Regulation of las and rhl quorum sensing in Pseudomonas aeruginosa

The production of several virulence factors by Pseudomonas aeruginosa is controlled according to cell density through two quorum-sensing systems, las and rhl. The las system is comprised of the transcriptional activator protein LasR and of LasI, which directs the synthesis of the autoinducer PAI-1. Similarly, the rhl system consists of the transcriptional activator protein RhlR and of RhlI, which directs synthesis of the autoinducer PAI-2 (formerly referred to as factor 2). To study the interrelation between the two P. aeruginosa quorum-sensing systems, we fused a lacZ reporter gene to lasR, rhlR, and rhlA and monitored expression of these three genes under various conditions. Our data indicate that lasR and rhlR are expressed in a growth-dependent manner, with activation of each gene occurring during the last half of log-phase growth. We also show that the las quorum-sensing system controls the rhl quorum-sensing system in two ways. First, we found that LasR and PAI-1 activated rhlR transcription. Second, we showed that PAI-1 blocked PAI-2 from binding to RhlR, thereby inhibiting the expression of rhlA. Our data thus indicate that the las system exerts two levels of control on RhlR, transcriptional and posttranslational.

At least two complete quorum-sensing systems, las and rhl, are present in the opportunistic human pathogen Pseudomonas aeruginosa. These systems are known to control the expression of a number of virulence genes in response to bacterial cell density (5), but their specific effect on each other has not been studied. The las and rhl systems each contain homologs of the LuxR and LuxI proteins of the prototypic lux quorum-sensing system from Vibrio fischeri (see reference 5 for a review). The las system consists of the transcriptional activator protein LasR and of LasI, which directs the synthesis of the autoinducer PAI-1 [N-(3-oxododecanoyl)-L-homoserine lactone] (6,20,22). This system has been shown to activate the expression of lasI, lasB, lasA, apr, and toxA (7,20,27,30). Similarly, the rhl system consists of the transcriptional activator protein RhlR and RhlI, which directs the synthesis of the autoinducer PAI-2 (N-butyryl-L-homoserine lactone; formerly known as factor 2) (16,17,23). This system controls the expression of rhlI and rhlAB, which codes for a rhamnosyltransferase required for rhamnolipid (heat-stabile hemolysin) production (11,(15)(16)(17). It has also been reported that rhl quorum sensing activates the expression of rpoS, a stationary-phase sigma factor that controls numerous genes (11).
The general model for quorum sensing (5) begins with the autoinducer, which is a diffusible molecule, being produced at a basal level at low cell densities. The autoinducer concentration then increases with cell density until a threshold concentration is reached. At this concentration, the autoinducer binds to its specific target protein (i.e., LasR or RhlR), and the autoinducer-protein complex activates genes that it controls.
Experiments on the interchangeability of the las and rhl system components showed that they were not compatible, in that PAI-2 does not activate LasR nor does PAI-1 activate RhlR in Escherichia coli (23,24). However, it was apparent that these two systems were not completely independent of one another. It was first indicated that the las and rhl systems may be linked when Pearson et al. (23) showed that PAI-2 was poorly expressed in a P. aeruginosa lasR strain, which led to the conclusion that LasR may control PAI-2 production. After this, it was reported that the rhl system was working in tandem with the las system to control the production of elastase in P. aeruginosa (3,17). While this paper was being written, an article was published that showed, as we do below, that rhlR transcription was controlled by LasR-PAI-1 (11). For our study, we sought to determine the effect, if any, of the las and rhl systems on each other. We report here that the las quorumsensing system controls RhlR, the transcriptional activator of the rhl system, at both the transcriptional and posttranslational levels. In addition, we show that lasR and rhlR are expressed in a growth-dependent manner, with each gene being activated during the second half of log-phase growth.

MATERIALS AND METHODS
Bacterial strains and plasmids. Bacterial strains and plasmids used in this study are listed in Table 1. Plasmid pECP60 was constructed by ligating the 623-bp, rhlA promoter-containing (24) BamHI/BglII fragment of pUO101 into the BamHI site of pSW205 to create a rhlAЈ-lacZ translational fusion. Plasmid pUO101 was kindly provided by U. Ochsner and carries a 5.8-kb HindIII/EcoRI fragment of P. aeruginosa PG201 DNA which contains the entire rhl regulon (rhlABRI). Transcriptional fusions of lasRЈ-lacZ and rhlRЈ-lacZ were constructed in the vector pLP170. A lasR promoter-containing fragment was generated through PCR, using primers that contained an EcoRI or BamHI site. The 3Ј end of the primer containing the EcoRI site (5Ј-GCGTGGGGAATTCCGCGTGC GCCGCGC-3Ј) corresponded to the nucleotide 395 bp upstream from the lasR start codon. The 3Ј end of the primer containing the BamHI site (5Ј-CCGTCG GATCCACCCGCGCGTAGCC-3Ј) corresponded to the nucleotide 197 bp downstream from the lasR start codon. The transcriptional start sites of lasR have been mapped to nucleotides 201 and 231 bp upstream from the lasR start codon (1), which ensured that the lasR regulatory region is contained on the sequence used to make the lasRЈ-lacZ fusion. The PCR product obtained by the use of a lasR template was digested with EcoRI and BamHI and ligated into the transcriptional fusion vector pTL61T, which had been digested with the same enzymes. The resulting lasRЈ-lacZ transcriptional fusion was released from this vector by digestion with EcoRI and EcoRV and ligated into pLP170 that had been digested with the same enzymes to yield pPCS1001. The PCR-generated fragment was sequenced to ensure that mutations had not been incorporated. The rhlR promoter-containing fragment was released from pUO101 by using the enzymes PstI and BamHI and was ligated into PstI/BamHI-digested pTL61T. The recognition sites for the enzymes PstI and BamHI are located at nucleotides Ϫ500 and ϩ242 relative the rhlR start codon. Through deletion analysis of the DNA upstream from rhlR, Ochsner et al. (16) mapped the promoter region of rhlR to the region within 80 nucleotides upstream from the rhlR start codon. This confirmed that the sequence used to construct the rhlRЈ-lacZ fusion contains the rhlR promoter. The rhlRЈ-lacZ transcriptional fusion was released by digestion with EcoRI and EcoRV and ligated into pLP170 digested with the same enzymes to yield pPCS1002. Nucleotide fusions were specifically verified by sequencing to ensure integrity of the fusion junction.
DNA techniques. Standard techniques were used for DNA manipulation (13). PCR was performed using Taq polymerase (Gibco/BRL, Gaithersburg, Md.). Oligonucleotides used for PCR were synthesized by G. Kampo and J. Maniloff at the Core Nucleic Acid Laboratory at the University of Rochester Medical Center. Restriction endonucleases were purchased from Gibco/BRL or New England Biolabs (Beverly, Mass.). Plasmids were introduced into E. coli and P. aeruginosa by transformation (13) or electroporation (28), respectively. E. coli containing plasmid DNA was selected on L-agar (13) plates containing ampicillin (100 g/ml) and/or chloramphenicol (30 g/ml) when appropriate. P. aeruginosa containing plasmid DNA was selected on peptone tryptic soy broth (PTSB) (18) agar plates containing carbenicillin (200 g/ml). Nucleotide sequencing was accomplished by using the Sequenase kit (U.S. Biochemical Corp., Cleveland, Ohio) and [␣-35 S]dATP (NEN Research Products, Boston, Mass.).
Media and culture conditions for ␤-Gal activity assays. (i) P. aeruginosa. P. aeruginosa cultures were grown for 18 h at 37°C with shaking in PTSB (supplemented with 200 g of carbenicillin per ml when appropriate) and subcultured into the same medium to a starting A 660 of 0.05. When indicated, PAI-1 and/or PAI-2 was added at the start of subculturing to achieve a final concentration of 1 M. For Fig. 2, cultures were assayed for ␤-Gal activity throughout the growth cycle and growth was monitored by measuring cell density (absorbance at 660 nM). For Fig. 1 and Fig. 3, cultures were assayed for ␤-Gal activity after 7.5 h of growth. This time point was chosen because it was the time point when all cultures were always at approximately the same cell density (see Fig. 2A) and therefore the assays would be comparable.
(ii) E. coli. E. coli cultures were grown in A medium (13) supplemented with 0.4% glucose, 1 mM MgSO 4 , and 0.05% yeast extract. When appropriate, 100 g of ampicillin per ml and/or 30 g of chloramphenicol per ml was added to the medium. Cultures were shaken at 37°C for 18 h and subcultured to a starting A 600 of 0.08. At an A 600 of 0.3, autoinducers were added where indicated. IPTG (5-bromo-4-indole-3-chloro-isopropyl-␤-D-galactopyranoside) was also added at this time to a final concentration of 1 mM for Fig. 4. For Fig. 6., IPTG was omitted because the tac promoter controlling rhlR allows enough RhlR to be produced to activate rhlA in the presence of PAI-2. When IPTG was added, the massive overexpression of RhlR lessened the effect shown here (data not shown). After additions were made to E. coli cultures, growth was continued for 1.5 h, at which time assays for ␤-Gal activity were completed.
PAI-2 binding assays. Binding assays followed the method of Hanzelka and Greenberg (8) with the following modifications. E. coli DH5␣ cultures containing two plasmids, pJPP8 and pGroESL, were grown (starting A 600 of 0.5) for 2 h in Luria-Bertani (LB) medium (containing 100 g of ampicillin per ml and 30 g of chloramphenicol per ml) at 37°C with shaking in the presence of 1 mM IPTG. Plasmid pGroESL was included because GroES and GroEL were shown to enhance binding of the Vibrio autoinducer VAI-1 to LuxR (8). Culture aliquots were preincubated (25°C, 30 min) in the presence or absence of PAI-2 or PAI-1 (1 or 10 M) followed by a 30-min incubation in the presence of 250 nM [ 3 H]PAI-2 (24). Radioactivity remaining with the cells was determined as described previously (8). All binding experiments were completed twice in duplicate. We have shown elsewhere that E. coli cells which contained pJPP8 and pGroESL bound approximately 20-fold-more [ 3 H]PAI-2 than cells which contained pEX1.8 (control vector) and pGroESL (24).

RESULTS AND DISCUSSION
The las quorum-sensing system affects the rhl quorum-sensing system. It became clear that the two quorum-sensing systems of P. aeruginosa were communicating when we studied the effects of exogenous autoinducers on rhlA expression in cells grown in PTSB medium. It has been shown that PAI-2 and RhlR are required for the activation of rhlA in E. coli (24) and that rhlR is required for rhamnolipid production in P. aeruginosa (16). To study rhlA expression in P. aeruginosa, ␤-Gal activity was assayed to monitor activation of an rhlAЈ-lacZ fusion in early-stationary-phase cultures of P. aeruginosa PDO100 rhlI(pECP60) and PAO-JP2 rhlI lasI(pECP60) (Fig.  1). As expected, strain PDO100(pECP60) exhibited rhlA activation only when exogenous PAI-2 was provided. However, in strain PAO-JP2(pECP60), maximal rhlA activation required both PAI-2 and PAI-1. The addition of PAI-2 alone allowed only 8% (100 Miller units) of the rhlA activation seen when both PAI-1 and PAI-2 were added (1,212 Miller units) (Fig. 1). We have also qualitatively shown that a P. aeruginosa lasI mutant strain does not make rhamnolipid (24). It was obvious  from these data that rhlA expression was dependent on components of both the rhl and las quorum-sensing systems. This result was a surprise because we had expected that activation of rhlA in P. aeruginosa would occur in the presence of PAI-2 and RhlR as it had in E. coli (24). However, in our E. coli experiments, rhlR expression was controlled by the inducible tac promoter instead of its own promoter, thus eliminating the wild-type regulatory sequences upstream from rhlR. This led us to speculate that perhaps RhlR was not expressed in strain PAO-JP2 and that rhlR expression may be regulated through quorum sensing. To test this theory, we examined the activation of our rhlRЈ-lacZ and lasRЈ-lacZ transcriptional fusions in P. aeruginosa strains PAO1 and PAO-JP2 grown in PTSB medium (Fig. 2). We feel that it is very important to study quorum sensing in a well-defined P. aeruginosa mutant strain in which a critical portion of each quorum-sensing system is absent. The use of strain PAO-JP2 lasI rhlI allowed us to test our fusions in the absence of the two known P. aeruginosa autoinducers or in the presence of exogenously added autoinducers (see below), thus ensuring that our data did not result from partial effects caused by interference between the las and rhl systems. This is in contrast to a similar study by Latifi et al. (11) in which they used strain PAO1 with a lasR mutation and the undefined mutant strain PAN067 to test the effects of quorum sensing on lasR and rhlR expression.
In our experiments, we found that activation of rhlR and lasR in strain PAO1 was very similar (Fig. 2B), with both genes exhibiting a basal level of transcription until they became activated in the last half of log-phase growth (culture A 660 , Ͼ1.0) ( Fig. 2A). This result differs significantly from that of Latifi et al. (11), who report that lasR is expressed constitutively in P. aeruginosa. We found that rhlR was also expressed in a growthdependent manner (Fig. 2B), while they (11) report that expression is constant until stationary phase. The reason for this conflict is not apparent. It could result from the use of different media (PTSB for this work and LB medium by Latifi et al. [11]), from differences between the specific P. aeruginosa PAO1 strains used by the two laboratories, or from an unknown factor.
In the lasI rhlI mutant (strain PAO-JP2), basal lasR transcription was unaffected and maximal expression was decreased by approximately 50% compared to that seen in the parental strain, PAO1 (Fig. 2B). However, rhlR transcription was critically affected in the absence of both PAI-1 and PAI-2. In strain PAO1, both basal and maximum levels of rhlR transcription were approximately fivefold higher than the levels seen in strain PAO-JP2 (Fig. 2B). These results indicate that even the basal levels of rhlR, but not those of lasR, are dependent on the presence of PAI-1 and/or PAI-2. These data also show that both lasR and rhlR are expressed to at least some extent in the absence of autoinducers, indicating that these genes are controlled by more than just quorum sensing. Most importantly, these data showed that during all phases of P. aeruginosa growth, maximum rhlR expression was dependent on PAI-1 and/or PAI-2, indicating that rhlR was controlled by one or both quorum-sensing systems.
Transcription of rhlR is positively regulated by LasR-PAI. To determine which quorum-sensing system controlled rhlR, PAI-1 and/or PAI-2 was added exogenously to cultures of strain PAO-JP2 containing pPCS1001 (lasRЈ-lacZ) or pPCS1002 (rhlRЈ-lacZ) (Fig. 3). Interestingly, these data showed that PAI-2 had no effect on rhlR or lasR transcription, but PAI-1 alone was capable of restoring expression of both gene fusions to normal levels (Fig. 3). The ability of PAI-1 to restore full rhlR expression to strain PAO-JP2(pPCS1002) suggests that the las quorum-sensing system controls rhlR transcription. In this scenario, the rhl system indirectly responds to cell density through the las system and only the las system directly "senses a quorum." To ensure that the affect of PAI-1 on rhlR required LasR, we determined the effects of LasR and PAI-1 on rhlRЈ-lacZ expression in E. coli (Fig. 4). In the absence of LasR or PAI-1, rhlR expression remained at background level (Fig. 4). In the presence of both LasR and PAI-1, rhlR expression increased approximately threefold (to the same level as that seen for our lasRЈ-lacZ fusion in E. coli) (Fig. 4), indicating that rhlR was directly controlled by LasR and PAI-1.
(Experiments in which we tested the ability of RhlR to activate lasR and rhlR in the presence of 50 nM, 1 M, and 10 M PAI-1 or PAI-2 showed no activation of either gene in E. coli [data not shown].) Furthermore, the expression of lasR was unaffected by the presence of LasR and PAI-1 in E. coli (Fig. 4). However, it is apparent that in P. aeruginosa PAO-JP2, PAI-1 caused a slight increase (approximately twofold) in expression from the lasRЈ-lacZ fusion ( Fig. 2B and 3). This distinct difference in lasR expression between P. aeruginosa and E. coli makes it difficult to determine the role of autoregulation in lasR expression. It has also been reported that lasR transcription was unaffected in a strain PAO1 lasR mutant but negatively affected by the presence of LasR in E. coli (11), phenomena which we have also observed (2). Additionally, we have shown elsewhere that lasR transcription requires vfr (1), which encodes a homolog of the E. coli Cap protein. Taken together, these data do not lead to a conclusion concerning the effect of quorum sensing on LasR but suggest that lasR expression is complex and controlled by multiple factors. The data in Fig. 2, 3, and 4 show that the las quorum-sensing system controls rhlR at the transcriptional level, providing direct evidence of communication between two quorum-sensing systems of the same organism. This role for LasR-PAI-1 adds to the importance of the complex as a global regulator of P. aeruginosa virulence. In addition, the control of rhlR by the las system implies that the las system is activated before the rhl system, indicating that, as suggested by Latifi et al. (11), a hierarchy exists between the two systems where las quorum sensing is dominant. This conclusion appears to be in direct conflict with earlier reports from our laboratory that indicated lasR could require a factor other than PAI-1 (possibly PAI-2) to function properly (20,23). Subsequent investigation has shown that these conclusions were invalid because two P. aeruginosa strains, PAO-R1(pTS400-1.7) and PAO-R1(pTS400), used to produce data for those papers contained improper plasmids.
Often, gene control through quorum sensing is associated with the presence of a "lux box" consensus sequence which is usually found very close to, or overlapping with, the promoter of a quorum-sensing-controlled gene (5). Upon inspection, we discovered that the DNA sequence upstream from rhlR contains a potential lux box consensus sequence centered 81 nucleotides upstream from the rhlR translational start site. This sequence (5Ј-TTTTGCCGTATCGGCAAGGC-3Ј) matched 10 of 20 nucleotides with the V. fischeri lux box and 11 of 20 nucleotides with OP1 of lasB from P. aeruginosa (26). The importance of these sequences in quorum sensing was emphasized when Stevens et al. (29) showed that the lux box of V. fischeri specifically associates with the DNA-binding domain of LuxR and Rust et al. (26) demonstrated that OP1 of lasB is important for LasR-PAI-1-mediated activation of lasB. Additionally, Ochsner et al. (16) mapped the rhlR promoter region by deletion analysis of the DNA upstream from rhlR and identified a potential rhlR promoter that begins only 7 nucleotides from the 3Ј end of the possible rhlR lux box. The discovery and location of this sequence upstream from rhlR lead us to speculate that it may be involved in the control of rhlR by LasR-PAI-1.
RhlR activity is also controlled at the posttranslational level by the las quorum-sensing system. To further explore the control of rhl by las, we tested the specificity of RhlR-autoinducer interactions. Tritium-labeled PAI-2 ( 94%, respectively (Fig. 5). This result was quite surprising and showed that PAI-1 also competed with PAI-2 for binding to cells expressing RhlR. We have shown elsewhere that PAI-1 and RhlR will not activate rhlA and that PAI-2 will not compete with PAI-1 for binding to LasR (21,24). With this in mind, we speculated that perhaps PAI-1 blocked RhlR activity in vivo. To test this hypothesis, we assayed rhlA expression in the presence of RhlR, 1 M PAI-2, and various concentrations of PAI-1 (Fig. 6). The ability of RhlR and PAI-2 to activate rhlAЈ-lacZ in E. coli decreased in a dose-dependent manner with increasing concentrations of PAI-1 (Fig. 6), suggesting that PAI-1 could block the PAI-2-binding site(s) of RhlR and cause the inhibition of a RhlR-controlled gene. This indicated that PAI-1 controls RhlR activity at a posttranslational level. Kuo et al. (10) have reported that a secondary autoinducer of V. fischeri inhibits activation of the lux operon, but it is not known whether this autoinducer is part of the lux system or primarily involved in a separate quorum-sensing system. The phenomenon shown here is the first report of an autoinducer from a defined quorum-sensing system within a bacterium that inhibits the activity of a second LuxR-type protein within the same bacterium.
To help clarify the circuitry involved in this complex regulation hierarchy, we present a model of P. aeruginosa quorum sensing (Fig. 7). We speculate that in P. aeruginosa, the posttranslational control of RhlR by PAI-1 occurs before rhlI is induced, when the concentration of PAI-1 is higher than the concentration of PAI-2. PAI-1 could block PAI-2 from associating with RhlR until enough RhlR and/or PAI-2 was present to overcome the PAI-1 blocking effect. At that point, RhlR-PAI-2 could autoinduce rhlI, which would allow the concentrations of PAI-2 and PAI-1 to become approximately equal, as we have seen in stationary-phase cultures (5 to 10 M [22,23]). This would allow P. aeruginosa to delay the induction of genes controlled by rhl quorum sensing and provide this organism complex on lasR is unclear, so this was indicated by "ϩ/Ϫ?". The minus symbol by the arrow extending from PAI-1 to the arrow between PAI-2 and RhlR indicates blocking of the association between PAI-2 and RhlR. To begin the quorum-sensing cascade, LasR and PAI-1 are both produced at a basal level. As culture density increases, lasR is activated by Vfr (1), and PAI-1 reaches a threshold concentration and binds to LasR. We speculate that at low culture densities, the PAI-1 concentration is in excess of the PAI-2 concentration, allowing PAI-1 to block the interaction of RhlR and PAI-2. The autoinduction of lasI by LasR-PAI-1 could keep the PAI-1 concentration well above that of PAI-2 until enough RhlR and/or PAI-2 is produced to overcome the blocking effect of PAI-1. Once RhlR associates with PAI-2, autoinduction of rhlI occurs and the remainder of the RhlR-PAI-2-controlled genes are activated.
with yet another mechanism to temporally control the activation of important factors.
The control of RhlR at both the transcriptional and posttranslational levels provides P. aeruginosa with an elegant mechanism through which it can use the las quorum-sensing system to control the rhl quorum-sensing system. The importance of this discovery becomes evident when one considers both the multiple virulence factors controlled by these two systems and the report (11) that RhlR-PAI-2 activates transcription of the sigma factor rpoS. The results presented here also lead us to speculate that it may be possible to develop a single therapeutic autoinducer analog of PAI-1, rather than two different analogs, which will inhibit expression of virulence genes controlled by both LasR and RhlR.