Reduced Intracellular c-di-GMP Content Increases Expression of Quorum Sensing-Regulated Genes in Pseudomonas aeruginosa

Cyclic-di-GMP (c-di-GMP) is an intracellular secondary messenger which controls the biofilm life cycle in many bacterial species. High intracellular c-di-GMP content enhances biofilm formation via the reduction of motility and production of biofilm matrix, while low c-di-GMP content in biofilm cells leads to increased motility and biofilm dispersal. While the effect of high c-di-GMP levels on bacterial lifestyles is well studied, the physiology of cells at low c-di-GMP levels remains unclear. Here, we showed that Pseudomonas aeruginosa cells with high and low intracellular c-di-GMP contents possessed distinct transcriptome profiles. There were 535 genes being upregulated and 432 genes downregulated in cells with low c-di-GMP, as compared to cells with high c-di-GMP. Interestingly, both rhl and pqs quorum-sensing (QS) operons were expressed at higher levels in cells with low intracellular c-di-GMP content compared with cells with higher c-di-GMP content. The induced expression of pqs and rhl QS required a functional PqsR, the transcriptional regulator of pqs QS. Next, we observed increased production of pqs and rhl-regulated virulence factors, such as pyocyanin and rhamnolipids, in P. aeruginosa cells with low c-di-GMP levels, conferring them with increased intracellular survival rates and cytotoxicity against murine macrophages. Hence, our data suggested that low intracellular c-di-GMP levels in bacteria could induce QS-regulated virulence, in particular rhamnolipids that cripple the cellular components of the innate immune system.


INTRODUCTION
Pseudomonas aeruginosa can cause opportunistic infections in humans, such as cystic fibrosis lung infections, burn wounds and urinary tract infections (Bodey et al., 1983). This is attributed to its ability to form biofilms and produce an abundance of virulence factors to impair the host immune response Fazli et al., 2011).
Similar to many Gram-negative bacteria species, the biofilm and planktonic lifestyles in P. aeruginosa are controlled by the secondary messenger bis-(3 ′ -5 ′ )-cyclic-dimeric-GMP (c-di-GMP) (Romling et al., 2005). C-di-GMP is synthesized by diguanylate cyclases (DGCs) and degraded by phosphodiesterases (PDEs) (Hengge, 2009). High intracellular c-di-GMP content enhances biofilm formation, whereas low intracellular c-di-GMP content leads to biofilm dispersal and the return to planktonic phase (Hisert et al., 2005;Romling et al., 2005;Kulasakara et al., 2006;Chua et al., 2014;Yu et al., 2015). The redundancy of DGC and PDE genes in the genome confers P. aeruginosa the survival advantage to respond to various stresses from the environment. For instance, the wspR DGC is important in the sensing of reactive oxygen species (ROS) and formation of biofilms resilient to ROS stress (Chua et al., 2016a).
Another system that plays important roles in biofilm formation and virulence is quorum sensing (QS), which is the intercellular communication system positively dependent on cell density and QS autoinducer (AI) concentrations (Fuqua et al., 1994;Whitehead et al., 2001;Ng and Bassler, 2009). P. aeruginosa possesses four major QS systems, encoded by the las, rhl, pqs and iqs systems, with the las and rhl systems employing homoserine lactones, namely the N-(3oxododecanoyl)-homoserine lactone (OdDHL) and N-butanoyl-L-homoserine lactone (BHL) respectively as their AIs (Gambello and Iglewski, 1991;Passador et al., 1993;Ochsner and Reiser, 1995;Pearson et al., 1995), while pqs and iqs sytems using the 2-heptyl-3-hydroxy-4(1H)-quinolone (PQS) and 2-(2hydroxyphenyl)-thiazole-4-carbaldehyde respectively (Cao et al., 2001;Diggle et al., 2007;Lee et al., 2013). The AIs will bind and activate the transcriptional regulators, resulting in the transcription of downstream QS operons. The QS systems interregulate one another, notably the las system and pqs system activate the rhl system McKnight et al., 2000;Farrow et al., 2008). The QS systems control the production of many virulence factors, such as pyocyanin (by the pqs operon) and rhamnolipids (by the rhl operon) Xiao et al., 2006). The rhamnolipids are biosurfactants which are highly cytotoxic to eukaryotic cells, as previously demonstrated by the induction of rhl operon-controlled gene expression in biofilm bacteria exposed to polymorphonuclear leukocytes (PMNs) and subsequent destruction of these important defensive immune cells (Alhede et al., 2009).
While the impact of high c-di-GMP content on biofilm formation is well understood, the consequences of low intracellular c-di-GMP content other than biofilm dispersal remain unclear. Our previous study showed that freshly dispersed cells, during the short-term liberation process, appeared to be highly virulent as compared to biofilm cells (Chua et al., 2014). It remains elusive whether reduced c-di-GMP content may have a long-term impact on bacterial physiology and virulence.
Hence, we aimed to investigate the impact of low vs. high cdi-GMP concentrations on P. aeruginosa virulence mechanisms. We compared the transcriptomes of P. aeruginosa PAO1 cells "locked" in a condition with high c-di-GMP content (by using the wspF mutation to induce constitutive expression of WspR) and the cells "locked" in a condition with low c-di-GMP content (by over expressing the YhJH PDE) cultivated under the similar growth conditions. As the WspF protein is the inhibitor of the WspR DGC, the wspF mutation will cause expression of WspR, thereby promoting the synthesis of c-di-GMP leading to high internal levels (Hickman et al., 2005). The PAO1/p lac -yhjH strain contains the constitutively expressed PDE gene yhjH leading to low internal levels of c-di-GMP, a condition important in swarming and swimming motility (Pesavento et al., 2008;Chua et al., 2013).
We found that low intracellular c-di-GMP content induced expression of the QS systems, specifically the rhl and pqs systems, which led to increased production of several virulence factors, such as rhamnolipids and pyocyanin. This was correlated to increased killing of macrophages. We showed that the induction of rhl and pqs QS under conditions of low c-di-GMP levels, was mediated by PqsR, the transcriptional regulator of pqs QS.
Hence, our present study suggested that c-di-GMP-governed biofilm dispersal might liberate bacteria capable of producing virulence factors, so as to survive and protect themselves from the phagocytic immune cells in the host. Hence, as a strategy to prevent the dissemination of biofilm infections, the use of QS inhibitors (Hentzer et al., 2003) can potentially reduce the production of QS-related virulence factors.

Quantification of c-di-GMP
Bacterial cells in 5 ml ABTGC were harvested and pelleted by centrifugation at 13,000 g for 3 min. The supernatant was removed and the cell pellet was immediately snap-frozen in liquid nitrogen. The cell pellet was re-suspended in 1 ml of acetonitrile/methanol/water (40:40:20) mixture. An aliquot of cells (10 µl) was used for protein quantification. The cells were then lysed with a probe tip ultrasonicator (Amplitude 30%; 5 s ON, 5 s OFF) for 1 min in ice slurry. The cell debris was removed by centrifugation at 13000 g for 3 min. The liquid phase was then evaporated by using the vacuum concentrator, leaving behind the nucleotide precipitate. The samples were then re-suspended in 100 µl ddH 2 O and centrifuged at 10,000 g, 4 • C for 10 min. The solutions were transferred to glass vials and injected through liquid chromatography-mass spectrometry (LCMS).
For the LC portion, the samples in the glass vials were run through the BEH C18 (1.7 µm; 2.1 × 50 mm) column with injection volume of 5 µl at 0.3 ml min −1 for a total runtime of 6 min, with the mobile phase A as 10 mM ammonium formate in water + 0.1% formic acid and mobile phase B as methanol + 0.1% formic acid. For the MS portion, the samples were then analyzed by Xevo TQ-S, Waters mass spectrometer, under the ESI positive ion mode (capillary voltage: 3.8 kV, desolvation temperature: 400 • C). The cyclic di-GMP compound was detected by monitoring ion transition of 691.2 m/z to 152.0 m/z at collision energy 36 eV.
For protein quantification, the cell aliquot was treated in 1 ml of 5 M NaOH at 95 • C for 5 min. After cooling the samples for 15 mins, the proteins were processed with the Qubit R protein assay kit (NanoOrange dye) and quantified by the Qubit R 2.0 Fluorometer (Invitrogen). The concentration of c-di-GMP was then normalized with protein quantity. Experiments were performed in triplicate, and results were shown as the mean ± s.d.

Quantification of PDE Activity
Bacterial cells in 5 ml ABTGC were harvested and pelleted by centrifugation at 13,000 g for 3 min. The supernatant was removed and the cell pellet was resuspended in 5 ml 0.9% NaCl. The cells were lysed with a probe tip ultrasonicator (Amplitude 30%; 5 s ON, 5 s OFF) for 3 min in ice slurry to obtain a crude extract. As previously described (Kuchma et al., 2007), the crude extracts were incubated with 5 ml of 5 mM bis(p-nitrophenyl) phosphate (bis-pNPP) in buffer (5 mM MgCl 2 , 50 mM Tris-HCl [pH 9.3], 50 mM NaCl). The release of p-nitrophenol was quantified by using a microplate reader (Tecan Infinite Series 2000) at OD410 every 15 min for 16 h.
As described in the previous section, protein concentration was determined by the Qubit R 2.0 Fluorometer (Invitrogen).The PDE activity was then normalized with protein quantity. Experiments were performed in triplicate, and results were shown as the mean ± s.d.

RNA Preparation
PAO1, PAO1 wspF and PAO1/p lac -yhjH were grown in 1 ml ABTGC in each well (triplicates) within a 24-well microplate (Nunc) for 7 h till late logarithmic phase in 37 • C, 200 rpm shaking. Bacterial cells were first treated with RNA Protect (Qiagen, Netherlands) and then treated with lysozyme. Total RNA was extracted using RNeasy Mini Kit (Qiagen, Netherlands). On-column DNase digestion with the RNase-free DNase Set (Qiagen) was used to remove DNA. The DNA contamination levels were assessed by using the Qubit R dsDNA High Sensitivity (HS) assay (PicoGreen dye) and the Qubit R 2.0 Fluorometer (Invitrogen). The integrity of total RNA was assessed by using the Bioanalyser RNA analysis kit (Agilent Technologies) and the Agilent 2100 Bioanalyzer (Agilent Technologies). The Ribo-Zero TM Magnetic Kit (Bacteria) (Epicentre) was used to deplete 16S, 23S, and 5S rRNAs from the samples.

RNA Sequencing and Data Analysis
Gene expression analysis of 2 biological replicates was conducted by RNA-Seq technology (Illumina). The RNA was fragmented to 200-300 bp fragments using divalent cations under elevated temperature.
First and second strand cDNA were then synthesized and treated by end repair and adapter ligation. After the 12-cycle PCR enrichment, the quality of the libraries was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies). The libraries were sequenced using the Illumina HiSeq2000 platform with paired-end protocol and read lengths of 100 nt.
The sequence reads were assembled and analyzed by "RNA-Seq and expression analysis" application of CLC genomics Workbench 6.0 (CLC Bio, Aarhus, Denmark). The PAO1 genome (http://www.ncbi.nlm.nih.gov/nuccore/110645304) was utilized as the reference genome. The following criteria were used to filter the unique sequence reads: maximum number of hits for a read of 1, minimum length fraction of 0.9, minimum similarity fraction of 0.8, and maximum number of two mismatches. Genes were annotated with Pseudomonas Genome Database (Winsor et al., 2011). The mapping results of RNA-Seq raw data from CLC genomics Workbench 6.0 were subjected to DESeq2 package for statistical analysis (Anders and Huber, 2010) by reading them into R/Bioconductor (Gentleman et al., 2004). The transcript counts were normalized to the effective library size. Hierarchical clustering analysis was performed and a heatmap was drawn for the 1000 most highly expressed genes of PAO1, PAO1 wspF, and PAO1/p lac -yhjH using heatmap.2 package of R/Bioconductor (Gentleman et al., 2004). Furthermore, the normalized counts were stabilized according their variance as outlined in the DESeq2 package tutorial and a principle component analysis (PCA) plot was generated. The differentially expressed genes among PAO1, PAO1 wspF, and PAO1/ p lac -yhjH were identified by performing a negative binomial test using the DESeq2 package of R/Bioconductor. Transcripts were stringently determined as differentially expressed when having a fold change larger than 5 and an adjusted p-value smaller than 0.05.
Accession number for the RNA-seq is PRJNA381683.

qRT-PCR Analysis
Total RNA from cells grown in 2 ml ABTGC was extracted using RNeasy Mini Kit (Qiagen) with on-column DNase digestion. The concentration and purity of the extracted RNA were measured by NanoDrop 2000 spectrophotometer (Thermo Scientific), while the integrity of RNA was analyzed by Agilent 2200 TapeStation System (Agilent Technologies). The elimination of contaminating DNA was confirmed via real time PCR amplification of the rpoD gene with total RNA as template.
First-strand cDNA was first synthesized from total RNA with the SuperScript R III First-Strand Synthesis SuperMix kit (Invitrogen). The cDNA was used as template for qRT-PCR with a kit of SYBR R Select Master Mix (Applied Biosystems, Life Technologies) on the StepOnePlus Real-Time PCR System (Applied Biosystems, Life Technologies). The gene rpoD was used as endogenous control. To verify specific single-product amplification, melting curves were analyzed.
Quantification of BHL by Using ∆lasI∆rhlI/p rhlA -gfp Reporter Fusion Supernatants (2 ml) from P. aeruginosa strains grown in ABTGC in 37 • C overnight were filtered through 0.2-µm filters, and the filtrates were collected. Overnight culture of the reporter strain lasI rhlI/p rhlA -gfp was adjusted to OD 600 = 0.2 using ABTGC medium. 100 µl of filtrate was added to 100 µl of lasI rhlI/p rhlA -gfp in a 96-well plate (Nunc, Denmark). Because lasI rhlI does not produce BHL, p rhlA -gfp was induced by the addition of serial diluted filtrates containing BHL. GFP fluorescence from p rhlA -gfp expression (expressed in relative fluorescence units, RFU) was measured for each well using a microplate reader (Tecan Infinite 2000) and was normalized to the OD 600 of each well. Experiments were performed in triplicate, and results are shown as the mean ± s.d.
Quantification of PQS by Using ∆pqsA/p pqsA -gfp Reporter Fusion Supernatants (2 ml) from P. aeruginosa strains grown in ABTGC in 37 • C overnight filtered through 0.2-µm filters and the filtrates were collected. Overnight culture of the reporter strain pqsA/p pqsA -gfp was adjusted to OD 600 = 0.2 using ABTGC medium. 100 µl of filtrate was added to 100 µl of pqsA/ p pqsAgfp in a 96-well plate (Nunc, Denmark). Because pqsA does not produce PQS, p pqsA -gfp was induced by the addition of serial diluted filtrates containing PQS. GFP fluorescence from p pqsAgfp expression (expressed in relative fluorescence units, RFU) was measured for each well using a microplate reader (Tecan Infinite 2000) and was normalized to the OD 600 of each well. Experiments were performed in triplicate, and results are shown as the mean ± s.d.

Quantification of Rhamnolipids by Orcinol Assay
Relative amounts of rhamnolipids produced by P. aeruginosa strains were quantified as previously described (Wittgens et al., 2011;Fong et al., 2016). The supernatant of strain grown in 2 ml ABTGC in 37 • C overnight was filtered with 0.2-µm filter and treated with equal volumes of ethyl acetate. Samples were then mixed with vortexing for 30 s, with a phase separation by putting samples briefly in centrifuge for 30 s at 5,000 g. The upper organic phase containing rhamnolipids were transferred to new tube. The organic solvent was then removed by evaporation with a vacuum concentrator. The residue containing rhamnolipids was then re-suspended in 100 µl ddH 2 O. 100 µl of sample was mixed with 100 µl of 1.6% orcinol in ddH 2 O and 800 µl sulphuric acid (60% v/v). The samples were then incubated at 80 • C for 30 mins and mixed periodically. After cooling the samples to room temperature for 15 mins, OD 420 was measured by the microplate reader (Tecan Infinite 2000). Experiments were performed in triplicates and results were shown as mean ± standard deviation.

Quantification of Pyocyanin
Relative levels of pyocyanin produced by P. aeruginosa strains were quantified as previously described (Frank and Demoss, 1959;Fong et al., 2016). The supernatant of strain grown in 2 ml ABTGC in 37 • C overnight was filtered with 0.2-µm filter and treated with equal volumes of chloroform. Samples were then mixed with vortexing for 30 s, with a phase separation by putting samples in centrifuge for 30 s at 5,000 g. The lower organic phase containing pyocyanin was transferred to new tube. 200 µl of 0.1 M hydrochloric acid (HCl) was then added to the chloroform phase and mixed with vortexing for 30 s, with a phase separation by putting samples briefly in centrifuge. The pink coloration which subsequently formed by acidification of pyocyanin, in the HCl phase at the top layer was then transferred to a 96well microplate (Nunc, Denmark) and OD 500 was measured by the microplate reader (Tecan Infinite 2000). Experiments were performed in triplicates and results were shown as mean ± standard deviation.
The cells were checked against Mycoplasma contamination by using the PCR Mycoplasma detection kit (Abmgood, USA) before experiments.

Macrophage Cytotoxicity Assay
To test the ability of cells to kill macrophages, cytotoxicity of macrophages was determined by monitoring cell integrity in 1 ml DMEM + 10% FBS + 20 µM propidium iodide (PI), as previously described (Chua et al., 2016b). Cells that stained by PI under the epifluorecent microscopy (Zeiss) with 20 × objective were observed as dead. The ratio of dead cells to live cells, enumerated from five images (each image contained approximately 200 macrophages), was then calculated. Experiments were performed in triplicates and results were shown as mean ± standard deviation.

Comparing the c-di-GMP Content of PAO1 Cells with High and Low c-di-GMP Levels
We used RNA-sequencing to compare the transcriptomes of P. aeruginosa PAO1 cells "locked" in conditions of either high or low intracellular c-di-GMP contents. The late log phase P. aeruginosa wspF mutant cells possessed a high intracellular c-di-GMP content due to the constitutively expressed WspR DGC protein (D'argenio et al., 2002;Rybtke et al., 2012;Chua et al., 2016a; Figure 1A), contributing to the overproduction of exopolysaccharides and low motility in wspF cells. Hence, we used wspF cells to mimic the biofilm stage. On the other hand, the late log phase P. aeruginosa p lac -yhjH mutant cells contained reduced c-di-GMP content due to the constitutively expressed YhjH PDE protein (Gjermansen et al., 2010;Chua et al., 2013; Figure 1A), thus mimicking cells freshly dispersed from the biofilms. We also corroborated our findings by detecting increased enzymatic PDE activity in the p lac -yhjH harboring cells, as compared to wild type and wspF mutant cells (Figure 1B).

Comparing the Transcriptomes of PAO1 Cells with High and Low c-di-GMP Levels
The wspF and PAO1/p lac -yhjH mutants demonstrated distinct gene expression profiles according to the heat map diagram and PCA analysis (Figures 2A,B). 431 genes were up-regulated and 595 genes were down-regulated in the p lac -yhjH mutant as compared to the wspF mutant (Data Sheet 1), including  genes regulated by QS (in particular regulated by the rhl and pqs encoded systems). Specifically, rhlR, rhlA and rhlB were highly induced in the p lac -yhjH strain compared to the wspF mutant (Data Sheet 1). This finding was validated by qRT-PCR analysis ( Figure 2C) and the rhl QS reporter fusion p rhlA -gfp (Yang et al., 2009; Figure 3A). Accordingly, the p lac -yhjH strain was found to produce more rhamnolipids (Figure 3B), PQS ( Figure 3C) and pyocyanin ( Figure 3D) than the wspF mutant, and was thus more cytotoxic to murine macrophages than the wspF mutant ( Figure 3E). This showed that the reduction of c-di-GMP levels lead to induction of the QS systems, which in turn stimulate production of a subset of QS-controlled virulence factors. Conversely, the wspF mutant with the "mimicked" biofilm phenotype had higher expression of biofilm-related genes important for exopolysaccharide production (pelB, pelC, pelD and PelF) and siderophore production (pvdD, pvdJ, pvdL, pvdO), as compared to the PAO1/p lac -yhjH mutant.

PqsR Induces the pqs and rhl QS
We also noticed that genes from the pqs operon, such as pqsA, pqsB, pqsC, and pqsH were highly induced in cells with low intracellular c-di-GMP content (Data Sheet 1). The qRT-PCR analysis confirmed these findings ( Figure 2C). Since the pqs operon is controlled by PqsR, we next hypothesized that PqsR could be crucial to the induction of pqs QS system under conditions of low c-di-GMP levels. Since the rhl encoded QS system can be induced by the pqs QS system (McKnight et al., 2000), it was also possible that activation of the pqs QS system boosted rhamnolipid production. Furthermore, a pqsR/p lac -yhjH mutant produced lesser rhamnolipids than the p lac -yhjH mutant ( Figure 4A). Accordingly, we showed that a PqsR deficient pqsR/p lac -yhjH mutant expressed only low levels of pyocyanin, BHL and PQS compared with the p lac -yhjH mutant (Figures 4B-D). This correlated well with lower cytotoxicity to macrophages by pqsR ( Figure 4E). Hence, PqsR appeared to be a key regulator for induction of QS systems by low intracellular c-di-GMP content.

DISCUSSION
While most studies focused on the effects of high c-di-GMP levels on biofilm formation, there is a paucity of research on the physiology of cells undergoing conditions of low cdi-GMP signaling. Our previous study had shown that cells freshly dispersed from biofilms contained lower c-di-GMP levels  than planktonic cells and biofilm cells (Chua et al., 2013), implying that dispersed cells possess a different physiology from biofilm and planktonic cells. This raised the question of how differing c-di-GMP levels impact the physiology in P. aeruginosa. As biofilm cells and dispersed cells had high physical and physiological heterogeneity (Stewart and Franklin, 2008), we used the wspF and PAO1/p lac -yhjH mutants to imitate biofilm and dispersed cells respectively, and cultivated them as planktonic cultures which were easy to manipulate in controlled conditions.
In this work, we compared the transcriptomics of cells with high and low c-di-GMP levels. Other than biofilm dispersal, we had shown using transcriptomics that low c-di-GMP levels could lead to the induction of the pqs and rhl QS, with PqsR acting as mediator to activate both QS systems. Although we do not show that conditions of low c-di-GMP mediate increased PqsR, a previous study had shown that RsmA from the c-di-GMP-mediated Gac/Rsm pathway, was important in pqs and rhl QS (Burrowes et al., 2006). Hence, the result of activating both QS systems was the increased production of pyocyanin and rhamnolipids, which were correlated to higher virulence to immune cells. Interesting, it was previously observed that rhamnolipids acted as surfactant to facilitate biofilm cells to disperse from biofilms (Bhattacharjee et al., 2016).
Several research groups, such as ours (Chua et al., 2015;Yu et al., 2015) are currently investigating the possibility of exploiting the lowering of the c-di-GMP content in bacteria and dispersal as a biofilm control strategy. Our study had several implications for clinical and environmental applications of this biofilm dispersal strategy. Firstly, liberated bacterial cells could attain a unique physiological state if the c-di-GMP content is maintained at a lower level than planktonic cells and biofilm cells. This state could be reached after long-term growth of dispersed biofilm cells in the presence of agents that cause biofilm dispersal, thus warranting further studies on the biofilmdispersed cells. Secondly, it appeared that a constitutively low cdi-GMP content renders the bacterial cells highly virulent, which might be essential for dispersed cells to survive the encounter with immune cells and cause development of sepsis.
Hence, it is important to evaluate the potential virulence outcome which applying c-di-GMP mediated biofilm dispersal during the eradication of biofilms, especially in infections. The use of QS inhibitors (Hentzer et al., 2003) can effectively negate the induction of QS pathways and production of virulence factors, to be used concurrently with c-di-GMP-mediated biofilm dispersal.

AUTHOR CONTRIBUTIONS
SL and LY designed methods and experiments, analyzed the data, interpreted the results and wrote the paper. YL designed RNA-seq experiments, discussed analyses, interpretation, and presentation. YYL and HJ performed experiments for qRT-PCR work for analysis and interpretation. GK performed data processing for RNA-seq. SL, TT, LY and MG defined the research theme and discussed project outline. All authors have contributed to, seen and approved the manuscript.