The Single-Nucleotide Resolution Transcriptome of Pseudomonas aeruginosa Grown in Body Temperature

One of the hallmarks of opportunistic pathogens is their ability to adjust and respond to a wide range of environmental and host-associated conditions. The human pathogen Pseudomonas aeruginosa has an ability to thrive in a variety of hosts and cause a range of acute and chronic infections in individuals with impaired host defenses or cystic fibrosis. Here we report an in-depth transcriptional profiling of this organism when grown at host-related temperatures. Using RNA-seq of samples from P. aeruginosa grown at 28°C and 37°C we detected genes preferentially expressed at the body temperature of mammalian hosts, suggesting that they play a role during infection. These temperature-induced genes included the type III secretion system (T3SS) genes and effectors, as well as the genes responsible for phenazines biosynthesis. Using genome-wide transcription start site (TSS) mapping by RNA-seq we were able to accurately define the promoters and cis-acting RNA elements of many genes, and uncovered new genes and previously unrecognized non-coding RNAs directly controlled by the LasR quorum sensing regulator. Overall we identified 165 small RNAs and over 380 cis-antisense RNAs, some of which predicted to perform regulatory functions, and found that non-coding RNAs are preferentially localized in pathogenicity islands and horizontally transferred regions. Our work identifies regulatory features of P. aeruginosa genes whose products play a role in environmental adaption during infection and provides a reference transcriptional landscape for this pathogen.


Introduction
The outcomes of the broad spectrum of diseases caused by the opportunistic pathogen Pseudomonas aeruginosa are determined by several factors related to the predisposing conditions of the infected individual [1], the mechanism of entry [2] and consequences of interaction with the innate defense mechanisms [3]. The versatility of this organism in adapting to a wide range of environments including that of the human body is attributed to a large genomic repertoire consisting of a diverse set of genes encoding metabolic functions suited for proliferation in environments with a wide range of available nutrients [4][5][6]. Moreover, P. aeruginosa has an impressive armament of virulence factors, explaining why this species is the only known human pathogen among the members of the genus Pseudomonas [7,8]. Interestingly, virulence genes are highly conserved among different P. aeruginosa strains and are part of the core genome, leading to speculations that, in addition to rare infections of humans, some form of pathogenic interaction occurs in natural environments providing the evolutionary pressure for their maintenance [9,10].
Rapid adaptation of bacteria to changing environments is accompanied by reprogramming of their regulatory networks to activate the expression of genes essential for their survival in the new environment while repressing those that are unnecessary or potentially deleterious [11]. This can be accomplished at the transcriptional level by responding to inputs from various environmental cues that are often mediated by specific signal transduction pathways [12]. Recent advances in high-throughput sequencing approaches allow more accurate quantification of RNA levels in bacteria (RNA-seq) providing significant advances over microarrays [13,14]. RNA-seq can provide complete coverage of protein-coding genes and intergenic regions to a single nucleotide resolution, and, with adaptation in the library preparation protocols, also allows strand-specific mapping of transcription start sites (TSSs) [15][16][17][18]. Driven mainly by the data generated from RNA-seq, a number of new regulatory mechanisms were uncovered based on activities of non-coding RNAs [19]. RNA-based regulation (riboregulation) is now recognized as an important mechanism for controlling gene expression by altering the translation of mRNA and/or modulating transcript turnover [20]. Such regulation is performed by trans-acting small regulatory RNAs (sRNAs), cis-antisense transcripts (asRNAs), and riboswitches. Not unexpectedly, virulence gene expression is also subject to riboregulation [19,[21][22][23].
In addition to small molecule chemical signals, such as nutrients, bacteria also sense and respond to environmental cues, such as the availability of oxygen, altered osmolarity, or pH [24,25]. Facultative pathogens are also exposed in their natural environments to temperatures that are usually lower than the relatively constant temperature of mammals, 37uC [26]; exposure to this temperature could represent an important signal of environmental change, and trigger expression of specific virulence determinants. Modulation of gene expression by the temperature experienced by microorganisms has been described for a number of facultative pathogens [27,28]. Moreover, pathogens with a more restricted lifestyle, involving different hosts, can display differential expression of genes depending on the host. For example, a number of proteins encoded on the virulence plasmid of Yersinia pestis are expressed at low temperature during growth in fleas, while a different set of proteins is induced at 37uC during the human infection cycle; each set of these proteins is required for growth and survival during a particular stage of infection [29]. Interestingly, bacteria lacking a reservoir outside of the human host can also display temperature-regulated gene expression [30], possibly due to colonization of exposed surfaces such as skin, or during episodes of fever.
Here we present the results of a global transcript analysis using RNA-seq to generate a high resolution map of transcription start sites and identify mRNAs and non-protein coding transcripts (sRNAs and asRNAs) for P. aeruginosa PA14. We compare transcript abundance in bacteria grown at the mammalian body temperature (37uC) and an arbitrarily-selected reduced temperature (28uC). Our results present several unexpected findings in regards to temperature and quorum sensing control of gene expression. Moreover, our data provide a useful tool for studies of transcriptional and post-transcriptional regulation of gene expression in P. aeruginosa by identifying transcriptional units, accurately mapping transcription start sites and identifying sRNA and asRNA transcripts.

Results/Discussion
A number of P. aeruginosa virulence genes are regulated by temperature We used RNA-seq to examine the global gene expression of P. aeruginosa PA14 at 37uC and 28uC, with an objective to assess whether virulence genes respond to alterations in temperatures encountered in different environments (Table S1). Using replicate experiments, we identified a total of 144 genes that showed statistically significant over-expression of transcripts at 37uC, and 234 genes whose mRNA levels were reduced at 37uC compared to 28uC (Table S2). These represent transcripts of genes that may be activated or repressed during infection of warm-blooded mammals and possibly respond to temperature as one of several input signals for their coordinated expression. Therefore, the levels of transcripts for 6.4% of the genes of the P. aeruginosa PA14 genome were affected by temperature. Gene ontology (GO) enrichment analysis showed significant enrichment for genes related to protein secretion, phenazine biosynthesis and regulation of alginate biosynthesis among the genes upregulated at 37uC (p = 5.1610 26 , 4610 23 , and 6610 23 respectively; Materials and Methods). Elevated expression of the regulatory gene cluster (mucABCD) leads to repression of alginate production. However, this effect can be overcome in isolates from cystic fibrosis patients in which common mutations in mucA lead to overproduction of the alginate exopolysaccharide. At 28uC, we observed upregulation of ribosomal protein genes, histidine catabolic genes, and aromatic and catechol-related genes (p = 6.4610 212 , 2.9610 23 , 1610 23 , and 4610 23 , respectively). We found a strong enrichment for virulence-related genes among those upregulated at 37uC (hypergeometric test p = 8.6610 215 ). Indeed, the expression of 43 out of 238 annotated virulence genes was altered in response to the temperature shift, with most of these (32, 74%) being upregulated at 37uC (Figure 1; Materials and Methods). The vast majority of the genes, 94%, were transcribed only from the strand containing the ORF, thus allowing quantification and comparison of the gene expression with strand-insensitive sequencing.
At the two temperatures used in our transcriptome analysis, we observed few significant changes in the expression of cold-response (capB and PA14_05960) or heat-shock genes (htpG), and did not find evidence for extensive cold-or heat-shock response that is indicative of thermal stress. This suggests that P. aeruginosa has evolved to thrive under a range of conditions that may include a variety of environmental temperatures without the need to activate protective survival responses.
The type III secretion system is preferentially expressed at 37uC The mRNA levels for the majority of the T3SS components (regulators, secretion machinery and effectors) were higher at 37uC (Figure 2A) suggesting that this important P. aeruginosa virulence mechanism is subject to temperature regulation. Since the RNAseq was carried out in calcium replete media, which is not optimal for the expression of the T3SS genes, we analyzed the production of the secretin component PscC by Western blots in two different strains, PA14 and PAO1, grown on calcium-depleted media, which promotes T3SS expression ( Figure 2B). In both strains we observed immunoreactive PscC at 37uC but not at 28uC. Similarly, we observed temperature-regulated production of species-specific effectors of the T3SS, ExoS and ExoU in PAO1 and PA14, respectively, where they were detected only at 37uC

Author Summary
Identifying coordinately regulated genes and their control by environmentally-initiated signal transduction pathways is important for understanding bacterial virulence mechanisms. The work reported here provides a comprehensive, high resolution, transcriptome map of the opportunistic pathogen Pseudomonas aeruginosa using RNA-seq. The results suggest that P. aeruginosa senses the temperature during the transition from its natural environment to a mammalian host, and this plays a key role in regulating the coordinated expression of several virulence factors. A large number of antisense transcripts and non-coding RNAs were identified, with preferential clustering in the regions acquired through horizontal gene transfer, suggesting that a part of the non-coding genome has a distinct evolutionary origin. We created an online data viewer, the Pseudomonas transcriptome browser, to facilitate access to the transcriptome data from this study as well as the subsequent results of work deposited by other investigators. The resources generated through our analyses provide a valuable tool to the P. aeruginosa research community and set the foundation for a systems biology approach towards understanding the complexity of the regulatory networks controlling the multiple lifestyles of this highly versatile organism.
( Figure 2B), confirming that the temperature regulation is not a PA14-specific trait.
The observed regulatory effect of temperature on the T3SS suggests that this virulence mechanism should play an important role in P. aeruginosa infections of warm-blooded mammals. The importance of the T3SS for acute human nosocomial infections has been demonstrated and T3SS-deficient P. aeruginosa strains are attenuated in murine models of pneumonia, bacteremia, keratitis and burn wound infections [31]. However, the T3SS has been shown to contribute to virulence in models where the hosts were infected at their optimal growth temperatures, such as Drosophila melanogaster [32], Acanthamoeba castellani [33], Galleria mellonella [34] and more recently, Danio rerio (zebrafish) embryos [35,36] although this virulence mechanism appears non-essential for the infection of adult gnotobiotic fish [37]. P. aeruginosa can also infect Caenorhabditis elegans, although its T3SS is expressed but not required for virulence [38]. Since the T3SS contributes to the virulence of P. aeruginosa even at reduced temperatures, it is conceivable that the environment of specific hosts provides additional signals that overcome those provided by the temperature or other environmental factors. Previous work has shown that contact between P. aeruginosa and mammalian cells can trigger the expression of T3SS even in the presence of calcium concentrations that in vitro inhibit the expression of these genes [39].
Effect of growth temperature on the expression of genes encoding phenazine biosynthesis enzymes The second group of genes showing a significant increase in mRNA levels encodes enzymes responsible for the biosynthesis of phenazines, which are well characterized virulence factors of P. aeruginosa [40]. Phenazines are nitrogen-containing heterocyclic secondary metabolites that serve as signaling molecules influencing gene expression during environmental adaptations including biofilm formation [41,42]. Phenazines are capable of producing reactive oxygen species toxic to eukaryotic cells and other bacteria [43,44]. These molecules are also involved in electron shuttling to Figure 1. Expression of virulence-related genes at different temperatures. The relative expression of genes known to be related to virulence (downloaded from the Virulence Factors Database [89]) was measured and compared at 28uC and 37uC by RNA-seq. Expression is represented as sequencing read count normalized by gene length and library size for our duplicate samples. Each dot represents a single gene, which is color-coded according to its virulence-related function. The dashed gray lines represent a 2-fold overexpression in 37uC (upper line) and 28uC (lower line) conditions. doi:10.1371/journal.ppat.1002945.g001 alternate terminal acceptors particularly during anaerobic growth [45]. Moreover, the production of phenazines in other Pseudomonas species is temperature regulated [46]. In P. aeruginosa, two unlinked gene clusters, phzA1-G1 and phzA2-G2 encode enzymes catalyzing the synthesis of the core molecule phenazine-1-carboxylic acid (PCA) [47]. Although levels of transcripts from both of the phz clusters were elevated at 37uC, the temperature effect was more pronounced for phzA1-phzG1 ( Figure 2C). Two additional genes located adjacent to the phzA1-G1 operon encode enzymes involved in the modification of PCA: PhzM (a methyltransferase) and PhzS (a monooxygenase). These two enzymes modify PCA to give pyocyanin, while PhzS alone can convert PCA to 1-hydroxyphenazine [48]. The levels of phzM and phzS mRNAs were also increased at 37uC. Finally, the P. aeruginosa chromosome contains the phzH gene at an unlinked site; it encodes the enzyme for the conversion of PCA to phenazine-1-carboxamide. However, we did not detect significant expression of this gene at either temperature, and, therefore, it is unlikely that this modification step takes place in P. aeruginosa under the conditions tested.
Phenazine is synthesized by a pathway utilizing metabolites that are also precursors for aromatic amino acids, TCA cycle intermediates and other bioactive molecules [49]. We therefore  [78] (Materials and Methods) for each gene in the two growth temperatures is represented by bars for each gene (green and red, 37uC and 28uC, respectively). The log 2 -ratio for 37uC and 28uC expression is represented by a diamond for each gene [red diamonds represent a statistically significant difference (x 2 , p,0.05) in expression for two replicates while genes marked with blue diamonds did not pass the significance threshold]. Genes connected by a line between their representing diamonds are part of the same transcriptional unit (TU). Error bars represent a single standard deviation from the mean. Horizontal red arrows represent TUs and their strand-orientation (right and left, forward and reverse strand, respectively). (B) Analyses of expression of the secretin and exoproteins of the T3SS in strains PA14 and PAO1. Cellular (top panel) and secreted (middle and lower panel) proteins from cultures growing at 37 or 28uC (labeled 37 or 28 respectively) were separated by SDS-PAGE followed by Western immunoblotting using antibodies against PscC (a-PscC), ExoU (a-ExoU), or ExoS (a-ExoS). The migration of molecular protein standards are indicated in kilodaltons. The common band on all blots probed with the a-PscC antibody represents a cross-reactive protein. (C) Expression of genes for the phenazine biosynthesis operons. The normalized number of sequencing reads mapped to the genes (black arrows) in the phenazine biosynthesis operons is shown in 37uC and 28uC (red and blue lines respectively). The TSSs that were detected in the operons are shown as black vertical lines (arrowheads representing the strand of the transcription, right and left, forward and reverse, respectively; numbers represent the genomic position of the TSS). A palindromic LasR binding site was identified 52 nt upstream to the TSS of the phzA1-G1 operon and 54 nt downstream of the TSS of phzM. (D) Expression of genes for enzymes in pathways that contribute precursors to the biosynthesis of phenazines. doi:10.1371/journal.ppat.1002945.g002 examined the levels of transcripts for enzymes of the entire pathway and its branches, starting from condensation of erythrose-4-phosphate and phosphoenol pyruvate to various end products ( Figure S1). Genes found in several operons encoding enzymes of these pathways were significantly more expressed (ranging from 6.6-to 35-fold) at 28uC than 37uC. Specifically, antABC, the genes for enzymes catalyzing the synthesis of catechol from anthranilic acid, and genes encoding CatABC and PcaDIJF, responsible for the conversion of catechol to the TCA cycle intermediates succinate and acetyl-CoA, were preferentially expressed at 28uC when compared to 37uC ( Figure 2D). We also detected temperature regulation of the antR gene encoding the transcriptional activator of the ant operon, which was not the case for mRNA levels of catR, the regulatory gene of the cat operon. Although numerous regulatory inputs are involved in regulating these genes at the transcriptional and post-transcriptional levels [50], appears that the conversion of anthranilate to TCA cycle intermediates may occur less effectively at 37uC than at 28uC, making more precursors available for other biosynthetic pathways.
Our results therefore show that at 37uC there should be a significant increase in the levels of two metabolic intermediates affecting the synthesis of important signaling and virulenceenhancing molecules ( Figure S1). An increase in the concentration of chorismic acid would be the consequence of higher levels of PhzC at 37uC and constitutive, temperature-independent transcription of genes encoding enzymes of the shikimate pathway (AroF1BQ1EKAC). Chorismic acid can be also converted to anthranilic acid, which would accumulate at 37uC, in part because of a decrease in its flow through the breakdown pathway to the TCA cycle intermediates succinate and acetyl-CoA, resulting from lower levels the enzymes (CatABC, PcaDIJF). An important consequence of this redistribution of various metabolic intermediates at 37uC is the diversion of chorismic acid towards the biosynthesis of phenazine. Anthranilic acid is also a precursor for the biosynthesis of the quorum sensing molecules alkyl-4 quinolones (AQs). Therefore, the reduction of the levels of the ant, cat and pca transcripts at 37uC (and by inference, levels of corresponding catabolic enzymes) could lead to an increase of the production of 2-heptyl-4-quinolone (HHQ) and 2-heptyl-3-hydroxy-4-quinolone (PQS), two important regulators of P. aeruginosa cell-to-cell communication and virulence gene expression [51,52]. We tested these predictions directly by comparing the levels of pyocyanin and PQS in P. aeruginosa supernatants of cultures grown at 28uC and 37uC ( Figure 3). We found increased concentrations of both of these molecules at 37uC, although the effect on PQS production was much more pronounced when cultures reached the stationary phase of growth. Moreover, the pyochelin siderophore pathway also benefits from the altered levels of the various precursors of the aromatic amino acid pathway, since increased levels of chorismic acid as the consequence of a decrease in its flow towards the TCA cycle makes more of it available for the synthesis of pyochelin, although mRNA concentrations encoding the enzymes of this pathway were not affected by temperature.
The temperature-dependent redirection of transcripts encoding the components of the aromatic amino acid biosynthetic machinery, towards the synthesis of phenazines and signaling molecules such as AQs, provides new insight into the role of environmental modulation of global survival strategies of P. aeruginosa. Although our study focused on the influence of temperature on transcript levels, the precise molecular mechanisms that accomplish thermoregulation are unclear. However, the results of this RNA-seq study provide a basis for the design of rational genetic and biochemical experiments to probe the molecular details of signal transduction, gene expression and protein function at different temperatures that are undoubtedly coordinated with other physical or chemical environmental inputs.
The transcriptome structure of P. aeruginosa PA14 Understanding the transcriptome structure and operon organization in bacteria is essential for understanding bacterial RNAbased regulation [19]. To produce a detailed transcriptome structure map of P. aeruginosa, we used the mapping of the RNAseq reads, which provides transcript coverage in a non-strandspecific manner. To complement this approach, we also used a strand-specific 59-end sequencing method, which identifies active transcription start sites (TSS) at a single nucleotide resolution across the entire genome [15,17]. The combination of the two sequencing methods provided a comprehensive view of the P. aeruginosa PA14 transcriptome in an unbiased manner.
We were able to map transcription start sites (TSSs) for 2,117 transcriptional units (TU), spanning 3,325 protein-coding genes (55% of all protein coding genes, Table S3). A total of 1,854 genes (56%) were found to be included in polycistronic TUs ( Figure  S2A). Over 61% of the multi-gene operons are bicistronic, and only 19% contain 4 or more genes. This general operon organization is highly similar to that found in other bacteria. For example, in the Gram-positive Listeria monocytogenes, in which the operon structure was determined by hybridization of RNA to tiling-arrays, it was found that 60% of the genes were transcribed in multi-gene operon structures, with about half of the multi-gene TUs transcribed as bicistronic mRNAs [53]. In Geobacter sulfurreducens, polycystronic mRNAs accounted for approximately 50% of all transcripts [54] and a recent determination of Escherichia coli K12 found that 35% of the TUs are polycistronic [55]. Since these organisms vary greatly in genome size, growth conditions and physiology, there appears to be a general design principle for bacterial operon organization, which has no correlation with the genome coding capacity. We note that TU structures can be flexible and may change when conditions change [56,57]; however, since in this study we tested only two growth conditions, a global analysis of alternative TU structures could not have been performed.

Definition of transcriptional units and the analysis of 59 untranslated regions (59 UTRs)
The capacity of 59 UTRs to regulate transcriptional and posttranscriptional processes in cis has been demonstrated in numerous organisms [58]. Until now, the mapping of 59 UTRs has not been performed systematically on P. aeruginosa, thus limiting the discovery of new cis-regulatory elements. Since our TSS mapping defines the 59 ends of TUs, it also reveals the 59 UTRs of the immediate downstream genes. Inspection of the 59 UTRs for the 2,117 TUs we defined showed a median 59 UTR length of 47 nt ( Figure S2B), similar to the 59 UTR lengths reported in E. coli (20-40 nt) [59] and Synechococcus elongatus (30 nt) [18]. Most of the 59 UTRs (77%) were shorter than 100 nt, with 60 genes completely lacking 59 UTRs. Interestingly, 115 59 UTRs were longer than 200 nt, which suggests that a significant fraction of these leader regions may function as cis-regulatory RNA elements. We scanned the sequences of all 59 UTRs longer than 100 nt using RFAM, and found that 13 of those contain known riboswitches and RNAleaders (Table S3). Since most of the well-characterized riboswitches models derive from highly divergent bacteria, such as Bacillus subtilis, the presence of these long 59 UTRs suggests that many post-transcriptional regulatory sequences in the P. aeruginosa are yet to be discovered.
The availability of genome-wide TSS maps can allow the correction of false computational gene annotations [19]. Indeed, we were able to correct the annotation of 46 genes (Table S3).

LasR-regulated genes
Since gene annotation in P. aeruginosa is largely limited to the description of ORFs, a genome-wide characterization of promoter regions and binding sites of regulators of transcription is often difficult. The availability of a comprehensive list of TSSs in P. aeruginosa now allows an accurate search for promoter and regulatory regions for specific genes or operons. To demonstrate the utility of this approach, we selected LasR, the transcriptional regulator of quorum sensing and virulence genes. LasR independently regulates the expression of a number of genes. It is positioned at the top of a regulatory network that includes additional quorum sensing systems such as Rhl, PQS and Qsc and, therefore, its regulon is extensive [60][61][62]. The number of genes directly or indirectly controlled by LasR was shown to include over 300 genes or about 6% of the genome [63,64]. A recent chromatin immunoprecipitation coupled to microarray hybridization study (ChIP-chip) [65] has characterized 35 LasR binding sites in the P. aeruginosa PAO1 genome, but the lack of TSS information prevented accurate localization of the LasR sites relative to the transcript. By analyzing the TSSs of genes marked as LasR-bound by the ChIP-chip study, we were able to establish that this transcription factor has a preferential positioning of 51-52 bases upstream to the start of transcription ( Figure 4A). Based on this analysis, we computationally identified 17 new, previously unrecognized putative LasR binding sites on the P. aeruginosa PA14 genome (Table S4; Materials and Methods). An additional 20 sites, previously shown to be bound by LasR in strain PAO1 [7,64,65] were also identified in the PA14 strain. The vast majority of the sites recognized by LasR were immediately adjacent to the 235 element of the promoter, making it likely that it contacts RNA polymerase during transcription initiation ( Figure 4B). Two genes, PA14_21030 (ATP-dependent Clp protease subunit) and PA14_16250 (lasB, previously identified by Gilbert et al [65]) contained multiple LasR binding sites at their promoters ( Figure 4A).
To confirm that the new LasR binding sites identified by our computational methods represent direct targets of regulation by this transcription factor, we analyzed these promoters for direct binding by LasR using electrophoretic mobility shift assays (EMSAs) and for LasR-dependent expression using lacZ transcriptional fusions. When purified LasR was used to analyze DNAprotein interactions, we detected binding of this protein to all but one fragment ( Figure 4C). Based on the low concentrations of LasR (0.3 pM) that were required to give a complete or nearcompete shift, relatively high affinity binding was detected for at least five DNA fragments corresponding to the promoter regions of PA14_09480, PA14_09490, PA14_33830, and PA14_48530. For the rest of the DNA fragments, the extent of binding of LasR varied and in a number of cases, only a weak shift was observed even at the highest concentrations (30 pM) tested. No binding of LasR to the promoter region of PA14_23220 was detected.
To determine whether the newly identified LasR-dependent binding sites indeed respond to LasR in vivo, we have cloned the same fragments used in the EMSAs into a transcriptional lacZ reporter plasmid and incorporated the gene fusion into the chromosome of wild-type P. aeruginosa PA14 and its isogenic lasR mutant. LasR-dependent lacZ expression was assessed at two time points corresponding to early stationary phase (8 hr growth) and late stationary phase (12 hr growth). In general, the levels of expression for most of the fusions, and correspondingly LasR dependence, were higher at later stages of growth ( Figure S3). A number of fusions were poorly expressed in P. aeruginosa, and because of this, no LasR-dependent regulation was observed for the fusions transcribed from the promoters of PA14_03490, PA14_23220, PA14_18800, and PA14_33890. This lack of expression is consistent with the transcriptome data (RPKM values of 30 or less for these genes) suggesting they are not expressed under these conditions. Interestingly, we observed LasR binding to three of these promoters (PA14_03490, PA14_18800, and somewhat weakly, PA14_33890) in our in vitro assay. Two conclusions could be drawn from these results. First, the predicted LasR binding sites adjacent to promoter regions do not necessarily indicate that a particular gene is expressed under all conditions. For example, PA14_03490, PA14_18800, and PA14_33890 could be expressed in environments that are different from these used in our work (LB, aerobic and at 37uC). Moreover, full transcriptional activation of certain LasR-regulated genes could require the input of additional regulatory factors. This regulatory complexity of quorum sensing would permit a more fine-tuned response of groups of genes to specific environmental inputs.

Regulation of phenazine biosynthetic operons
The P. aeruginosa PA14 genome contains two temperature regulated gene clusters, encoding the enzymes for the synthesis of phenazine-1 carboxylic acid (phzS, phzA1-G1, phzM, and phzA2-G2; Figure 2C). Regulation of the phzA1-G1 operon and phzM by acylhomoserine-lactones (HCLs) has been shown previously using DNA microarrays [63,64]. Using the precisely-mapped TSSs, we independently identified in the phzA1-phzM intergenic region a conserved palindromic LasR consensus binding sequence CTAC-CAGATCTTGTAG, with the C positioned 52 nt from the site and on the opposite strand, the sequence CTACAAGATCTGG-TAG with the C 54 nt from the TSS of phzM ( Figure 2C, 5A). Since the consensus LasR binding site is palindromic, the binding of this transcription factor to a single site in an intergenic region could allow activation of the two divergent promoters, a rather uncommon arrangement of regulatory sequences bound by positive transcription regulators in bacteria. Moreover, since the LasR binding site resembles the predicted binding site for a second quorum sensing regulator RhlR [61,63], and LasR regulates the expression of rhlR, it is conceivable that the same palindromic double stranded DNA sequence serves as a binding site for either LasR or RhlR, or both for these two transcriptional units.
Promoter-lacZ fusion experiments in wild-type, lasR, and rhlR mutants of P. aeruginosa showed strong LasR-and RhlR-dependent expression of the lacZ reporter fused to the phzA1 promoter fragment ( Figure 5B). Similarly, the phzM promoter directed transcription in a LasR-and RhlR-dependent manner, although the fold decrease of transcription of the fusion in lasR and rhlR mutants was not as large as seen for the phzA1-lacZ fusions, owing to its high level of expression in wild-type P. aeruginosa. There was only modest regulation by LasR or RhlR for the phzS and phzA2 fusions ( Figure 5B).
Because LasR controls the expression of rhlR, it was impossible to differentiate which of the two quorum sensing transcription factors regulated the expression of phzA1 or phzM. We therefore engineered phzA1-lacZ and phzM-lacZ promoter fusion plasmids and the expression of the reporter gene was measured in response to N-(3-oxododecanoyl) homoserine lactone (3O-C12-HSL) or Nbutyryl homoserine lactone (C4-HSL), in E. coli, co-expressing LasR or RhlR on an inducible plasmid. DNA fragments containing the promoter regions of rsaL and rhlA were used as controls for specificity of expression of activation of the reporter constructs by LasR and RhlR, in response to 3O-C12-HSL and C4-HSL, respectively ( Figure 5C).
An approximately 3.7-fold and 6-fold induction of phzA1-lacZ was seen in E. coli expressing either LasR or RhlR, in response to their cognate autoinducers. The phzM-lacZ fusion showed a significant, albeit modest, effect of the homoserine lactone autoinducers and LasR, RhlR (1.6-and 1.3-fold, respectively). The results from the expression of the phzA1-lacZ and phzM-lacZ fusions in P. aeruginosa and heterologous expression in E. coli indicate that binding of LasR and RhlR to the same site, in the context of the respective promoters, leads to the enhanced expression of phzA1 over phzM in overall magnitude and dependence on these two quorum sensing regulators.
There are several explanations for these observations. First, the two overlapping regulatory sequences found on the opposite DNA strands are not functionally identical although the differences are rather minor ( Figure 5A). There are two differences in the regulatory sequences adjacent to each promoter (T/A at position 5 and C/G at position 12). These could account for strong preference for LasR and RhlR activation towards the transcription of phzA1. Moreover, the predicted promoters (210 and 235 regions) are not identical and these could also provide additional context for the extent and magnitude of LasR and RhlR activation. Finally, the distances of the LasR element from the TSSs of phzA1 and phzM are different, although well within those seen for other LasR activated genes, and this could also contribute to the overall transcriptional level and extent of influence of quorum sensing regulators and autoinducers. Conceivably, all of these factors could incrementally contribute to regulated expression.

Detection of expressed non-coding RNAs
Non-coding RNAs (ncRNAs) are now appreciated as important regulators of diverse processes in bacteria. Although P. aeruginosa has been studied extensively, only 38 intergenic ncRNAs (sRNAs) have been identified in its genome so far [66]. By combining our TSS mapping and the whole-transcriptome data, we determined 223 events of intergenic transcription based on the current genome annotation ( Figure 6). Of these, 23 represent protein-coding genes that were annotated in other genomes but so far escaped detection in PA14 (Table S5; Materials and Methods), and 35 additional transcripts that might represent non-conserved protein coding genes (Table S6). The sequences of the remaining 165 transcripts lack an ORF and therefore most likely represent non-coding transcripts (Table S6). Most of the previously described sRNAs (31/38, 86%) were detected in our sRNA set, and we could define the exact TSS for 26 of them. Most of the sRNAs we identified (69.5%) are conserved only within other P. aeruginosa strains, and only a few (10.5%) are conserved outside of the genus Pseudomonas (Table S6).
We next examined the distribution of the DNA segments that are transcribed into asRNAs and sRNAs in the genome of P. aeruginosa PA14. Specifically, we assessed whether they are part of the flexible or core genome, as determined previously [9] (Figure 6). We noted an over 2-fold enrichment in the location of sequences specifying both sRNAs (Fisher's exact test p = 0.003) and antisense RNAs (Fisher's exact test p = 8.83610 218 ) in regions that have been previously designated as regions of genome plasticity (i.e. the flexible genome); sRNAs and asRNAs were found less frequently in the conserved core genome. For example, the annotated 115 kb pathogenicity island PAPI-1 carries only two genes with predicted function in transcriptional regulation, RL0037 (a RcsB ortholog) and a predicted transcription regulator RL0012 [67]. Strikingly, in addition to the 175 annotated genes in PAPI-1, we were able to detect additional 46 ncRNAs (10 sRNAs and 36 asRNAs) representing 21% of the genes in the island; this number of asRNAs in PAPI-1 is enriched by over 5-fold compared to the core genome. Since the flexible genome represents, in most instances, horizontally-acquired genes, it is likely that these segments specify regulatory elements controlling the expression of co-acquired target genes. However, it could be that genes expressed from other genomic islands or from the core genome could be subject to regulation by the trans-acting sRNAs at a post-transcriptional level [68]. These findings raise the possibility that the ability to acquire ncRNA regulators via horizontal gene transfer contributes significantly to bacterial evolution, where new phenotypes emerge by post-transcriptional activation or repression of genes already present in the recipient cells. An alternative explanation is that horizontally acquired sequences in the flexible genome were newly introduced to the genome, and due to lack of evolutionary time to ameliorate their coding-and non-coding sequences, the acquired sequences contain random promoter sequences that are recognized by Pseudomonas transcriptional machinery.
Interestingly, two of the predicted sRNAs (Lrs1 and Lrs2) were identified as having a conserved binding site with similar binding Figure 5. Role of LasR and RhlR in regulating the biosynthesis of pyocyanin. (A) The location of the LasR and/or RhlR binding sites are presented on top. Binding sequences are highlighted in red, asterisks in the binding sequences represent nucleotide identity between the forward and reverse sequences of the palindromic binding site; non-identical positions are marked with arrows. Also shown are the consensus sequences recognized by LasR, RhlR and the sequence of the LasR-RhlR box taken from previous studies [63,65,90]. (B) Growth (lines) and activities of the lacZ reporter (bars) fused to various promoters fragments of the phz operon in wild-type P. aeruginosa PA14, DlasR and DrhlR strains. (C) Promoter activity in E. coli; (i) a fusion of a promoter fragment and lacZ transcriptional reporter (indicated in the upper left corner of the panel) and (ii) plasmid expressing either lasR (pJTT201, indicated as pLasR) or rhlR (pJTT202, indicated as pRhlR), or the empty vector pMMB67 [73]. The acyl-homoserine lactones 3OC 12 -HSL (2 mM) and C 4 -HSL (10 mM) were added as indicated. The rsaL promoter previously shown to directly bind LasR was used as a positive control while a fragment containing the mvfR gene's promoter served as the negative control. doi:10.1371/journal.ppat.1002945.g005 Figure 6. Distribution of temperature regulated genes, and segments transcribing antisense RNAs and intergenic small RNAs in the genome of P. aeruginosa PA14. The outer circle represents the core (blue) and accessory genes (yellow) as defined previously [9]. The innermost two circles represent genes that were differentially expressed at 28uC and 37uC (blue and orange, upregulated in 28uC and 37uC, respectively); the middle two circles show the expression of sRNAs from intergenic regions (red dots) and asRNAs (green dots). For each category, relative ranking of expression was grouped into five color-coded bins (darker colors represent higher expression). doi:10.1371/journal.ppat.1002945.g006 affinity for LasR ( Figure 4A, 4C). Monitoring the expression of lrs1 and lrs2 with transcriptional lacZ fusions showed LasR dependence ( Figure 7A) throughout the entire growth phase of the cultures, although there was a slight increase in the expression of the lrs1-lacZ fusion and lrs1 transcript levels ( Figures S3 and S4B) when cultures reached a late stationary phase of growth. A Northern blot with a probe designed for lrs1 with RNA extracted from wild-type PA14, a DlasR mutant, and a DlasR strain complemented with LasR on a plasmid indicated that lrs1 expression is indeed dependent on LasR ( Figure S4B). We also note that during the late stationary phase Lrs1 can be detected even in a lasR mutant, suggesting that its regulation may be directed by another transcription factor expressed or activated at this late growth phase. We further demonstrated using EMSA that the concentrations of Lrs1 (but not Lrs2) in P. aeruginosa were dependent on the RNA chaperone Hfq through a direct interaction ( Figure 7B). We also observed that Lrs1 transcript levels, but not Lrs2 levels, are reduced in a hfq mutant, possibly due to reduced transcript stability ( Figure 7C). These results point toward an overlapping regulatory network that includes a transcriptional regulatory cascade and a sRNA-regulated post-transcriptional regulatory component. The quorum sensing regulatory system therefore very likely integrates, in addition to cell density, a number of regulatory inputs and directs these to the most optimal responses for survival in a particular environment.
To further investigate the possible roles of Lrs1, we constructed an isogenic deletion mutant lacking the lrs1 sequence. We first determined the precise 59 and 39 ends of the transcript following intramolecular ligation and synthesis of cDNA across the junction. The sequence of the cDNA was used to design PCR primers for deletion of the entire lrs1 in P. aeruginosa PA14. Surprisingly, the culture of the mutant strain was devoid of the characteristic bluegreen pigment of P. aeruginosa, a color that stems from production of pyocyanin ( Figure S4C). Measurements of levels of pyoverdine, the fluorescent pigment produced by P. aeruginosa, showed no detectable differences between culture supernatants of wild-type and the lrs1 mutant (data not shown). To detect the transcriptional changes that might contribute to this phenotype we constructed additional RNA-seq libraries from wild type and Dlrs1 isolates. Interestingly, the gene expression in both libraries was almost identical (Pearson r.0.99), except for a single operon, the anthranilate dioxygenase operon (antABC), which was upregulated by more than 2-fold in the mutant strain. This might explain a highly efficient metabolic flux to this pathway, which in turn depleted the production of pyocyanin, due to lack of chorismic acid ( Figure S1); in turn this may provide an explanation for the lack of the color in the Dlrs1 cultures. Strikingly, Dlrs1 also resulted in a significant (over 4-fold) upregulation in the expression of the two PrrF sRNAs, which are regulators of a number of iron-related genes [69].

The Pseudomonas transcriptome browser
As part of this study, we created a comprehensive graphical transcriptome map of P. aeruginosa. This map provides a highresolution view of the Pseudomonas transcriptome structure and its regulatory sequences. To make this data an accessible resource for further research we generated an online repository and a dataviewer, The Pseudomonas transcriptome browser, which is available at http://www.weizmann.ac.il/molgen/Sorek/ pseudomonas_browser/. The transcriptome browser allows navigating the transcriptome and genome data, and provides direct links to additional Pseudomonas resources.
In conclusion, the high-resolution analysis of the transcriptome of the opportunistic pathogen P. aeruginosa presented here provides a framework for studying the regulatory mechanisms that allow this common soil organism to become a successful human pathogen in various compromised hosts. In this work, we provide a detailed snapshot of the operon organization, the abundance of coding and non-coding transcripts and their sites of transcription initiation, taken at temperatures encountered most likely in the natural environment and in the human body. Taking advantage of this data, we could significantly expand the number of genes that are under the direct control of LasR, the master quorum-sensing regulator, including two novel sRNAs. In addition to creating a valuable tool for the P. aeruginosa research community, The Pseudomonas transcriptome browser, our work demonstrates an important function for temperature in controlling the expression of many of P. aeruginosa virulence factors. Considering the wide range of reservoirs, including the human body, this detailed transcriptome analysis should provide insights into the interactions between the various mechanisms that direct the assimilation of various signals and directing selective and coordinated gene expression to favor survival of this organism in often challenging environments.

Materials and Methods
Strains, mutant construction and culture conditions P. aeruginosa PA14 and PAO1 are standard laboratory strains. P. aeruginosa PAK DpscC was described previously [70]. Unmarked mutations in P. aeruginosa PA14 in lrs1 and hfq were generated using pEXG2 and the sucrose counter-selection method [71]. The pET29a-hfq recombinant plasmid was constructed by PCR amplification of the 249 bp coding sequence of P. aeruginosa hfq gene and cloning the DNA fragment into pET29a vector. P. aeruginosa lasR and P. aeruginosa rhlR mutant strains were constructed by introducing pJTT4 and pEXG2-RhlR into P. aeruginosa PA14 as previously described [72]. The complementation plasmid pLasR was constructed by PCR amplification of the lasR coding sequence and cloning the amplicon into pMMB67EH [73]. Strains defective in type III secretion system proteins and effectors were described previously [74]. E. coli and P. aeruginosa strains were grown in Luria broth (LB; Lennox formulation) unless otherwise noted. For monitoring the expression of the type III secretion effectors ExoS, ExoU, and the secretin PscC, the MinS medium lacking calcium supplement was used [75].
RNA harvesting and cDNA synthesis RNA was isolated from P. aeruginosa as described previously [76]. Briefly, bacteria were grown in LB to early stationary phase [optical density at 600 nm (O.D. 600 ) of 1.8] and total RNA was isolated from ,7610 8 cells using hot phenol followed by ethanol precipitation. The samples were digested with DNase I for 2 hr, followed by phenol-chloroform-isoamyl alcohol extraction and ethanol precipitation. The precipitated sample was resuspended in 20 mL of RNase-free water. The quality of the resultant RNA was determined using an Agilent Bioanalyzer. For samples with RNA Integrity Numbers of greater than 7.5, the MICROBExpress Kit (Ambion) was used to deplete rRNA from the total RNA samples. The loss of intact rRNA was verified using the Agilent Bioanalyzer. cDNA was generated by using the methods from the Superscript Double-Stranded cDNA Synthesis Kit (Invitrogen) per manufacturer's instructions.

Library preparation and sequencing
Illumina RNA-seq whole-transcriptome libraries were prepared using the Paired End Sequencing Sample Preparation Guide (Illumina) per the manufacturer's instructions as previously described [15,77].
RNA from the same extractions was subjected to the strandspecific 59-end sequencing protocol as previously described [17]. Briefly, RNA samples of both conditions were divided into two sub-samples (hereby called TAP+ and TAP2) containing 2.5 mg from rRNA-depleted RNA. The RNA of TAP+ sub-sample was incubated with 2 U of tobacco acid pyrophosphatase (TAP, Epicentre) for 1.5 hr at 37uC to generate 59 monophosphate RNAs. The other sub-sample (TAP2) was not treated with TAP. cDNA sequencing libraries were prepared as previously described [15,17,77] with an addition of a unique 4-base index sequence to the 59 RNA adapter used for each sub-sample followed by 2 random bases for ligation quality control. cDNA libraries of every two sub-samples were mixed in equal amounts prior to Illumina sequencing. The resulting sequences were sorted to the correct sub-sample by their index bioinformatically.

Read mapping and identification of differentially expressed genes
Sequencing reads were mapped to the corresponding reference genome (Genbank: NC_008463) as described previously [15,17,77]. The number of overlapping mapped reads was counted for every ORF in each sample of whole-transcriptome sequencing (2 biological replicates for the 28uC and 37uC samples). For each ORF the number of reads per kilobase of gene model per million mapped reads (RPKM) was calculated as previously described by Mortazavi et al. [78]; this procedure also normalizes samples having different numbers of initial reads. Differentially expressed genes were called by performing x 2 test on the RPKM values and applying Bonferroni correction for multiple-testing. Genes with adjusted p-values smaller or equal to 0.05 in both replicates were determined to be differentially expressed.

Gene ontology enrichment assay and determination of virulence related genes
Gene ontology (GO) annotation for Pseudomonas aeruginosa PAO1 was downloaded from the Gene Ontology repository website (http://cvsweb.geneontology.org/cgi-bin/cvsweb.cgi/go/geneassociations/gene_association.pseudocap.gz). Gene identifications (gene ids) for strain PAO1 were substituted with homolog gene ids for strain PA14 (homologs were downloaded from http://www. pseudomonas.com/download.jsp). Differentially expressed genes that have homologs in strain PAO1 (224/234 and 137/144 for 28uC and 37uC samples, respectively) were tested for GO enrichment against the entire set of ORFs of PAO1, which have homologs in PA14 (n = 5374) using Ontologizer [79]. GO categories with adjusted p-values smaller than 0.05 were considered significant. A list of virulence-related genes in P. aeruginosa was downloaded from the Virulence Factors Database (VFDB) (http://www.mgc.ac.cn/VFs/) and genes in the list were considered to be virulence-related for subsequent analyses.

Determination of TSSs
For each position in the genome the number of mapped reads derived from the 59 end sequencing experiments was recorded in a strand-specific manner. Positions with five or more uniquely mapped reads (encompassing 78% of the reads) were considered as reproducible sites. Reproducible sites in intergenic regions or overlapping ORFs were associated with the closest gene according to the gene annotation (Genbank: NC_008463). In cases where the site was positioned over 250 nt from the downstream gene the site was not associated with any gene. All reproducible sites that were not associated with genes were inspected manually using the Pseudomonas transcriptome browser and in cases which had a continuous whole-transcriptome expression from the site to the beginning of a downstream gene, the site was associated with the corresponding gene. The TSS of each gene was defined as the reproducible site upstream to the beginning of the gene with the highest number of reads in comparison to other sites that were associated with the same gene, as previously described [15,17]. This was validated by randomly selecting 200 TSSs and checking for their agreement with the whole-transcriptome sequencing data, searching for continuous expression from the TSS in the direction of the transcription. TSSs (n = 2,117) were further validated by comparing the ratio of reads derived from the TAP+ sample divided by the number of reads from TAP2 sample. The TAP+/ TAP2 ratio of putative TSSs were found to be enriched by an average of 2.6-fold (Wilcoxon sum rank p,1 2193 ) as compared to non-TSS sites, indicating that they were preceded by a triphosphate, a hallmark of primary transcripts [80].

Generation of transcriptional unit map
Genes that begin a transcriptional unit (TU) were defined as (1) genes with a defined TSS; (2) genes with an upstream gene located on the opposite strand; or (3) expressed genes that lacked a defined TSS were defined as beginning a TU, if the upstream gene was not expressed, or was expressed by at least 2-fold difference. When genes were not expressed under the conditions tested, operon structures were determined based on distance between pairs of genes, where genes occurring on the same strand with a distance shorter than 72 nt considered as belonging to the same operon. This threshold was empirically determined based on the distances between genes in expressed polycistrons with defined TSSs (n = 646; median distance = 10 nt), where the 90th-percentile distance between genes was 72 nt.

Genome-wide detection of sRNAs
Small RNAs were identified based on the presence of a TSS and expression in intergenic regions (IGRs) that was not associated with genes. The TSS site was marked as the 59 end of the putative sRNA and the 39 end was estimated from the whole-transcriptome data (Table S6). Previously described sRNAs in Pseudomonas aeruginosa PA14 that were expressed in the 59 end-sequencing and/or the whole-transcriptome data were included in the list. In addition, the sequence of known sRNAs from the P. aeruginosa PAO1 were searched for in the PA14 genome by a BLAST search (parameters '-p blastn -e 0.001 -F F -W 4 -q 1') and sRNAs that were found and expressed in the PA14 genome were also included in the list. All expressed regions in the list were examined manually in the Pseudomonas transcriptome browser and were validated to be transcribed independently from their downstream genes; putative sRNAs that were not transcribed independently (i.e., transcribed along with the downstream or upstream gene) were removed from the list. All remaining intergenic transcripts represented independent expression of either a non-coding RNA or an un-annotated ORF. Therefore, all intergenic transcripts were scanned for putative ORFs and the longest possible ORF lacking a stop-codon in every region of expression was subsequently searched for homology against the entire protein collection (nr) using BLAST+ with a minimal e-value of 10 24 [81]. Putative ORFs with homologs in other organisms were removed from the list and moved to a list of un-annotated ORFs in P. aeruginosa PA14 genome (Table S5). sRNAs containing putative ORFs larger than 50 amino acids and lacking sequence homology in the protein database were marked as putative ORFs in the list of sRNAs. Terminator sequences for the sRNAs were predicted using TransTermHP v2.09 [82].
The conservation of the sRNAs was tested using BLASTN search against the entire nucleotide collection (nt) with the following parameters ('-W 4 -e 0.001 -F F -q 1') demanding that homologs would span at least 50% of the sRNA sequence. BLAST reports were analyzed and each sRNA was assigned a category as follows: (1) PA14-specific when homologs were not identified; (2) P. aeruginosa-specific when homologs were found only in the P. aeruginosa lineage; (3) Pseudomonas-specific when homologs were found also in other Pseudomonas species; and (4) widespread when the sRNA sequence was found outside of the Pseudomonas lineage.

Computational identification of LasR sites in gene promoters
The sequences of promoters known to be directly regulated by LasR were downloaded from the Prodoric database [83]. The sequences upstream to all defined TSSs of ORFs and ncRNAs (200 nt) were extracted from the PA14 genome and searched for significant matches to the PSSM model using matrix-scan [84] with the following parameters ('-pseudo 1 -decimals 1 -1str -origin end -bginput -markov 0 -bg_pseudo 0.01 -return sites -lth score 0'). The process was performed iteratively to refine the PSSM for 3 rounds until convergence of the matches was achieved (Table S4).

Quantitative Real-Time PCR
Unique primers were designed for ,100 bp and ,400 bp segments of transcripts using MacVector software. Triplicate qRT-PCR reactions were set up using 1X concentration of SYBR FAST One-Step qRT-PCR Universal (Kapa Biosystems), optimized concentrations of appropriate primers, and one of the following templates: 10 8 , 10 6 , 10 5 , 10 4 , 10 3 , 10 2 molecules of ,400 bp PCR-generated standard curve template, 20 ng of the total RNA used for Illumina sequencing, 80 ng of genomic DNA, or water (no template control). Reactions were run in a Mastercycler realplex RT-PCR machine (Eppendorf) and the PCR was performed with the following conditions: 42uC for 5 min, 95uC for 5 min, (95uC for 3 sec, 60uC for 20 sec)640 cycles. After determining the absolute number of copies in each sample for each gene, relative concentrations were calculated using Microsoft Excel.

Western immunoblotting
Wild-type and the pscC, exoU and exoS mutants of P. aeruginosa were grown from fresh overnight cultures to an O.D. 600 of 1.8. Cells were collected from 1 mL aliquots by centrifugation. Five hundred microliters of supernatants were concentrated 20-fold using an Amicon Ultra 10 kDa MW limit column (Millipore). Cell pellets and concentrated supernatants were dissolved in 4X SDS-PAGE buffer and normalized according to their optical density at time of harvesting. Equal volumes of each were loaded on 8-10% SDS polyacrylamide gels, electrophoresed, and transferred to PVDF membranes for Western immunoblotting. Blots were blocked in 5% non-fat skim milk, and incubated with the following dilutions of the primary antibodies, as described (Hoang et al., 2001): anti-PscC at 1:100, anti-ExoS and anti-OprF at 1:5000 and anti-ExoU (generously provided by Dr. Dara Frank, Medical College of Wisconsin) at 1:10,000. Blots were washed with 1X PBST, followed with probing by goat anti-rabbit-HRP conjugate antibodies (BioRad). After washing again, blots were visualized following treatment by SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific).

Determination of 59 and 39 ends of transcripts following intramolecular ligation for lrs1
The procedure for mapping the 59 and 39 ends of transcripts was carried out as previously described [85] with minor modifications. Total RNA was isolated from early stationary phase of P. aeruginosa PA14 and residual DNA was removed with TURBO DNase (Ambion). Total RNA (6.5 mg) was treated with 5 U of tobacco acid pyrophosphatase (Epicentre) for 1 hr at 37uC. Following phenol/chloroform extraction and ethanol precipitation, RNA was treated for 16 hr at 16uC with 20 U of T4 RNA ligase (NEB) for RNA circularization. After phenol/chloroform extraction and ethanol precipitation, 1.5 mg of self-ligated RNA was reverse transcribed using an lrs1 specific primer (R_lrs1+151 ;  Table S7) and SuperScriptIII First-Strand Synthesis system (Invitrogen). RNA was removed with 25 U of RNase If (NEB) in addition to RNase H treatment after termination of the reaction. Approximately 10% of the reaction was used as the template for PCR amplification using GoTaq Green Master Mix (Promega) and the lrs1 primer pairs of R_lrs1+126/F_lrs1+152. Cycling conditions were: 94uC/3 min; 35 cycles of 94uC/45 sec, 61uC/ 30 sec, 72uC/30 sec; 72uC/3 min. The PCR products were separated by agarose gel electrophoresis; DNA bands were eluted and cloned into pCR2.1 TOPO vector (Invitrogen) and the inserts were sequenced using the M13R primer.

RNA isolation and Northern blot analysis
Overnight cultures grown at 37uC were diluted in LB containing gentamicin (75 mg/mL) to an initial O.D. 600 ,0.02. Cell growth was monitored at each time point and IPTG was added at the defined time points (4, 8, and 12 hr) at 2 mM final concentration. After IPTG induction for 30 min, total RNA was isolated by the hot phenol method as previously described [76]. Total RNA (5 mg) was fractionated on a 5% polyacrylamide gel/ 7M urea gel and transferred onto a Hybond-XL membrane (GE Healthcare). RNAs were UV cross-linked to the membrane and hybridized at 42uC with oligonucleotides R_lrs1+126 and R_5S+67 as probes for Lrs1 and 5S rRNA, respectively, and then 59-end radiolabeled using [c-32 P]-ATP (Perkin Elmer) and T4 polynucleotide kinase (NEB). Signals were visualized with a phosphoimager (Typhoon 9400, GE Healthcare) and analyzed with ImageQuant TL software (GE Healthcare).

Electrophoretic mobility shift assays (EMSA)
We assessed LasR binding to PCR-generated DNA using EMSA as previously described [86]. PCR products were endlabeled with [c-32 P]-ATP (Perkin Elmer) with T4 polynucleotide kinase (NEB). Binding reactions contained 1 pM of both a specific and nonspecific probe and 30, 3, or 0.3 pM purified LasR in a total volume of 20 mLDNA binding buffer (20 mM Tris-HCl [pH 7.5], 50 mM KCl, 1 mM EDTA, 1 mM DTT, 100 mg/ml bovine serum albumin, 10% glycerol). For each DNA target, a control reaction contained probe but no LasR. Mixtures were incubated for 20 min at room temperature and separated on a non-denaturing 5% Tris-glycine-EDTA polyacrylamide gel for 120 min at 50 V using a mini-PROTEAN tetra-cell (Bio-Rad). The gels were dried and exposed to a storage phosphor screen; exposure was visualized using a Storm phosphorimager and ImageQuant TL software (GE Healthcare).

Hfq gel shift assay
Hfq was purified from lysates of E. coli BL21(DE3) pLysS cells harboring pET29a-hfq grown in LB medium at 37uC. Expression of the protein was induced by the addition of IPTG to 1 mM when the cultures were at O.D. 600 of 0.4. After an additional 3 hr incubation at 37uC, cells were collected and lysed by sonication in 30 mL of buffer A (50 mM sodium phosphate, 300 mM NaCl) and a protease inhibitor cocktail tablet (Complete mini, EDTAfree, Roche Diagnostic) was added. Cells were treated with 20 U of TURBO DNase (Ambion) and 100 mg of RNase A and incubated on ice for 1 hr. Centrifugally-clarified lysate was passed over a Co 2+ affinity column (Clontech). Eluted fractions was pooled and dialyzed with buffer B (10 mM Tris-Cl [pH 7.5], 50 mM NH 4 Cl, 0.2 mM EDTA, 2% glycerol).

Construction of lacZ fusions and b-galactosidase assays
Promoter fragments including the putative LasR binding sites were amplified by PCR from P. aeruginosa PA14 genomic DNA using each specific primer pairs shown in Table S7. Each fragment was cloned into the integration vector mini-CTX-lacZ. These constructs and the vector control were integrated into the chromosome of P. aeruginosa PA14 as described previously [88]. The assay for activation of transcription by homoserine lactones in E. coli was carried out as described previously [72]. The promotercontaining fragments of phzA1, phzM, and rhlA were cloned into pZE21-LacZ and transformed into strains harboring plasmids expressing lasR, rhlR, or an empty vector. b-galactosidase activities were assayed as previously described [72].

Pyocyanin assay
Pyocyanin was extracted from 1 mL of culture supernatant following addition of 0.6 mL chloroform. After extraction, the organic phase was removed and added to 1 mL of 0.2 N HCl. Following centrifugation, 0.8 mL of the aqueous phase was removed. Concentration of pyocyanin was determined by multiplying the absorbance at 520 nm by 30 (the dilution factor) and then by 17.072 [46].

PQS assay
To quantify PQS levels, a bioassay was developed where a lacZ fusion responds to exogenous PQS. A pqsA-lacZ transcriptional fusion was constructed by cloning a DNA fragment corresponding to pqsA (2246 to +231 relative to the start site of transcription) into mini-CTX-lacZ and conjugating it into PAO1DpqsA strain. The bgalactosidase activities in this strain were measured by addition of supernatants from culture grown at two different temperatures and at two different growth phases. For supernatant isolation, the equal amounts of diluted PA14 (O.D. 600 ,1.0) grown overnight at 28uC and 37uC were diluted 1:100 in 25 mL LB medium and cultured to early stationary and stationary phase by monitoring the O.D. 600 . Next, 1 mL of the culture was isolated by centrifugation at 10,0006g for 5 min and the supernatant was passed through a 0.2 mm syringe filter. For the bioassay, the reporter strain (PAO1 pqsA-lacZ DpqsA) was grown overnight at 37uC and diluted to 1:200 in 10 mL LB medium containing 10 mg/ml tetracycline. The culture was incubated at 37uC until it reached an O.D. 600 of 1. One hundred mL of the reporter strain were mixed with 100 mL of each supernatant and then incubated at 37uC for 30 min. bgalactosidase activities were assayed as previously described [72].

Pyoverdine assay
Pyoverdine was measured by growing the bacteria in King's B medium. Strains grown on King's B agar were cultured in 2 mL liquid King's B medium at 37uC overnight. Cultures were diluted to an initial O.D. 600 at ,0.05 in 25 mL King's B liquid medium and then incubated at 37uC with shaking. At defined time points, the O.D. 600 was measured and supernatants were isolated from 1 mL of each culture by centrifugation at 1,0006g for 3 min. The absorbance (A) at 400 nm was measured and expressed as the ratio of A 400 /A 600 . Figure S1 Effects of temperature on the enzymes of various metabolic pathways providing precursors of quinolones, pyochelin, phenazine and TCA cycle intermediates. Enzymes whose transcript levels were increased in P. aeruginosa grown at 37uC are boxed in red; those that were decreased at this temperature are boxed in blue. Transcripts for enzymes in black boxes and crossed boxes were unchanged or not detected, respectively. (TIF) Figure S2 The transcriptome structure of P. aeruginosa PA14. (A) Number of genes in TUs. Shown is a histogram of the numbers of genes per TU. The vast majority of genes are found in mono-and bi-cistrons, and only a minority of the genes is transcribed as longer operons. (B) Length distribution of 59 UTRs in P. aeruginosa. Only a small number of the P. aeruginosa 59 UTRs are longer than 100 nt (median length of 47 nt). The unbiased TSS mapping allowed the detection of a small group of genes with extraordinarily long 59 UTRs (.250 nt), which might represent cis-regulatory sequences. (TIF) Figure S3 LasR-dependent expression of LacZ reporter fusions. The same fragments used in the EMSA ( Figure 3C) were cloned upstream of a lacZ gene and introduced into P. aeruginosa PA14 wild-type and DlasR. Overnight cultures grown at 37uC were diluted in LB to an initial O.D. 600 of 0.02. The b-galactosidase activity was measured in cultures at (A) late logarithmic (8 h) and (B) stationary phase (12 h) of growth. A two-tailed student's t-test assuming equal variance on the three replicates of wild-type and mutant strains was used to evaluate the significance of the differences in b-galactosidase activity between samples. Asterisks indicate pvalues less than 0.01 between wild-type and mutant strains. (TIF) Figure S4 Expression and regulation of pigment production by Lrs1. Shown are growth (A) and Northern blot analysis of Lrs1 expression (B) in (i) P. aeruginosa PA14 carrying an empty vector; (ii) a PA14 DlasR strain carrying an empty vector, and (iii) a PA14 DlasR strain containing the complementation plasmid pLasR. Northern blot performed separately on the 3 strains validated the expression of the Lrs1 sRNA in strains expressing LasR either from chromosomal or extra-chromosomal origins. (C) Absence of the green pigment from a culture of an lrs1 mutant (left), compared to wild-type P. aeruginosa PA14 grown for 8 hours at 37uC. (TIF)