Hfq-assisted RsmA regulation is central to Pseudomonas aeruginosa biofilm and motility

Expression of biofilm and motility genes is controlled by multiple regulatory elements, allowing bacteria to appropriately adopt a sessile or motile lifestyle. In Pseudomonas aeruginosa, the post-transcriptional regulator RsmA has been implicated in the control of various biofilm- and motility-associated genes, but much of the evidence for these links is limited to transcriptomic and phenotypic studies. RsmA binds to target mRNAs to modulate translation by affecting ribosomal access and/or mRNA stability. Here we trace a global regulatory role of RsmA to the inhibition of Vfr - a transcription factor required for efficient production of two other transcriptional regulators, namely FleQ and AlgR. FleQ and AlgR, in turn, directly control flagella and pili genes, respectively. FleQ also directly controls biofilm-associated genes that encode the PEL polysaccharide biosynthesis machinery. Furthermore, we show that RsmA cannot bind vfr mRNA alone, but requires the RNA chaperone protein Hfq. This is the first example where a RsmA protein family member is demonstrated to require another protein for RNA binding. SIGNIFICANCE STATEMENT Microorganisms are subjected to dynamic changes in their environment and rapidly adapt their gene expressions accordingly. Such environmental cues influence planktonic/biofilm and acute/chronic infection modes of growth. In the opportunistic pathogen Pseudomonas aeruginosa, RsmA facilitates this switch by controlling gene expression on the RNA level. In this study, we present evidence that RsmA lies central to several downstream regulatory cascades that govern biofilm formation, flagellar motility, and Type IV pilus-mediated motility - all previously described as part of the RsmA regulon. Importantly, we show that RsmA-binding to the vfr mRNA requires Hfq. These findings have implications for the global regulatory role of RsmA and necessitates a genetic and biochemical re-evaluation of what is currently thought to be the RsmA regulon.

limited to transcriptomic and phenotypic studies. RsmA binds to target mRNAs to 23 modulate translation by affecting ribosomal access and/or mRNA stability. Here we trace a 24 global regulatory role of RsmA to the inhibition of Vfra transcription factor required for 25 efficient production of two other transcriptional regulators, namely FleQ and AlgR. FleQ 26 and AlgR, in turn, directly control flagella and pili genes, respectively. FleQ also directly 27 controls biofilm-associated genes that encode the PEL polysaccharide biosynthesis 28 machinery. Furthermore, we show that RsmA cannot bind vfr mRNA alone, but requires 29 the RNA chaperone protein Hfq. This is the first example where a RsmA protein family 30 member is demonstrated to require another protein for RNA binding. 31

SIGNIFICANCE STATEMENT 33
Microorganisms are subjected to dynamic changes in their environment and rapidly adapt 34 their gene expressions accordingly. Such environmental cues influence planktonic/biofilm 35 and acute/chronic infection modes of growth. In the opportunistic pathogen Pseudomonas 36 aeruginosa, RsmA facilitates this switch by controlling gene expression on the RNA level. 37 In this study, we present evidence that RsmA lies central to several downstream regulatory 38 cascades that govern biofilm formation, flagellar motility, and Type IV pilus-mediated 39 motilityall previously described as part of the RsmA regulon. Importantly, we show that 40 RsmA-binding to the vfr mRNA requires Hfq. These findings have implications for the 41 global regulatory role of RsmA and necessitates a genetic and biochemical re-evaluation of 42 what is currently thought to be the RsmA regulon. 43

INTRODUCTION 45
The opportunistic pathogen Pseudomonas aeruginosa can be isolated from a wide range of 46 environmental niches, in large part owing to its versatile metabolic capabilities. It is also 47 proficient in colonizing various eukaryotic organisms and can cause both acute and chronic 48 infections, with the latter often associated with biofilm-like modes of growth at the sites of 49 infection (1). P. aeruginosa is also highly competitive against other microbial species, with 50 its adaptability and competitive fitness being aided by the production of various secondary 51 metabolites and virulence factors (2, 3). Many of these cellular processes are inter-regulated 52 by multiple regulatory pathways, presumably to provide appropriate response mechanisms 53 to a wide range of environmental cues. One of the central regulators that determine P. 54 aeruginosa behavior is the post-transcription factor RsmA (Regulator of Secondary 55 Metabolite) as evidenced by its null mutant phenotypes linked to motility (4), virulence (5-56 7), biofilm (5, 8), growth (8), and secreted products (4,9). 57 58 RsmA belongs to the RsmA/CsrA family of dimeric RNA-binding proteins (10-12), 59 homologues of which are found across a wide range of both Gram-negative and Gram-60 positive bacterial species (13). RsmA/CsrA proteins generally function as translation 61 inhibitors by binding to the ribosome binding sites of target mRNAs, typically overlapping 62 either the Shine-Dalgarno or the start codon (14). However, in a few cases described, they 63 can also serve as positive regulators by altering the secondary structures of bound RNAs 64 (15,16)  aeruginosa biofilm polysaccharide operon psl without any effect on the level of 73 transcriptional changes between the WT and ΔrsmA strains were actually a consequence of 117 post-transcriptional RNA stability changes. However, quantification of pelA transcripts at 118 progressive time points after blocking de novo transcription revealed no differences in 119 stability between the WT and ΔrsmA strains (Fig. S1B). These results suggest that the 120 observed differences in pel expression likely occur at the promoter activity level. Because 121 the RNA-binding protein RsmA is not a transcription factor, we next turned our attention to 122 the possibility of RsmA affecting expression of a known transcriptional regulator of pel, 123 namely FleQ. As a well-characterized transcriptional activator of pel (25, 31), potential 124 RsmA-mediated alterations in expression of FleQ would be anticipated to affect 125 transcription of the pel genes. Using a similar approach as described for pel above, we 126 found that a transcriptional reporter fusion of fleQ gave lower levels in the absence of 127 RsmA (Fig. S2A). This is consistent with previous data showing pel transcript levels are 128 up-regulated in ΔfleQ backgrounds (25). 129

130
The data above could lend themselves to the interpretation that RsmA serves as a positive 131 regulator of fleQ. While RsmA is more commonly known to be a translational repressor, 132 this possibility is not without precedence. Examples of RsmA/CsrA proteins stimulating 133 translation of target mRNAs include phz2 and moaA in P. aeruginosa (15,16) and flhDC in 134 E. coli (17). While P. aeruginosa FleQ and E. coli FlhD4C2 share no sequence or 135 mechanistic similarities, they are functional counterparts, with both being class I master 136 regulators of flagellar biosynthesis in their respective hosts (32). In addition, flagellar 137 motility was previously found to be positively controlled by RsmA in P. aeruginosa (4), 138 while in E. coli, CsrA has been shown to up-regulate flagellar gene expression through 139 protecting flhDC mRNA from degradation (17). It follows that if RsmA regulated fleQ in 140 an analogous manner, fleQ mRNAs would be hyper-stabilized by RsmA and thus, less 141 stable in its absence. However, no evidence of altered mRNA turnover rate was found 142 between WT and ΔrsmA strains (Fig. S2B). Therefore, we reasoned that the direct action of 143 RsmA may lie even further upstream in a regulatory cascade that involved control of fleQ 144 by Vfr. 145 146 Vfr acts as a transcriptional repressor of the fleQ promoter (33). Therefore, to be consistent 147 with the above phenotypes, RsmA would have to serve as a repressor of vfr. In line with 148 this idea, comparison of vfr transcriptional reporter activities in WT and ΔrsmA strains 149 revealed up-regulation in the absence of RsmA (Fig. 1A). To ensure that the regulatory 150 cascade functioned as anticipated, we also tested the prediction that over-expression of Vfr 151 should ultimately result in increased activity of the pel transcriptional reporter. As shown in 152 Bacterial migration is an important aspect during P. aeruginosa biofilm development as 163 initial attachment to a surface is thought to require flagella and mature microcolony 164 formation has been attributed to Type IV pili functions (36, 37). Thus, fine-tuned regulatory 165 control of motility and biofilm genes is a necessity for effective P. aeruginosa adaptation, 166 particularly when switching between motile and sessile lifestyles. 167 168 Because FleQ is a class I master regulator of flagellar genes (32) and RsmA and Vfr lie 169 upstream of FleQ in the regulatory cascade, lack of either protein should predictively 170 impact flagellar motility, but in opposite ways. As would be predicted, ΔrsmA has reduced 171 motility while Δvfr is hyper-motile ( Fig. 2A). Mutants in vfr have previously been 172 documented to be defective in Type IV pilus-dependent twitching motility (38, 39) due to 173 Vfr's positive regulation of the AlgZR two-component system (40, 41) required for 174 expression of the Type IV pili genes (42,43). Based on the results that RsmA serves as a 175 repressor of vfr ( Fig. 1), we predicted that the over-expression of RsmA and lack of Vfr 176 should have epistatically similar phenotypic effect on twitching motility. As shown in Fig.  177 2, Δvfr and RsmA over-expressing strains were both defective in twitching motility. We infer from the findings above that Hfq-binding is a pre-requisite for RsmA to bind to its 231 target within vfr mRNA. Together with the data in preceding sections, these results lead us 232 to conclude: 1) that the direct binding of RsmA on vfr RNA requires the RNA chaperone 233 protein Hfq and 2) that RsmA is an indirect regulator of biofilm polysaccharide locus pel, 234 flagellar motility and Type IV pili-mediated twitching motility, through its direct action on 235 vfr propagated through a regulatory cascade from Vfr to FleQ and AlgR. 236 237 DISCUSSION 238 In this study, we report that P. aeruginosa RsmA requires the RNA chaperone Hfq to assist 239 its binding to vfr mRNA to initiate a regulatory cascade that ultimately impacts both 240 motility-and biofilm-associated genes. The requirement for Hfq was unexpected, since all 241 other biochemical analysis of RsmA/CsrA family members document unaided binding to 242 their target RNAs (e.g.: 8,[15][16][17][56][57][58][59]. We identified two independent Hfq-binding sites 243 upstream of the RsmA-binding motif, which overlaps the ribosome-binding site of vfr (Fig.  244   3A). Based on a mFOLD-predicted secondary structure, a large hairpin loop base-pairs the 245 identified U-rich and A-rich Hfq-binding sites, resulting in a dsRNA structure that would 246 predictively block both RsmA and the ribosome from accessing the mRNA (Fig. 4A). As 247 depicted in Fig. 4B, the simplest model that arises from our findings is that Hfq-binding is 248 required to open the stem-loop to expose the RsmA-binding site. Subsequent binding of 249 RsmA would directly block the Shine-Dalgarno sequence, preventing ribosome access and 250 thereby inhibiting translation. We also provide evidence that vfr mRNA is hyper-stable in 251 the absence of RsmA (Fig. 1C), indicating that Hfq-and RsmA-bound vfr mRNA may 252 undergo more rapid degradation. It was previously suggested that Hfq may be able to 253 directly recruit RNase E for the degradation of bound RNA in E. coli (60), and it is possible 254 that a similar mechanism causes vfr mRNA instability. 255 256 Due to its histidine-rich C-terminus, E. coli Hfq is a common contaminant of His-tagged 257 proteins purified by nickel affinity chromatography after over-expression in E. coli, and 258 visually undetectable Hfq levels can critically influence in vitro analysis of RNA-binding 259 (51, 61). Because the proteins used in this study were purified from either P. aeruginosa 260 (RsmA) or a Hfq null strain of E. coli (Hfq proteins), our analyses were not complicated by 261 this issue. However, a large majority of analyses performed in past publications use C-262 terminal His-tagged RsmA/CsrA proteins derived from over-expression in E. coli, raising 263 the possibility of E. coli Hfq contamination. Given that Hfq binds vfr RNA with apparent 264 high affinity (Figs. 3C and S3), it is plausible that Hfq-assisted RsmA-binding may be 265 prevalent. Although such a possibility would not alter major conclusions and remains to be 266 experimentally verified, it is provocative that comparison of transcriptomic studies done for 267 RsmA, Hfq, and Vfr regulons in P. aeruginosa identify numerous overlaps of genes that 268 were differentially expressed between WT and the respective null mutants (18, 19, 29, 62, 269 63). Based on our findings, it will be crucial to determine and distinguish direct versus indirect 298 regulatory routes to gain a greater understanding of the RsmA regulon. 299

MATERIALS AND METHODS 301
Bacterial strains and growth conditions 302 Table S1 lists the bacterial strains used in this study. E. coli and P. aeruginosa strains were 303 grown in lysogeny broth (LB) at 37 o C unless specified otherwise. VBMM citrate medium 304 (70) was used for selecting P. aeruginosa post-conjugation. For E. coli, the following 305 antibiotics concentrations were used: 50 μg·ml -1 carbenicillin, 10 μg·ml -1 gentamicin, and 306 10 μg·ml -1 tetracycline. For P. aeruginosa strains: 300 μg·ml -1 carbenicillin, 100 μg·ml -1 307 gentamicin, and 100 μg·ml -1 tetracycline were used. Sucrose counter-selection for plasmids 308 carrying the sacB gene used in P. aeruginosa strain constructions was performed by 309 streaking colonies on LB agar (no salt) supplemented with 10% w/v sucrose. Plates were 310 incubated at 30 o C for 24 hours, after which the counter-selected colonies were confirmed 311 for the loss of antibiotic resistance and mutations confirmed by PCR for double-cross-over 312 genomic mutants. 313 314

Strain constructions 315
Transcriptional fusion constructs were generated in single copy on the chromosome of P. 316 aeruginosa strains via integration into the chromosomal attB site as previously published 317 (8). In brief, promoter regions were PCR amplified using oligonucleotides in Table S2. 318 PCR products were ligated between the EcoRI and BamHI sites of mini-CTX lacZ (Fig.  319   S4). These plasmids were then introduced into P. aeruginosa by conjugation. After double 320 site recombination, plasmid backbones were removed by FLP recombinase, and then strains 321 cured of the pFLP2 plasmid by sucrose counter-selection. 322 323

Motility assays 324
Swimming motility assays were performed essentially as previously described (71). LB 325 plates containing 0.3% Bacto agar were inoculated with overnight cultures with a sterile 326 inoculation needle, ensuring the needle tip was inserted approximately halfway into the 327 agar but not to the plastic petri dish bottom, and incubated for 24 hours at 30 o C. Swim ring 328 diameters were measured for quantitation. For twitching motility, we followed the protocol 329 of Haley et al. (72) by using LB plates supplemented with 5 mg/ml porcine gastric mucin 330 (Sigma-Aldrich). A sterile inoculation needle was inserted through the agar until it touched 331 the plastic bottom. Plates were incubated for 48 hours at 37 o C, then at room temperature for 332 48 more hours. Subsequently, the agar was peeled off, and the plates were stained with 1% 333 w/v crystal violet to visualize the twitching zone diameters prior to measurement (73). All 334 motility experiments were performed in biological quadruplicates. 335 336 β-galactosidase assays 337 Quantitative β-galactosidase activities were assayed using Galacto-Light Plus kit (Thermo-338 Fisher) as previously published (74). All P. aeruginosa cultures were grown in VBMM 339 citrate at 37 o C to exponential phase, and lysed using chloroform as previously described 340 (8). β-galactosidase activity units were normalized to total proteins per ml as determined 341 using Bradford assay reagents (Bio-Rad). Assays were performed in biological triplicates. 342 343

Quantitative real-time PCR and RNA stability analyses 344
Real-time PCR was performed as previously described (30), using the oligonucleotides 345 listed in Table S2. For RNA stability experiments, exponential phase P. aeruginosa 346 cultured in VBMM citrate at 37 o C were treated with 200 μg·ml -1 rifampicin (75). RNAs 347 were extracted from 1 ml of the cultures at various time points as previously described (8)  348 after the addition of rifampicin. RNA extractions were performed using the RNeasy Mini 349 Kit (Qiagen) after treating the harvested cells with RNAprotect (Qiagen). Genomic DNA 350 was removed using DNase I (Promega) and removal confirmed by PCR using primers 351 designed against the rplU gene (Table S2)