain PAO1(pME9655), amiE::lacZ translation was even further repressed in strain PAO1DcrcZ (pME9655) when grown in BSM medium supplemented with succinate and acetamide (CCR).
Moreover, amiE::lacZ translation during CCR was indistinguishable in the absence of Hfq in strain PAO1hfq-(pME9655) and in the double mutant PAO1hfq-DcrcZ(pME9655). Taken together, these experiments showed that the hfq deletion is epistatic to crcZ, in other words that CrcZ exerts regulation on amiE::lacZ only in the presence of Hfq. Hence, these in vivo experiments lend further support to the hypothesis that CrcZ titrates Hfq, and thereby abrogates Hfq-mediated translational repression.

Target genes of Hfq in PAO1
In this study, we provided evidence that Hfq represses three CCR regulated genes, amiE, estA and phzM. Given that Hfq blocked translation initiation of amiE mRNA by binding to A-rich stretches encompassing the RBS, we revisited catabolic and transport genes that were found to be up-regulated in a comparative transcriptome analysis of a PAO1 and a PAOhfqmutant [31]. Out of 126 putative target mRNAs, 72 transcripts contained A-rich stretches in the TIR (Table S1). Out of the latter, 28 transcripts contain several single non-consecutive A-R-N motifs, whereas 44 transcripts contained consecutive (A-R-N) n repeats with n$3 either within or in close proximity to the RBS (Table S1), indicating that Hfq could interfere with ribosome binding. Although it remains to be tested for either candidate transcript it is tempting to speculate that Hfq is involved in translational regulation of several genes controlled by CCR. In addition, it is conceivable that several CCR controlled genes, which do not contain A-rich regions in the TIR [39], are indirectly controlled by Hfq.

Hfq in CCR of Pseudomonas aeruginosa: regulating a few at a time
Based on our in vivo and in vitro studies with the amiE model mRNA, we suggest that Hfq acts as a translational repressor of transcripts subjected to CCR in PAO1 ( Figure 7). As there appear to be several Hfq regulated genes (Table S1) and our preliminary results show that Hfq levels are more or less constant during growth in different carbon sources (E. Sonnleitner, unpublished), the question arises how Hfq can regulate numerous mRNAs. Many if not all catabolic genes including the corresponding transporter genes are primarily regulated at the transcriptional level [4,40], and their transcription usually requires the presence of the respective catabolite. Therefore, it is rather unlikely that many of the putative target genes (Table S1) are concomitantly induced at the same time. Thus, in the presence of a preferred carbon source, e.g. succinate, only a few other catabolites may induce concomitant transcription of the corresponding catabolic genes. Moreover, translational repression most likely leads to degradation of mRNAs [41,42] encoding catabolic enzymes other than those required for the breakdown of succinate, and thus to recycling of Hfq. Therefore, CCR control may not require vast amounts of Hfq. This hypothesis is supported by the experiment shown in Figure S8. When compared with the absence of CCR (presence of acetamide only) or with the PAO1hfq-strain, amiE mRNA was faster degraded in PAO1 during CCR in the presence of succinate and acetamide. In addition, we have estimated the number of Hfq hexamers/cell in BSM-succinate medium with 2160+/256 per PAO1 cell ( Figure S9), which might suffice to silence ''a few catabolic transcripts'' that are induced during CCR (Figure 7).
In the presence of non-preferred carbon sources, the two component system CbrAB is activated; phosphorylated CbrB binds to the RpoN-dependent promoter of crcZ and stimulates CrcZ synthesis [7,43]. Our data (Figures 5 and 6) suggest that CrcZ then binds to and titrates Hfq by virtue of its A-rich binding motifs. This in turn would allow translation of transcripts that are induced by available catabolites, and thus synthesis of the cognate degradative enzymes (for example aliphatic amidase) (Figure 7).
For the following reasons we favor the hypothesis that no other sRNAs are required for regulation of the catabolic genes examined in this study. Most E. coli sRNAs bind to the proximal site of Hfq, and binding is usually abrogated in the Hfq K56A mutant protein [13]. However, in contrast to the Hfq Y25D variant, the Hfq K56A mutant protein was still capable to repress the fusion genes governed by CCR in PAO1 ( Figure 4) and in E. coli ( Figure S5B). It seems therefore reasonable to suggest that regulatory sRNAs other than CrcZ are not involved in Hfq-mediated regulation of these catabolic genes.
A novel concept for Hfq function: from mediating riboregulation to being regulated Although binding of E. coli Hfq to A-rich stretches in mRNA has been demonstrated in several model systems (see below), Hfq seems to work predominantly in conjunction with sRNAs [44][45][46]. Hfq-mRNA binding may recruit the sRNA to the target mRNA and stimulate sRNA-mRNA pairing. In this scenario, the sRNA competes with initiating 30S ribosomes, whereas Hfq has a rather indirect function.
Some deviations from the canonical model of Hfq assisted and sRNA-mediated regulation of target mRNAs have been reported. In E. coli evidence has been provided that Hfq acts as an autogenous repressor on its own mRNA. As for PAO1 amiE mRNA (Figure 2), E. coli Hfq was shown to bind to an A-rich sequence encompassing the SD sequence of hfq mRNA [47]. Another example for translational repression by Hfq entails regulation of the E. coli sdhC mRNA by the sRNA Spot 42. Desnoyers and Massé [48] reported that Spot 42 binds upstream of the RBS and recruits Hfq, which then directly represses translation. Moreover, E. coli Hfq was recently shown to act as a repressor of cirA mRNA translation in the absence of a sRNA [49]. Interestingly, the translational block exerted by Hfq was shown to be abrogated by RyhB RNA pairing to cirA mRNA [49].
Similarly, the model shown in Figure 7 entails direct repression of target mRNAs by PAO1 Hfq. However, in contrast to Spot 42 [48] and RyhB [49] RNA, which recruit Hfq to and abrogate Hfqmediated translational repression of the target mRNA, respectively, CrcZ RNA acts as a decoy for Hfq. The CrcZ RNA contains six (A-R-N) n repeats (n$4) ( Figure S6A) that can potentially be exploited for binding to the distal face of Hfq. Thus, in the absence of CCR Hfq is most likely sequestered by CrcZ, and therefore not available to act as a translational repressor on target mRNAs. In this way the Hfq/CrcZ regulatory system is reminiscent to the Pseudomonas RsmA/RsmY/RsmZ system and the CsrA/CsrB system of several Gram-negative Bacteria [30,50,51]. The ability of Hfq to bind to RsmY [31] in fact poses the question whether the regulatory systems are interlinked. This could provide an explanation why different carbon sources can affect virulence traits such as biofilm formation and antibiotic resistance [39,52].
Moreover, given its constellation with Hfq, it seems possible that CrcZ on the one hand can interfere with any mRNA-dependent role of Hfq. On the other hand, the role of Hfq in sRNA-mediated riboregulation remains ill defined in PAO1. Although Hfq appears to be required for the PAO1 sRNAs PrrF1/PrrF2 to regulate target mRNAs [53], it is unclear whether Hfq exerts this function by stabilizing the sRNAs and/or by stimulating sRNA-mRNA pairing. As CrcZ binds to the distal site of Hfq it is conceivable that Hfq might even still be able to bind sRNAs such as PrrF at the proximal site, and thus to protect them from degradation. In any case it will be interesting to study whether sequestration of Hfq by CrcZ indirectly affects other Hfq-mediated processes in P. aeruginosa.

The elusive function of Crc in CCR of Pseudomonas
We have recently shown that the Crc protein is devoid of RNA binding activity [10], and that the previously observed RNA binding activity of Crc could be attributed to Hfq [11]. However, Crc contributes to Hfq-mediated repression during CCR (Figure 1). In addition, Crc was shown to impact on biofilm formation [54,55], virulence and antibiotic susceptibility [52], traits which are also affected by Hfq [31,33]. This raises the question how Crc impacts on Hfq function. Crc appears not to interfere with hfq expression. The translation of a hfq::lacZ reporter gene, whose transcription is driven by the authentic hfq promoter, was indistinguishable in PAO1 and in the isogenic PAO1Dcrc strain ( Figure S10A). In addition, the levels of Hfq, as determined by quantitative western-blot analysis, were not significantly altered in PAO1Dcrc when compared with the wild-type strain ( Figure  S10B). Conversely, Hfq seems not to affect the cellular concentration of Crc ( Figure S10B). Thus, Crc seems not to impact on the Hfq levels, and vice versa Hfq seems not to affect that of Crc.
RelA was recently shown to enhance multimerization of E. coli Hfq, and thereby to stimulate binding to sRNAs [56]. We therefore considered the possibility that Crc might interact with Hfq to increase the specificity of Hfq for A-rich sequences. However, as revealed by EMSA assays, the presence of Crc did not increase the affinity of PAO1 Hfq for amiE 2134-+20 RNA (E. Sonnleitner, unpublished), making it less likely that Crc acts similar to RelA. Nevertheless, we are currently exploring the possibility whether Crc is associated with Hfq in vivo.

Bacterial strains, plasmids and growth conditions
The strains and plasmids used in this study are listed in Table  S2. Unless indicated otherwise, the cultures were grown at 37uC in BSM minimal medium supplemented with 40 mM succinate. The strains PAO1hfq-Dcrc and PAO1hfq-DcrcZ were constructed by homologous recombination. Briefly, plasmid pME9672 and plasmid pME9673, respectively, were mobilized into strain PAO1hfq-with the aid of E. coli strain HB101(pRK2013), and then chromosomally integrated through selection for tetracycline resistance. Excision of the vector by a second crossover event was achieved by enrichment for tetracycline-sensitive cells [57]. If required E. coli and PAO1 were grown in the presence of 100 mg ml 21 ampicillin, 25 mg ml 21 tetracycline or 25 mg ml 21 kanamycin and 50 mg ml 21 gentamicin, 100 mg ml 21 tetracycline or 250 mg ml 21 carbenicillin, respectively. Details on the construction of plasmids used in this study are provided in Text S1.

b-Galactosidase assays
The b-galactosidase activities were determined as described by Miller [58]. The cells were permeabilized with 5% toluene.

Western-blot analyses
The protein levels of Hfq and Crc fused to C-terminal Flag-tags (DYKDDDDK) were determined in the respective strains and under the growth conditions as specified in the legends to the Figures. 1 ml aliquots of the respective cultures were withdrawn; the cells were harvested by centrifugation, resuspended and boiled in protein sample buffer. Equal amounts of total protein were separated on 12% SDS-polyacrylamide gels and then electroblotted to a nitrocellulose membrane. The blots were blocked with 5% dry milk in TBS buffer, and then probed with rabbit anti-DYKDDDDK polyclonal antibody (Roth). 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-3indolyl phosphate (BCIP).

Enzymatic probing of RNA
Five pmol of in vitro transcribed full length amiE RNA was incubated in RT-buffer (50 mM Tris pH 8.3, 60 mM NaCl, 6 mM Mg-acetate, 10 mM DTT) at 37uC for 30 min in a 10 ml reaction with 0, 5 and 20 pmol of purified Hfq protein. Then 2 U RNase T1 (1 ml), 10 pg RNase A (1 ml) or 1 ml of RNase free H 2 O was added and incubated for additional 10 min followed by phenol/chloroform extraction and precipitation. 200 fmol RNase treated or untreated RNAs were further used for the primer extension reaction with AMV reverse transcriptase (Promega), which was primed with the 59-[ 32 P]-labeled C78 oligonucleotide to test for protection by Hfq of the proximal part of amiE mRNA. For sequencing amiE RNA and ddNTPs (Fermentas) were used in primer extension reaction(s).

Toeprint analysis
The amiE 2134-+76 RNA used for toeprinting was obtained as described above. The [ 32 P]-59-end labeled oligonucleotide Q99 was annealed to amiE mRNA (+57 to +76 with regard to the A (+1) of the start codon) and used to prime cDNA synthesis by MMuLV reverse transcriptase (Thermo Scientific). The toeprinting assay was carried out with purified E. coli 30S ribosomal subunits and E. coli initiator-tRNA (tRNA fMet ) as described by Hartz et al. [37]. The mRNA (0.05 pmol) was pre-incubated at 37uC for 10 min with or without 4 pmol 30S subunits and 16 pmol tRNA fMet before reverse transcriptase was added. To test whether Hfq interferes with translation initiation, 0.05 pmol amiE 2134-+76 mRNA was pre-incubated at 37uC for 10 min with 4 pmol 30S subunits, 16 pmol tRNA fMet and Hfq-hexamer (1 or 2 pmol, respectively) before the reverse transcriptase reaction was performed. To test whether CrcZ can abrogate the Hfq mediated repression of amiE translation initiation, 0.05 pmol amiE 2134-+76 mRNA was pre-incubated at 37uC for 10 min with 4 pmol 30S subunits, 16 pmol tRNA fMet and equimolar amounts of Hfqhexamer and CrcZ (1 or 2 pmol, respectively) before reverse transcriptase was added.

Electro mobility shift assays
The amiE 2134-+20 and crcZ 151 RNAs (see above) were dephosphorylated with FastAP thermo sensitive alkaline phosphatase (Thermo Scientific) and subsequently 59-end labeled using [c-32 P]-ATP (Hartmann Analytic) and polynucleotide kinase (Thermo Scientific). The labeled RNAs were gel-purified and dissolved in diethylpyrocarbonate-treated water. Labeled RNA (10 nM) was incubated with increasing amounts of purified Hfq, the Hfq Y25D or Hfq K56A mutant proteins in 10 mM Tris-HCl (pH 8.0), 10 mM MgCl 2 , 60 mM NaCl, 10 mM NaH 2 PO 4 , 10 mM DTT, and 25 ng tRNA in a total volume of 10 ml. Unlabeled RNA was used as competitor as stated in the legend to Figure 5A. The reaction mixtures were incubated at 37uC for 30 min to allow protein-RNA complex formation. The samples were mixed with 4 ml loading dye (25% glycerol, 0.2 mg/l xylencyanol and bromphenol blue) immediately before loading and separated on 4% polyacrylamide gels using Tris-borate buffer. The radioactively labeled bands were visualized with a PhosphorImager (Molecular Dynamics) and quantified with ImageQuant software 5.2.

In vitro translation
In vitro translation was performed with the PURExpress in vitro protein synthesis kit (New England BioLabs). 5 pmol in vitro transcribed amiE Flag mRNA was used in a 12.5 ml reaction. Increasing amounts of purified Hfq protein (as specified in legends of Figures 5B and S4) were added. For competition CrcZ (5 and 10 pmol) or RsmZ (10 pmol) RNA were added. After 1 h of incubation at 37uC, 5 ml of the reaction was mixed with 5 ml protein loading buffer and a western-blot was performed using anti-Flag antibodies as described above.

Esterase assay
Esterase activity was assayed as described previously [60]. Briefly, the cells were harvested by centrifugation and washed in 100 mM potassium phosphate buffer pH 7.2. The substrate (25 ml p-nitrophenyl-caproate dissolved in 5 ml ethanol) was added to 100 ml potassium phosphate buffer (100 nM; pH 7.2) containing MgSO 4 to a final concentration of 10 mM. 1 ml of the test solution and 50 ml of cells were used to determine esterase activity by monitoring the change in absorbance at 410 nM min 21 , which was normalized to the optical density (OD 600 ) of the culture.

Northern-blot analyses
Total RNA of the respective strains as specified in the legends to the Figures was purified using hot phenol. The steady state levels of CrcZ and 5S rRNA (loading control) was determined by Northern-blotting using 4 mg of total RNA. The RNA samples were denatured for 5 min at 65uC in loading buffer containing 50% formamide, separated on a 8% polyacrylamide/8 M urea gel, and then transferred to a nylon membrane by electroblotting. The RNAs were cross-linked to the membrane by exposure to UV light. The membranes were hybridized with gene-specific 32 P-endlabelled oligonucleotides (CrcZ: K3; 5S rRNA: I26; Table S3). The hybridization signals were visualized using a PhosphorImager (Molecular Dynamics).  Figure S2 Hfq binds to amiE mRNA and CrcZ RNA in vivo during mimicked CCR. Strains PAO1hfq-(pMMBhfq Flag ) and PAO1hfq-(pMMB67HE) (control) were grown under conditions of mimicked CCR in BSM medium supplemented with 40 mM succinate, 40 mM acetamide (transcriptional induction of amiE) and 1 mM IPTG (transcriptional induction of hfq Flag ). Then, lysates were prepared and RNAs associated with Hfq were co-immunoprecipitated with Hfq specific antibodies. RNA was extracted from the coimmunoprecipitate and from the remaining supernatant. Equal concentrations were used for RT-PCR with specific primers for amiE, CrcZ and RsmZ (control RNA) as described in Text S1. Lanes 1 and 2, RT-PCR with amiE, CrcZ and RsmZ specific oligonucleotides performed with RNA obtained from the supernatant after coimmunoprecipitation (S; not in complex with Hfq Flag ) and with RNA obtained after CoIP with Hfq specific antibodies (CoIP; in complex with Hfq) in lysates of strain PAO1hfq-(pMMBhfq Flag ). Lanes 3 and 4, RT-PCR with amiE, CrcZ and RsmZ specific oligonucleotides performed with RNA obtained from the supernatant after coimmunoprecipitation (S) and after mock co-immunoprecipitation (CoIP) in the absence of Hfq in strain PAO1hfq-(pMMB67HE).  Figure S7 Hfq is required for CrcZ function. The strains were grown to an OD 600 of 0.5 in BSM medium supplemented with 40 mM succinate and 40 mM acetamide. Then IPTG was added (1 mM final concentration). At an OD 600 of 2.0 the cells were harvested and the b-galactosidase activities were determined. The bars depict the b-galactosidase values conferred by the translational amiE::lacZ fusion encoded by plasmid pME9655 in strain PAO1 harboring either the control plasmid pMMB67HE (blue bar) or the crcZ encoding plasmid pMMBcrcZ (pink bar) (A) and in strain PAO1hfq-harboring either the control plasmid pMMB67HE (red bar) or the crcZ encoding plasmid pMMBcrcZ (pink bar) (B), respectively. The error bars represent standard deviations from three independent experiments. The CrcZ levels (top panel) were determined by Northern-blot analysis. 5S rRNA served as a loading control. (TIF) Figure S8 CCR results in destabilization of amiE mRNA. PAO1 was grown in BSM medium supplemented with 40 mM acetamide (No CCR) or supplemented with 40 mM acetamide and 40 mM succinate (CCR). PAO1hfq-was grown in BSM medium supplemented with 40 mM acetamide and 40 mM succinate (CCR). At an OD 600 of 1.0, rifampicin was added to a final concentration of 100 mg/ml and samples were withdrawn for total RNA extraction at the times indicated. (A) The remaining levels of amiE and 16S rRNA (control) were determined by RT-PCR with oligonucleotides specific for either RNA as described in Text S1. The result from one representative experiment is shown. (B) The amounts of amiE mRNA during CCR in PAO1 (blue squares) and in PAO1hfq-(red triangles), respectively, as well as in PAO1 in the absence of CCR (green diamonds) were normalized to that of 16S rRNA at different times after addition of rifampicin. The results are derived from three independent experiments. Error bars represent standard deviations. The half-life of amiE mRNA was determined with 6.8+/20.4 min in PAO1 and with 10.1+/2 0.5 min in PAO1hfq-during CCR. In the absence of CCR the half-life of amiE mRNA was determined with 11.4+/20.8 min. (TIF) Figure S9 Determination of the cellular Hfq concentration. PAO1 was grown in BSM medium supplemented with 40 mM succinate to an OD 600 of 2.0. ( = 3.260.3 10 9 CFU/ml). The Hfq concentration was determined in triplicate samples of PAO1 cell lysates corresponding to 50 ml of culture (lanes 2-4) using quantitative western-blotting with Hfq specific antibodies. Lane 1, marker protein. Lanes 5-7, 0.1, 0.4 and 0.6 pmol of purified Hfq 6 protein were loaded, respectively. The Hfq 6 concentration per cell was determined as described in Text S1. (TIF) Figure S10 Crc does not affect hfq expression. (A) Determination of the b-galactosidase activity conferred by a translational hfq::lacZ fusion encoded by plasmid pTLhfq during growth in strain PAO1 (blue squares) and in strain PAO1Dcrc (green circles). The strains were grown in BSM medium supplemented with 40 mM succinate. The error bars represent standard deviations from three independent experiments. (B) Levels of Hfq and Crc in PAO1 (wt), PAO1Dcrc (Dcrc) and PAO1hfq-(Dhfq) grown to an OD 600 of 2.0 in BSM medium supplemented with 40 mM succinate. Immunodetection of Hfq, Crc and of ribosomal protein S2 (loading control) was performed as described in Text S1. (TIF)   Text S1 Supporting materials and methods and references. (DOCX)