Bacteria and plasmids used in this study are listed in Supplementary Table S1. E. coli strains were grown at 37°C in Lysogeny Broth (LB). P. aeruginosa strains were grown at 37°C in LB or in Brain Heart Infusion Broth (BHI) or Artificial Sputum Medium (ASM) in flasks at 200 r.p.m.. When required, for E. coli strains the media were supplemented with 10 μg/ml gentamycin, 100 μg/ml ampicillin, 25 μg/ml kanamycin, and for P. aeruginosa strains with 50 μg/ml gentamycin and 300 μg/ml carbenicillin. For monitoring biofilm development in flow-chambers conditions, PAO1 wild-type and PAO1 ΔersA (Ferrara et al., 2015) were chromosomally tagged with green fluorescent protein (GFP) and grown in modified FAB medium (Heydorn et al., 2000) supplemented with 0.3 mM glucose.

ErsA overexpression was obtained from pGM-ersA plasmid (Ferrara et al., 2015) using arabinose 0.2% when required.

Oligonucleotides used in this study are listed in Supplementary Table S2. Translational fusions pBBR1 amrZ::sfGFP, amrZCIS1::sfGFP, amrZΔIS2::sfGFP and amrZCIS1ΔIS2::sfGFP under the P_LtetO-1 constitutive promoter were generated as follows. A DNA fragment of 161 bp including 56 nt of UTR-region and 35 codons of the open reading frame (ORF) of amrZ was amplified by PCR with oligos 1/2 (Supplementary Table S2), digested with NsiI-NheI and cloned into the sfGFP reporter vectors pXG10-SF resulting in the plasmid pXG10-amrZ::sfGFP. Likewise for amrZ::sfGFP, 161 bp including 56 nt of UTR-region and 35 codons of the ORF of amrZ were amplified by PCR with oligos 1/2 (Supplementary Table S2) from pUCIDT amrZCIS plasmid, carrying the synthetic and modified sequence of amrZ, digested with NsiI-NheI and cloned into pXG10-SF to generate the translational fusion pXG10-amrZCIS1::sfGFP. The translational fusion pXG10-amrZΔIS2::sfGFP was generated amplifying a fragment of 119 bp including 56 nt of UTR-region and 21 codons of the amrZ ORF with oligos 1/3 (Supplementary Table S2) digested with NsiI-NheI and cloned into pXG10-SF. AmrZCIS1ΔIS2::sfGFP was constructed amplifying a fragment of 119 bp including 56 nt of UTR-region and 21 codons of the amrZ ORF with oligos 1/3 (Supplementary Table S2) from pUCIDT amrZCIS plasmid. All the fragments from the P_LtetO-1 promoter to the end of the GFP reporter gene, including the different versions of amrZ, were amplified from pXG10-amrZ::sfGF, amrZCIS1::sfGFP, amrZΔIS2::sfGFP and amrZCIS1ΔIS2::sfGFP, using oligos 9/10, digested with ClaI-XbaI and cloned into the low-copy number shuttle vector pBBR1-MCS5 generating the pBBR1-amrZ::sfGFP, amrZCIS1::sfGFP, amrZΔIS2::sfGFP and amrZCIS1ΔIS2::sfGFP, respectively. All the plasmids were then transformed into P. aeruginosa strains as reported previously (Ferrara et al., 2015).

A PrrB1-gfp-a transposon cassette was inserted into the chromosome of PAO1 wild-type and ΔersA by conjugation using pBK-miniTn7-ΩGm as a delivery plasmid carrying the cassette inserted into NotI site as reported previously (Lambertsen et al., 2004).

A quantity of 200 μl of overnight bacterial cultures grown in BHI or ASM and diluted to OD₆₀₀ = 0.01, with the addition of carbenicillin 300 μg/ml and arabinose 0.2% when required, was aliquoted into 96-well peg-lid microtiter plates (Nunclon Delta Surface Cat. No.167008, Nunc TSP Cat. No.445497, Thermo Scientific) as reported previously (Harrison et al., 2010). The plates were incubated at 37°C in aerobic conditions with 100 r.p.m. stirring. After 20 h of incubation, growth was monitored by measuring the OD₆₀₀, and the ability of the P. aeruginosa strains to adhere to the polystyrene peg-lid was tested by crystal violet staining. Briefly, the peg-lid was washed twice with saline solution and then stained with 0.1% crystal violet for 20 min (O’Toole, 2011). Excess of stain was rinsed off by placing the peg-lid in saline solution before to solubilize the dye in absolute ethanol. The optical density of each well was measured at 590 nm. Biofilm formation was expressed in adhesion units as the result of the OD₅₉₀/OD₆₀₀ ratio and statistical analysis were performed using T-Test.

Biofilms were grown at 30°C in flow chambers composed of three individual channels as described previously (Møller et al., 1998). PAO1 wild-type and ΔersA overnight cultures diluted to OD₆₀₀ = 0.01 were inoculated into each flow channel with a small syringe. After 1 h without flow, each channel was supplied with a flow of 3 ml/h of FAB medium with glucose 0.3 mM, using a Watson Marlow 205S peristaltic pump. The mean flow velocity in the flow cells was 0.2 mm/s.

The microscopic analyses were performed using a Zeiss LSM510 confocal laser scanning microscope (CLSM; Carl Zeiss, Jena, Germany) equipped with an Ar/Kr laser and filter sets for GFP detection (excitation, 488 nm; emission, 517 nm). Images were obtained using a 40×/1.3 Plan-Neofluar oil objective.

Simulated shadow projection images and cross sections were generated using the IMARIS software package (Bitplane AG, Zürich, Switzerland).

The experiment was performed in triplicate for each strain acquiring seven random images for each channel every day for 3 days. Thus, 21 images for each time point were employed for the statistical analyses using COMSTAT 2.1 software¹ (Heydorn et al., 2000; Vorregaard, 2008).

Swarming assays were performed using Nutrient Broth (Nutrient Broth n°2 Oxoid) medium plates supplemented with 0.5% glucose and 0.5% Bacto-agar (Difco). Overnight cultures normalized at the same OD₆₀₀ of PAO1 wild-type and ΔersA were spotted on the same plate suitably spaced each other and placed at both 28°C and 37°C for 24 h.

Twitching was performed on LB plates supplemented with 1% Bacto-agar (Difco). The inoculation was performed with a sterile toothpick dipped in the overnight cultures and followed at 37°C for 24 h. Statistical analysis was performed on three independent replicates with GraphPad Prism software.

For RNA-Seq, cultures of wild-type PAO1 and ΔersA strains were grown to early stationary phase (OD₆₀₀ = 2.7) in BHI medium. For each strain, total RNA was extracted from at least two independent biological replicates using Trizol reagent (Thermo Fisher Scientific Inc.) followed by RNA clean and concentrator kit (Zymo Research, Irvin, CA, United States) accordingly to vendors’ protocols. RNA quality was checked using RNA Nano kit on an Agilent Bioanalyzer 2100 machine. Samples with an RNA integrity number (RIN) greater than 9 were used in downstream analysis. Strand-specific sequencing libraries were prepared using 50 ng of mRNA-enriched samples as input for TruSeq stranded mRNA library preparation kit (Illumina) following vendor’s recommendations. Sequencing was performed on an Illumina NextSeq 500 to a depth of 15–20 million reads per sample. After quality filtering, raw reads were aligned using BWA aligner against P. aeruginosa PAO1 genome (NC_002516.2). Read count for gene relative abundance was obtained using HTSeq-count tool from HTSeq package (Anders et al., 2015), while differential expression analysis and statistical analysis were performed as previously described (Peano et al., 2014). RNA-seq data have been deposited in the ArrayExpress database at EMBL-EBI² under accession number E-MTAB-6247.

Total RNA was extracted as reported previously (Ferrara et al., 2012). RNA for RNA/RNA interaction assays was prepared by T7 RNA polymerase transcription of gel-purified DNA fragments. DNA fragments for ErsA RNA and amrZ mRNAs (amrZ, amrZCIS1, amrZΔIS2, amrZCIS1ΔIS2) preparations were amplified from P.aeruginosa PAO1 genomic DNA with oligo pairs 4/5 or 4/6 and 7/8, respectively. The transcription reactions were performed using the Riboprobe^® System-T7 (Promega) with 300 ng of DNA template. DNA probe was 5′-end–labeled with (γ-³²P) ATP and T4 polynucleotide kinase (Promega) according to manufacturer’s instruction. Synthesized RNA was precipitated and resuspended in diethylpyrocarbonate-treated water. Purified RNA was checked by denaturing polyacrylamide gel electrophoresis and quantified using a Qubit Fluorometer.

To assess the ErsA/amrZ mRNA interactions in vitro, the binding reactions were set up as described previously (Ferrara et al., 2015). After the electrophoresis, the membrane was UV-crosslinked and hybridized with a [³²P]-labeled oligo and the radioactive bands were acquired using a Typhoon^TM 8600 variable mode imager scanner (GE Healthcare BioSciences) and visualized with ImageQuant software (Molecular Dynamics).

Non-radioactive EMSA were performed using Mini-Protean^® Electrophoresis System (Bio-Rad) at 4°C and 150 V for 45 min. The gel was stained in SYBR^TM Gold Nucleic Acid Gel Stain diluted in 0.5 × TBE. Images were acquired by Gel Doc^TM XR+ (Bio-Rad) imaging system.

Fluorescence measurements of P. aeruginosa strains carrying the reporters pBBR1-amrZ::gfp were carried out as previously reported (Ferrara et al., 2015). Abs₅₉₅ and fluorescence polarization FP_485/535 were measured in a Tecan Infinity PRO 200 reader, using Magellan as data analysis software (Tecan). GFP activities were expressed in Arbitrary Units (AU) as ratio FP_485/535/Abs₅₉₅. Statistical analysis performed on three individual clones per strain using T-test.

We investigated the effects of deleting the ersA gene on biofilm formation using a semi-quantitative microtiter “peg-lid” assay in Brain Heart Infusion medium (BHI). As shown in Figure 2A, the ErsA deletion resulted in decreased biofilm formation in BHI compared to PAO1 wild-type strain, and the complemented strain carrying the plasmid pGM-ersA produces more biofilm than the ersA deletion mutant strain carrying the pGM931 empty vector (Figure 2B).

To examine the role of ErsA in P. aeruginosa biofilm architecture development, we cultivated the PAO1 wild-type and the ΔersA GFP-tagged strain, in flow-chambers continuously supplied with modified FAB medium supplemented with glucose. Biofilm development stages were followed and visualized daily for 3 days by Confocal Laser Microscopy (CLSM). In agreement with biofilm formation in the microtiter “peg-lid” assays in BHI medium, the PAO1 ΔersA strain developed less biofilm biomass than the wild-type, which showed the mushroom-like structures typical of 3-days old P. aeruginosa biofilms in flow-cells system (Figure 2C). The statistically significant differences in biomass and spatial structure between PAO1 wild-type and ΔersA biofilms were determined by the COMSTAT 2.1 software (Heydorn et al., 2000; Vorregaard, 2008) as represented in Figure 2C. We further noticed the positive influence of ErsA on adhesion and biofilm formation when overexpressed in PAO1 wild-type and ΔersA strains, grown in ASM (Supplementary Figure S1), which is defined to reflect the chemical environment of CF lungs (Sriramulu et al., 2005; Haley et al., 2012).

Motility is crucial in cell-to-cell adherence and attachment in early biofilm stages and it has been suggested an inverse regulation of motility and biofilm during biofilm development (Caiazza et al., 2007; Wang et al., 2014). Several transcriptional and post-transcriptional regulators are involved in these pathways and some of them coordinate both sessile and motile lifestyles (O’Toole and Kolter, 1998; Ramsey and Whiteley, 2004; Shrout et al., 2006; Gloag et al., 2013). To further investigate the involvement of ErsA on these biofilm-related phenotypes, we performed co-swarming, swimming and co-twitching experiments comparing PAO1 wild-type with ΔersA strain. Our results reveal a negative influence of ErsA on both swarming and twitching motility (Figure 3) and the temperature conditions do not affect ErsA regulation on swarming motility (Figure 3A). No differences between PAO1 wild-type and ΔersA mutant strain were observed for swimming motility (Supplementary Figure S2).

Small RNAs are usually involved in post-transcriptional regulation, and the role of ErsA in biofilm development and motility shown in this study, suggested interference with the translation of transcriptional regulators as AmrZ. Thus, to expand the panel of ErsA targets in P. aeruginosa PAO1, and to have a better view of the effect of ErsA activity on the genome-wide gene expression, we performed an RNA-seq experiment comparing PAO1 wild-type to ErsA deletion mutant strains, grown to late exponential phase (OD₆₀₀ of 2.7) in BHI medium. We observed 168 genes (Supplementary Table S3 and the most representative genes listed in Table 1) differentially expressed in the ersA deletion mutant when compared to the wild-type strain. Among the 29 genes upregulated in the ersA deletion mutant we identified genes involved in denitrification and nitrate metabolism (narI, narJ, nirN) as well as type VI and III secretion systems effectors (tssA1, tsi4, tse6).

The majority of genes were downregulated in absence of ErsA (139 genes); the strongest negative effect was observed for narK1 involved in nitrate transport. The other hits with a change of Log₂(FC) ≤-1.5, comprise well described genes involved in biofilm formation and motility (algD, esrC, ppyR, pelCDEFG, roeA), energy and carbon metabolism (prpD, prpC, coxA, coxB), heat-shock proteins (htpG, hslU, hslV, ibpA, dnaK, dnaJ) and phzS involved in pyocyanin production.

In order to investigate the possibility that ErsA regulates biofilm modulating the expression of AmrZ at the post-transcriptional level through direct binding to the amrZ mRNA, we used a plasmid based GFP-reporter system and an electromobility-shift assay for the in vivo and in vitro validation, respectively. Before this, however, we used the full-length ErsA RNA sequence and the amrZ mRNA (including the 5′ untranslated region, 5′-UTR), as inputs in the web tool IntaRNA (Wright et al., 2014) to predict ErsA-amrZ mRNA interactions. The tool identified two putative interaction sites for ErsA on the amrZ mRNA. The interaction site 1 (IS1) involves part of the ErsA U-rich unstructured region, from nt 41 to 52 and is predicted to bind to amrZ mRNA in the region spanning +5 to +14 from the translational starting site AUG (Figure 4A). The ErsA interaction site 2 (IS2) on amrZ mRNA is predicted at positions +65 to +89 and covers a longer region on ErsA unstructured structure, from 26 to 56 nt (Figure 4A).

To test the ErsA post-transcriptional regulation on amrZ mRNA, we generated a translational fusion between the 5′-UTR along with the first 35 codons of amrZ mRNA and the superfolder variant gene of the green fluorescent protein (sfGFP) under the control of the heterologous constitutive promoter P_LtetO-1. This GFP reporter fusion was transformed into P. aeruginosa PAO1 wild-type and ΔersA strains, respectively. As shown in Figure 4B, the ErsA deletion caused a reduction in GFP activity of the amrZ::sfGFP translational fusion compared to the wild-type and it was possible to increase the amrZ::sfGFP translational levels in ΔersA mutant strain by inducing with arabinose the expression of ersA from the pGM-ersA plasmid (Supplementary Figure S3), suggesting a direct effect of ErsA on amrZ translation efficiency. The lack of a full genetic complementation could be explained by the fact that we observed by Northern blot that the ErsA levels expressed by pGM-ersA in a ΔersA strain are lower than those expressed by the chromosomal copy of ersA gene (data not shown). This scenario is different from the one observed for the expression of ErsA from pGM-ersA in a wild-type background where the ErsA levels resulted to be five–sixfold higher than those expressed by the chromosomal copy of ersA gene (Ferrara et al., 2015). This would suggest a higher ErsA degradation in a ΔersA background.

Interactions of ErsA with the GFP ORF were previously controlled using a plasmid carrying exclusively the gfp gene (Ferrara et al., 2015). These results strongly suggested a positive regulation by ErsA on translation of the amrZ gene. This regulation does not depend on Hfq (data not shown). Furthermore, to document the predicted ErsA–amrZ mRNA interaction also in vitro, ErsA RNA and amrZ mRNA were synthesized, mixed and analyzed by electrophoresis on native polyacrylamide gels. As shown in Figure 4C, ErsA specifically formed a complex with the amrZ mRNA.

To further document the specific ErsA-amrZ mRNA interactions, we generated three amrZ mRNA fragments, (i) amrZCIS1, in which the interaction site 1 has been substituted with its complementary sequence, (ii) amrZΔIS2 characterized by the deletion of the interaction site 2 and (iii) amrZ CIS1ΔIS2 containing both the modifications present in amrZCIS1 and in amrZΔIS2. The in vitro analysis showed that ErsA forms a complex with both amrZCIS1 and amrZΔIS2 (Figures 5A,B), and it does not bind to amrZ CIS1ΔIS2 mRNA (Figure 5C). This suggested that both interaction sequences are involved in ErsA-amrZ binding (Supplementary Figure S4). In vitro results were corroborated by in vivo experiments, measuring the translational levels of amrZCIS1::sfGFP, amrZΔIS2::sfGFP and amrZCIS1ΔIS2::sfGFP in PAO1 wild-type and ΔersA strains. The absence of the interaction sites for ErsA causes a reduction of translational fusions activity in both genetic backgrounds (Figure 5D), associated also to a transcriptional instability (data not shown).