|Home | About | Journals | Submit | Contact Us | Français|
In Pseudomonas aeruginosa quorum sensing (QS), the transcriptional regulator LasR controls the expression of more than 300 genes. Several of these genes are activated indirectly via a second, subordinate QS regulator, RhlR. Conserved sequence elements upstream of individual other genes have been shown to bind LasR in vitro. To comprehensively identify all regions that are bound by LasR in vivo, we employed chromatin immunoprecipitation in conjunction with microarray analysis. We identified 35 putative promoter regions that direct the expression of up to 74 genes. In vitro DNA binding studies allowed us to distinguish between cooperative and non-cooperative LasR binding sites, and allowed us to build consensus sequences according to the mode of binding. Five promoter regions were not previously recognized as QS-controlled. Two of the associated transcript units encode proteins involved in the cold-shock response and in Psl exopolysaccharide synthesis, respectively. The LasR regulon includes seven genes encoding transcriptional regulators, while secreted factors and secretion machinery are the most overrepresented functional categories overall. This supports the notion that the core function of LasR is to coordinate the production of extracellular factors, although many of its effects on global gene expression are likely mediated indirectly by regulatory genes under its control.
Quorum sensing (QS) allows populations of bacteria to communicate via the exchange of chemical signals, resulting in coordinated gene expression in response to cell density (Bassler, 2002; Taga and Bassler, 2003; Williams, 2007). This process has been extensively studied in Pseudomonas aeruginosa, a Gram-negative environmental bacterium capable of causing acute and chronic infections in immunocompromised individuals. In this organism, QS controls the expression of numerous virulence factors including extracellular enzymes and toxins (Bjarnsholt and Givskov, 2007; Girard and Bloemberg, 2008; Rumbaugh et al., 2000; Smith and Iglewski, 2003).
The P. aeruginosa QS circuitry is comprised of two complete systems, LasR-LasI and RhlR-RhlI. LasI and RhlI are acyl-homoserine lactone (acyl-HSL) synthases, producing 3-oxo-dodecanoyl (3OC12) HSL and butanoyl (C4) HSL respectively, while LasR and RhlR are the transcriptional regulators which bind to their cognate signals to activate target gene expression (Fuqua and Greenberg, 2002; Whitehead et al., 2001). Both systems are connected in a hierachical fashion, as LasR, under standard culture conditions, controls the expression of rhlI and rhlR (Latifi et al., 1996; Medina et al., 2003a; Pesci et al., 1997). Target genes respond to each system with varying degrees of specificity (Schuster et al., 2003; Whiteley et al., 1999; Whiteley and Greenberg, 2001). LasR and RhlR have been shown to recognize conserved palindromic sequences, las-rhl box sequences, of individual QS-controlled promoters, and more such sites have been predicted upstream of other QS-controlled genes (Anderson et al., 1999; Pessi and Haas, 2000; Rust et al., 1996; Schuster et al., 2003; Wagner et al., 2003; Whiteley et al., 1999; Whiteley and Greenberg, 2001). In vitro binding studies with purified LasR showed that some promoters bind LasR cooperatively, while others bind LasR non-cooperatively (Schuster et al., 2004b). Microarray analyses revealed that P. aeruginosa QS constitutes a global regulatory system in which the LasRI and RhlRI systems govern the expression of hundreds of target genes (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003), many of which encode central metabolic functions. In addition, acyl-HSL QS is highly interconnected with other cellular regulatory pathways, resulting in a complex network with the potential to integrate and respond to a multitude of environmental signals (Juhas et al., 2005; Schuster and Greenberg, 2006; Venturi, 2006).
The purpose of this study was to obtain further insights into the complexity of the QS network by separating direct from indirect regulatory effects. Such information is desirable to fully appreciate the mechanism of action of novel anti-virulence strategies that target QS. Here we report the genome-wide identification of direct targets of the central QS regulator LasR using chromatin immunoprecipitation in conjunction with DNA microarray analysis (ChIP-chip) (Aparicio et al., 2005; Buck and Lieb, 2004). LasR-target promoter interactions were independently confirmed by electrophoretic mobility shift assays (EMSA) and transcriptional reporter fusions. EMSA also allowed us to sort promoters according to the mode of binding, and to define putative LasR binding sites. Our experiments confirmed known members of the LasR regulon, and identified five novel LasR-regulated promoters.
While results from microarray studies indicated that at least 6% of the P. aeruginosa genome is regulated by LasRI and RhlRI QS (Schuster et al., 2003; Wagner et al., 2003), many may be indirect targets. We employed ChIP-chip to identify direct targets of LasR in vivo. We used early-stationary phase P. aeruginosa cultures grown in buffered Lennox LB. These conditions are identical to our previous transcriptome analysis, allowing direct comparison of data sets, and they result in high-level induction of QS target genes (Schuster et al., 2003). We chose the wild type strain expressing lasR in its native context to capture protein-DNA interactions under physiologically relevant conditions. The results from two independent ChIP-chip experiments are averaged and presented in Table 1. In total, 35 sites of the P. aeruginosa genome had enrichment values considered significant (average enrichment > 2-fold; p-value <0.0001). All of these sites were in intergenic regions upstream of open reading frames, consistent with LasR’s function as a transcriptional activator. Thus, these sites constitute putative LasR-dependent promoters.
Thirty ChIP-chip enriched regions are associated with genes that transcription studies identified as QS-controlled (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003; Whiteley et al., 1999; Yarwood et al., 2005). We divided these regions into two categories, based on whether independent evidence for direct regulation by LasR is available (Category I) or is not available (Category II). Several ChIP-chip regions are upstream of divergently expressed genes or operons. In some cases only one direction of transcription is QS-controlled, while in other cases both directions are QS-controlled (Table 1). Overall, the 30 ChIP-chip regions are associated with 56 QS-controlled genes. The most overrepresented functional classes are, as expected, secreted factors and protein secretion/export apparatus, although biosynthesis of cofactors, prosthetic groups and carriers, adaptation and protection, and transcriptional regulators are also overrepresented (Fig. 1).
In order to estimate the fraction of las-activated promoters that are directly regulated by LasR, we re-interrogated our transcriptome data (Schuster et al., 2003). Among the set of 315 QS-activated genes, we identified 195 as las-activated. This includes genes that only respond to LasR-3OC12-HSL, and those that respond to LasR-3OC12-HSL, but respond better to both LasR-3C12-HSL and RhlR-C4-HSL. We predicted that these genes are organized into 114 transcript units, which are controlled by 104 promoters (10 bi-directional promoters), and we identified conserved las-rhl box sequences in 29 of those promoters (Schuster et al., 2003). Twenty-five out of the 30 ChIP-chip regions, then, are associated with las-activated promoters. Among the five remaining promoters, two (PA0144 and PA0855) also respond to LasR-3OC12-HSL, albeit below the cut-off chosen in the transcriptome study (≥2.5-fold induction), and three (PA2426, PA2763, and PA5232) were only identified in other studies (Table 1). Therefore, 24% (25 out of 104) of all las-activated promoters identified in our transcriptome study appear to be regulated by LasR directly, and 10 contain a las-rhl box sequence as defined previously.
A small set of the QS target genes identified by microarray analysis have been previously shown to be regulated by LasR directly. The respective promoters should therefore be among the sites enriched by ChIP-chip. Thirteen promoter regions, controlling 43 genes, have been previously identified as direct targets of LasR either by mutagenesis of predicted promoter binding sites, heterologous expression, or biochemically using purified protein. These promoters include hcnABC, lasB, lasI/rsaL, mvfR (pqsR), rhlI, rhlR, rhlAB, qsc117 (PA1869), vqsR, xcpP/xcpR, PA0572, PA3904, and PA4677 (de Kievit et al., 2002; Latifi et al., 1996; Li et al., 2007; Pearson et al., 1997; Pesci et al., 1997; Pessi and Haas, 2000; Schuster et al., 2004b; Whiteley and Greenberg, 2001; Xiao et al., 2006). The primarily rhl-responsive promoter rhlAB, which does not bind LasR in vitro (Schuster et al., 2004b), was included here because it shows significant activation by LasR-3OC12-HSL not only in P. aeruginosa (Schuster et al., 2003), but also in a heterologous host system (Medina et al., 2003c; Pearson et al., 1997). We identified nine of the 13 regions by ChIP-chip (Table 1, Category I). The promoter region of las-specific gene PA0572, which binds purified LasR in vitro (Schuster et al., 2004b), showed enrichment but at levels below the cut-off (p-value <0.0001 but enrichment only 1.9-fold) . The promoter region of PA1869 showed no appreciable enrichment, but it is in fact primarily rhl-responsive and is only mildly activated by LasR-3OC12-HSL (Schuster et al., 2003). Curiously, the rhlI and hcnABC promoters also did not show any enrichment. These genes show substantial activation by LasR-3OC12-HSL alone although they require both LasR-3OC12-HSL and RhlR-C4-HSL for full activation (de Kievit et al., 2002; Pessi and Haas, 2000; Schuster et al., 2003).
Analysis of the ChIP-chip data also revealed five promoters for which there is no previous evidence of QS-dependent regulation (Category III). These five regions are upstream of 18 putative novel quorum-sensing genes. One of these promoter regions, PA2231, is upstream of the psl gene cluster, which has been shown to play a role in biofilm formation (Friedman and Kolter, 2004; Jackson et al., 2004; Matsukawa and Greenberg, 2004). Additionally, as two regions are upstream of divergently expressed genes, it is possible that one or both directions are under the control of LasR.
To further confirm the ChIP-chip results, a subset of promoter regions was selected for in vitro DNA binding analysis using purified LasR. Fourteen ChIP-chip enriched promoters from Categories I, II, and III were chosen. In addition, we included the lasI/rsaL promoter, which had previously been shown to bind LasR in vitro, and the PA2069 promoter, which has been shown to be responsive to RhlR but not LasR (Schuster et al., 2003), and was also not detected by ChIP-chip. A fragment internal to the lasB gene was included as a control for non-specific binding.
Nine of the 14 promoters bound LasR in vitro (Fig. 2). Two distinct binding patterns were observed. Five promoters (plus rsaL) bound LasR as a single-shift complex, while four promoters bound LasR as a multiple-shift complex. These binding patterns were virtually identical to those observed in our previous study for the rsaL promoter and other promoter fragments (Schuster et al., 2004b). In that study, a careful Hill-plot analysis revealed that the formation of single-shift and multiple-shift complexes corresponds to non-cooperative and cooperative LasR binding, respectively. As expected, the RhlR-specific fragment PA2069, and the non-specific intergenic control fragment did not bind LasR specifically. Non-specific binding was generally observed as a smear at concentrations at or above 10 nM. Unexpectedly, however, five promoter fragments identified by ChIP-chip also did not bind LasR in vitro. It is possible that binding in vivo requires additional factors not present in vitro, or that these sites simply represent false-positives. To distinguish between these possibilities, we performed a second ChIP. This time, enrichment by LasR was evaluated using qPCR (Table 2). The promoters upstream of rsaL and PA2305 were included as positive controls, while four intergenic fragments were used to calculate background levels. Four of the five fragments, which were negative in the EMSA experiment, showed significant enrichment using qPCR, suggesting that these sites are true positives. We conclude that LasR binding in vivo is complex and cannot always be reproduced accurately in vitro.
Based on confirmatory EMSA and qPCR data, we calculate a false-positive rate for our ChIP-chip experiment of 8% (only one out of 13 ChIP-chip enriched sites, excluding known Category I targets, was not confirmed by EMSA or qPCR). Typically, the false- positive rate of a successful ChIP-chip analysis ranges from 2 to 10% (Kim et al., 2007).
Next, we tested the prediction that ChIP-chip enriched fragments not previously associated with QS-dependent genes can function as promoters and drive transcription in vivo. Two putative promoter regions from Category III (PA1159, encoding a predicted cold-shock protein, and PA2231, encoding an enzyme involved in psl exopolysaccharide synthesis) were selected for gene expression analysis. The activity of plasmid-borne transcriptional lacZ reporter fusions was evaluated in the P. aeruginosa PAO1 wild type and an isogenic lasR mutant. The rsaL promoter was included as a positive control. The two Category III promoters showed modestly decreased, but statistically significant, expression in the lasR mutant strain as compared to the wild-type (Fig. 3). Re-interrogation of our microarray data, which were obtained under identical culture conditions, revealed that the expression of both genes was modestly QS-dependent, but this difference was below the threshold chosen (Schuster et al., 2003). It is possible that the contribution of lasR to the regulation of these genes is stronger under different growth conditions. Previous studies have shown that PA1159 is up-regulated during biofilm formation (Waite et al., 2006), conceivably in a QS-dependent fashion. Consistent with the notion that additional transcription factors mediate activation, the two promoter fragments were unable to drive LasR-3OC12-HSL-dependent expression of a lacZ reporter in the heterologous host E. coli (data not shown).
The program CONSENSUS (Hertz and Stormo, 1999) was used to identify a common DNA sequence upstream of the ChIP-chip positive promoters. Conserved binding sites, so-called las-rhl box sequences, have been previously identified upstream of several las and rhl-responsive genes (Schuster et al., 2003; Wagner et al., 2003), including those that have been shown to directly bind LasR in vitro (Li et al., 2007; Schuster et al., 2004b). Our ChIP-chip data set provided us with the opportunity to identify a consensus sequence specific to LasR (herein designated las-box) among a much larger set of binding sites. Thirty-four ChIP-chip enriched sites (excluding the false-positive PA2763) plus PA0572 were used as input to generate a LasR consensus sequence (Fig. 4A). The resulting sequence contains the expected CT-(N12)-AG motif, consistent with previous alignments (Schuster et al., 2003; Schuster et al., 2004b; Wagner et al., 2003; Whiteley et al., 1999; Whiteley and Greenberg, 2001). Almost the entire sequence exhibits dyad symmetry, although conservation at some positions is weak. For all regions in Category I, except PA3477 (rhlR), the identified las-box sequences match published sequences. Because CONSENSUS identifies a sequence pattern in every input sequence without assigning significance, we used PATSER to scan all input sequences for those that closely match the consensus sequence. All but four sequences identified by CONSENSUS were re-identified and considered significant by PATSER (weight score >5, p-value <0.0005) (Table 1).
Next, we analyzed sequences according to their mode of binding as determined by EMSA. We used CONSENSUS to identify separate consensus sequences for all nine cooperatively bound and all six non-cooperatively bound sites. The promoter regions that bind LasR in a non-cooperative manner include five fragments from this study (Fig. 2A), as well as rsaL and three others previously shown to bind LasR in vitro, PA0572, PA2591, and PA3904 (Li et al., 2007; Schuster et al., 2004b). The promoter fragments that bind LasR in a cooperative manner include four regions identified in this study (Fig. 2B) as well as two sequences identified previously, lasB OP2 and PA4677 (Schuster et al., 2004b). In both cases, a 16 bp consensus sequence was identified (Fig. 4B, C). Both contain the conserved CT-(N12)-AG motif. Differences at individual positions that would distinguish cooperative from non-cooperative binding sites are not readily apparent. Because cooperatively bound LasR footprints a region of more than 40 bp in length (Schuster et al., 2004b), we re-interrogated all cooperatively bound sites for the presence of two adjacent las-box sequences. However, a second sequence was not identified, suggesting that, if present, it is too degenerate to be recognized by CONSENSUS.
Microarray technology has identified more than 300 genes responsive to acyl-HSL QS in P. aeruginosa, including those regulated by LasR (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003). However, many of these genes are predicted to be further downstream in the QS hierarchy and as such would not be directly regulated by LasR. To identify the direct targets of this DNA-binding protein, ChIP-chip in conjunction with EMSA, qPCR and promoter fusions, was performed. ChIP-chip technology identified 35 regions as direct targets of LasR in vivo. Thirty of these promoters are associated with previously identified QS-controlled genes.
We identified many, but not all, of the promoters previously shown to interact with LasR directly. The only las-specific promoter region we did not identify was PA0572, which bound LasR in vitro (Schuster et al., 2004b). It showed enrichment but at levels below the cut-off chosen. The other promoters, hcnABC, rhlI, and PA1869, all respond to both LasR-3OC12-HSL as well as RhlR-C4-HSL (Schuster et al., 2003). Another promoter that belongs in this category is lasA. It was not identified by ChIP-chip, although the presence of a las-rhl box sequence suggests direct regulation (Schuster et al., 2003). In fact, only 10 of 29 las-responsive promoters with a las-rhl box sequence - as predicted previously - were isolated by ChIP-chip. A closer analysis of the transcriptome data reveals that most of these 19 regions not identified by ChIP-chip respond much better to LasR-3OC12 and RhlR-C4-HSL than LasR-3OC12-HSL alone. In many cases, only one las-rhl box sequence was identified in each of the corresponding promoter regions. This could mean that several of these promoters are directly regulated by RhlR, and only indirectly by LasR. It could also mean that several promoters are directly regulated by both transcription factors, which bind to the same sequence. In the case of hcnABC, the latter has in fact been shown experimentally using a heterologous host system (Pessi and Haas, 2000). Thus it is conceivable that in a stationary-phase wild type cell, in which both LasR and RhlR are expressed at high levels, RhlR displaces LasR at these “single-occupancy” binding sites. Consequently, the respective promoters would not be recognized by ChIP using a LasR-specific antibody. This would be in contrast to lasB, another las/rhl responsive promoter that contains two las-rhl box sequences, one of which is exclusively bound by LasR (Anderson et al., 1999), and that was indeed identified by ChIP-chip. An alternative explanation might be that LasR and RhlR bind to individual sites as heterodimers (as suggested by Pessi and Haas, 2000) or higher-order multimers, which might also preclude enrichment by ChIP using LasR-specific antibodies. Taken together, the available data favor a mechanistic, biological explanation and do not suggest a technical deficiency with ChIP-chip.
Our criteria chosen to filter significantly enriched promoters strike a balance between false-positive and false-negative discovery rates. As mentioned above, we estimated the false-positive rate to be 8%. A false-negative rate can be calculated by considering only las-specific promoters (those that respond to LasR-3OC12-HSL but not RhlR-C4-HSL) and those that bind LasR at distinct sites (i.e. lasB). Taken together, there are nine such Category I promoters for which independent evidence for direct regulation is available. Only one of those, PA0572, was not identified by LasR-ChIP, giving a false-negative discovery rate of 11%. For this calculation, we did not consider promoters that also respond to RhlR-C4-HSL as these may not be recognized by LasR-ChIP for biological reasons, namely binding site competition or oligomerization in vivo, and therefore should not be considered “false-negatives”.
Promoters that have been shown to be directly and exclusively activated by RhlR-C4-HSL or QscR-3OC12-HSL were not identified here, confirming that the immunoprecipitation procedure was highly specific for LasR. This includes lecA (PA2570) and PA1897 (Lee et al., 2006; Winzer et al., 2000). PA1897 is directly regulated by QscR rather than LasR, as had been suggested initially (Lee et al., 2006; Schuster et al., 2004b; Whiteley and Greenberg, 2001).
Furthermore, we identified five novel QS regulated promoters. Two of these, PA1159 and PA2231, were of particular interest to us as they suggested potential involvement of QS in previously unrecognized functions, the cold-shock response and Psl exopolysaccharide synthesis. QS has been shown to regulate the production of another P. aeruginosa EPS, Pel, in strain PA14 (Sakuragi and Kolter, 2007). Other studies have demonstrated that three of these novel QS promoters, PA0805, PA1159 and PA3384, respond to other stimuli, including osmotic stress (Aspedon et al., 2006), tributyltin (Dubey et al., 2006), biofilm formation (Hentzer et al., 2005) and aminoglycosides (Marr et al., 2007).
The ChIP-chip data were also evaluated for the presence of promoters targeted for down-regulation by LasR. The transcript data for down-regulated genes from Schuster et al. (2003) Wagner et al. (2001) are different in size, but suggest that upwards of 200 genes may be down-regulated in a QS-dependent manner. Because none of the genes listed in either study were identified by ChIP-chip, we conclude that LasR does not function as a transcriptional repressor. Rather, LasR appears to mediate repression indirectly by activating a transcriptional repressor. Seven genes annotated as transcriptional regulators were identified by ChIP-chip, including the well characterized MvfR, RhlR, RsaL, and VqsR (for one additional regulator in Category III, AmrZ, QS-dependent transcription has yet to be verified experimentally). RhlR functions as a repressor of rhlAB in the absence of C4-HSL (Medina et al., 2003b). VsqR controls the expression of 580 transcripts, 476 of which are repressed (Wagner et al., 2007), and RsaL controls the expression of 130 transcripts, 120 of which are repressed (Rampioni et al., 2007). RhlR, RsaL, and VqsR may therefore mediate the repression of many QS-repressed genes.
EMSA analysis not only confirmed binding to many ChIP-chip enriched sites, but also provided mechanistic insights into LasR binding to target promoters. We noted previously that LasR exhibits two modes of binding (Schuster et al., 2004b). LasR bound several promoter cooperatively, and a few others non-cooperatively. Cooperative binding may be the result of LasR multimerization. Here we show that these distinct binding patterns extend to many other LasR-controlled promoters. However, six ChIP-chip enriched promoters did not demonstrate any specific binding to LasR, including the four promoters from this study (excluding one false-positive; Fig. 2), and two promoters, rhlAB and pqsA, from previous studies (Schuster et al., 2004b; Wade et al., 2005). This finding is not unique to our work. In another ChIP-chip study, for example, 15% of promoters identified by ChIP-chip also did not bind the purified transcription factor by EMSA (Molle et al., 2003a). Such inconsistencies are not unexpected as EMSA detects binding in vitro while ChIP-chip detects binding in vivo. It is intriguing to speculate that additional transcription factors aid in the recruitment and/or binding of LasR in vivo. RhlR, the stationary phase sigma-factor RpoS (σS), and the LysR-type regulator MvfR may be such factors. Three out of the six ChIP promoters that did not bind LasR in vitro (PA0026, PA4117, and PA4306) show RpoS-dependent activation (Schuster et al., 2004a), two promoters, rhlAB and PA4306, depend on RhlR-C4-HSL for full activation (Table 1), and one promoter, PA0996 (pqsA), binds MvfR in vitro (Wade et al., 2005). Cooperative interactions between these transcription factors and LasR in vivo could explain why purified LasR did not shift several ChIP-chip enriched promoter fragments by EMSA. Taken together, interactions between LasR and RhlR might be particularly complex, involving either antagonism (as described above), synergy, or no interaction at all, depending on the specific promoter architecture.
Previous studies have identified conserved sequences upstream of QS genes, termed las-rhl box sequences, which included las-responsive and rhl-reponsive genes (Schuster et al., 2003; Wagner et al., 2003; Whiteley and Greenberg, 2001). LasR binding sites identified by ChIP-chip in this study provided an excellent source for a more refined characterization of conserved recognition sequences. We performed a bioinformatics analysis that treated promoter sites separately, according to their mode of binding, in addition to an overall analysis of all ChIP-chip enriched sites. All three analyses revealed the general CT-N12-AG motif, but did not confirm the notion from our previous study that cooperative sites are different from non-cooperative sites (Schuster et al., 2004b). For cooperative promoters, we predicted the presence of an elongated sequence or the presence of additional, adjacent binding sequences, but we were only able to identify the standard 16 bp motif. It is possible that a second, more degenerate, sequence is present, which is bound by additional copies of LasR, once the strong consensus sequence has been bound. There appears to be no correlation between conservation of the binding site (particularly the CT-N12-AG motif), binding strength of LasR in vivo (as revealed by the ChIP-chip enrichment factor), and binding strength of LasR in vitro. Interestingly, a las-box sequence with low similarity to the consensus was found in only 14% (2 of 14) of regions that bound LasR in vitro versus 50% (3 of 6) of regions that did not bind LasR in vitro. Thus, LasR recognition sites in promoters that did not bind LasR in vitro may be more degenerate and presumably mediate weaker binding, which is consistent with the notion that LasR requires additional factors for binding to these promoters in vivo.
The location of the LasR binding sequence with respect to the transcriptional start site in individual promoters, as determined previously (Table 1), suggests multiple modes of transcription activation. A proximal location in some promoters suggests that LasR could function as a class II-type activator by making multiple contacts with RNA polymerase (Nasser and Reverchon, 2007). A more distal location in other promoters suggests that LasR could also function as a class I-type activator by contacting only the α-subunit carboxy-terminal domain, or even as an enhancer similar to σ54-dependent promoters (Rappas et al., 2007). As for other LuxR-type transcription factors, it is known that Vibrio fischeri LuxR functions as a class II-type activator (Finney et al., 2002; Johnson et al., 2003), while Agrobacterium tumefaciens TraR functions both as a class I and class II-type activator (White and Winans, 2005).
Taken together, our ChIP-chip data presented here, combined with recent transcriptome data (Schuster et al., 2003), suggest that a principal function of LasR is to directly regulate expression of extracellular factors (Fig. 1). We propose that other functions controlled by LasR, many likely indirectly, help reprogram cellular metabolism for the production of these secreted factors. Only 24% of all las-activated QS promoters showed significant enrichment by LasR-ChIP. This probably underestimates the total number of genes directly controlled by LasR, primarily because several las/rhl-responsive promoters escaped enrichment. If RhlR indeed replaced LasR at “single-occupancy” binding sites, as proposed above, then a second ChIP-chip experiment with a rhlR mutant could potentially reveal additional targets. Nevertheless, the fact that seven directly controlled genes encode transcriptional regulators supports the notion that many genes are regulated by LasR indirectly. Such hierarchical regulatory control, together with co-regulation at the target promoter level - characterized as “dense-overlapping regulon” in a previous study (Schuster and Greenberg, 2007) - allows for a high level of signal integration. We suggest that the resulting network architecture is responsible for the exceptional adaptability of the QS response to different environmental conditions.
Bacterial strains used in this study include the Pseudomonas aeruginosa PAO1 wild-type, the P. aeruginosa mutants lasR::GmR, rhlR::GmR, lasR:: TcR rhlR::GmR (Rahim et al., 2001; Schuster et al., 2003), and Escherichia coli DH5α (Invitrogen, CA). All strains were cultured in Lennox LB medium at 37°C. Where indicated, LB was buffered with 50 mM 3-(N-morpholino) propanesulfonic acid (MOPS), pH 7.0. Antibiotics, when required, were used as follows: P. aeruginosa, 200–300 µg/ml carbenicillin; E. coli, 100 µg/ml ampicillin, 10 µg/ml gentamicin.
A chromatin immunoprecipitation protocol was developed for P. aeruginosa based on published procedures (Horak and Snyder, 2002; Molle et al., 2003a; Molle et al., 2003b), and a protocol available from Roche Nimblegen Technical Services (www.nimblegen.com). Details are as follows: Bacteria were grown in 40 ml of buffered LB to early stationary phase (OD600 = 2.0). Cells were inoculated from mid-logarithmic phase cultures to initial optical densities of 0.01. The experimental strain was P. aeruginosa PAO1, and the non-specific control strain was an isogenic lasR, rhlR double mutant. Ten ml aliquots were cross-linked by the addition of formaldehyde to a final concentration of 1% and incubation at room temperature for 15 min with gentle agitation. The cross-linking reaction was quenched by the addition of fresh glycine to a final concentration of 125 mM and incubation at room temperature for 10 min. Next, the cells were washed twice in 10 ml of cold 1xTBS (20 mM Tris-HCL, pH 7.5, 150 mM NaCl) and the pellets were frozen at −80°C. After thawing, cells were resuspended in 1 ml immunoprecipitation (IP) buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% deoxycholic acid) supplemented with 250x diluted Protease Inhibitor Cocktail (Set III, Novagen) and 1 mg/ml lysozyme. The suspension was incubated at 37°C for 15 minutes. Samples were chilled on ice and sonicated three times for 10 seconds each using a Branson microtip sonicator at an output of 0.4 with no pulse. A 25 µl aliquot from the cleared lysate was set aside as the total DNA control.
A 20 µl aliquot was retained for size analysis by agarose gel electrophoresis to ensure that fragments were in the 300 – 1000 bp range. To prepare samples, they were treated with SDS, Proteinase K, RNaseA, extracted with phenol-chloroform, and precipitated with ethanol as described for IP samples below.
The remaining supernatant was pre-cleared by incubation with one-tenth volume of a 50% protein A Sepharose slurry (nProtein A 4 Fast Flow, Amersham Biosciences) for 1 hour at 4°C. LasR-DNA complexes were precipitated by the addition of 6 µg of affinity-purified LasR rabbit polyclonal antibody to 850 µl of the supernatant and incubation on ice overnight. Then, 50 µl of the 50% slurry were added and incubated at 4°C for 1 h. Subsequently, the complexes were washed five times for 10 min each: twice with IP buffer, and once with each of Buffer II (50mM Tris-HCl pH 7.5, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% deoxycholic acid), Buffer III (10 mM Tris-HCl pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% deoxycholic acid, 0.5% Nonidet P-40), and 1xTBS. Finally, the slurry was resuspended in 100 µl of elution buffer (50 mM Tris-HCl pH 7.5, 10 mM EDTA, 1% SDS) and incubated at 65°C for 15 min to elute the antibody-protein-DNA complexes from the beads. The supernatant was transferred to a new microcentrifuge tube and the beads were washed with another 100 µl of elution buffer, which was added to the eluate. Cross-links between the protein and DNA were reversed by incubating the eluate at 65°C overnight. The total DNA control sample set aside earlier was also incubated with 175 µl of elution buffer overnight at 65°C. To further remove the protein, 200 µl TE supplemented with 100 ug proteinase K and 20 µg glycogen were added to each sample and incubated at 37°C for 2 h. DNA was purified using phenol-chloroform-isoamyalcohol extraction and ethanol precipitation. All samples were resuspended in 30 µl TE containing 0.33 µg/µl RNase A and incubated at 37°C for 2 h. The DNA was then purified using the QIAquick PCR Purification Kit (QIAGEN) and eluted in 50 µl of elution buffer. Blunting of DNA ends, ligation of linkers, and first-round linker-mediated PCR, was performed according to the Roche NimbleGen Chromatin Immunoprecipitation and Amplification protocol, Sections 8 – 10.
P. aeruginosa whole-genome microarrays, including all intergenic regions, were custom-made by Nimblegen and consisted of 60-mers tiled with a spacing of 30 bp. The ChIP DNA and control input DNA were differentially labeled, hybridized to one array, and scanned. Separate arrays were used for wild type (specific) and lasR rhlR mutant (non-specific) samples. These steps were performed by Nimblegen. The raw data were treated as follows (Kim et al., 2005). Intensity data (signals) were normalized within each array and also across all arrays, using Loess normalization. This procedure resulted in comparable array signals for the entire data set in terms of mean signal and variance. After this two-step normalization, replicate probes were fit and averaged using a linear model and Bayes Statistics. A Bayes statistical approach uses a weighted t-statistics that also considers within-treatment variation, which is particularly suitable for a low number of replicates. The non-specific signals were then subtracted from the specific signals to eliminate non-specific effects. This signal difference represents a net-enrichment resulting from specific binding of the respective transcription factor. The so-processed data were viewed in Nimblegen’s SignalMap software, which provides a graphical representation of the hybridization data and allows straight-forward identification of candidate transcription factor binding sites. To assign confidence values to these differentially enriched sites, we performed a two-tailed Student’s t-Test of five consecutive probes between specific and non-specific signals. We considered those sites as significantly enriched that showed p<0.0001 and an enrichment ratio >2-fold.
As mentioned above, we used a lasR, rhlR double mutant as a non-specific control. We chose this mutant instead of a lasR single mutant because our initial goal was to identify genes directly controlled by LasR and RhlR. The lasR, rhlR mutant was intended as a single, economical control of non-specific enrichment for ChIP with either LasR-specific antibody or RhlR-specific antibody. We decided to not pursue the RhlR-ChIP further at this time, because we have not been able to purify soluble, active RhlR for confirmation of ChIP-chip results by EMSA. Regardless, the use of a lasR single mutant as a control is expected to yield results identical to those of a double mutant, because under standard growth conditions such as those used for ChIP-chip, the las system strictly controls expression of the rhl system (Latifi et al., 1996; Medina et al., 2003a; Pesci et al., 1997). A potential problem with the use of any negative control that does not express rhlR is that it would not be able to detect any non-specific binding of the LasR-ChIP antibody to RhlR, if it occurred. Importantly, however, our affinity-purified antibody did not cross-react with RhlR (see below), and LasR-ChIP did not enrich known RhlR-specific promoter fragments, eliminating such concerns. The same is true for QscR, the third LuxR-type regulator in P. aeruginosa.
ChIP replicates were also analyzed by quantitative real-time PCR using the Applied Biosystems 7300 sequencing system as described (Sandoz et al., 2007; Schuster and Greenberg, 2007). Eight promoter fragments (rsaL, PA0026, PA2069, PA2305, PA2763, PA4117, PA4305/6, PA5184), and fragments internal to the lasB, PA0996, PA1914, and PA4677 genes were selected for analysis by qPCR. The latter served as background controls. Primers were designed such that the amplified sequence overlapped with the peak ChIP-chip signal (Supplemental Table 1). Experimental conditions for culture growth, immunoprecipitation, and further sample processing were identical to those described above. Specific enrichment was calculated as the difference in the critical threshold (ΔCt) of immunoprecipitated DNA from the wild type and the lasR rhlR mutant, normalized to the ΔCt of total, non-immunoprecipitated DNA. This approach assumed a direct correlation between Ct and DNA template abundance, which was validated by quantifying the relative abundance of individual templates at defined dilutions using the standard curve method. A baseline was established by calculating the average enrichment of four control fragments, which was subtracted from the specific enrichment. Enrichment was considered significant if the resulting value was > 2-fold, and if the resulting value was > 2 standard deviations above the average enrichment of the background control fragments (Kim et al., 2007). All qPCR reactions were performed in duplicate for two independent ChIP samples.
LasR polyclonal rabbit antiserum was purified by pre-adsorption and affinity chromatography. First, clarified antiserum was pre-adsorbed by incubation with an equal volume of the soluble, filtered fraction (20 × concentrate) of a stationary-phase culture of PAO lasR::GmR. After centrifugation, the supernatant was applied to a Sulfolink column (Pierce) to which purified LasR (Schuster et al., 2004b) had been cross-linked. Column preparation, washing, and elution were according to manufacturer’s instructions. The quality and specificity of the purified antibody was assessed by Western blotting (data not shown). A single band (LasR) was detected in whole-cell lysates of PAO1 wild type cultures, and there was no cross-reactivity with RhlR or with QscR.
Gel-shifts were performed as previously described (Schuster et al., 2004b) with the following modifications. Each reaction contained 25 pM of either a specific or a non-specific DNA probe. In total, 17 promoters were PCR-amplified with fragment sizes ranging from 269 bp to 301 bp. The 146 bp lasB intergenic fragment was used as a non-specific control. Primer sequences used for amplification are in Supplemental Table 1. The location of all fragments (excluding controls) was chosen such that it overlapped with the peak ChIP-chip signal. Promoters of categories II and III, on which the calculation of a false-positive rate is based, were randomly chosen for EMSA analysis.
LacZ reporter fusions were constructed as follows. Upstream regulatory regions of genes PA1159 and PA2231 were PCR-amplified from PAO1 genomic DNA, with primer pairs containing engineered NcoI/HindIII restrictions sites (see Supplemental Table 1). PCR fragments were cloned into the broad-host-range, low copy-number plasmid pQF50 (Farinha and Kropinski, 1990) using identical restriction sites. This plasmid contains a promoterless lacZ gene, allowing for the construction of transcriptional reporter fusions. Constructs were verified by DNA sequencing, and were transformed into the respective P. aeruginosa strains as described (Chuanchuen et al., 2002). An rsaL-lacZ fusion in pQF50 (pMW312) had been constructed previously (Whiteley and Greenberg, 2001).
Mid-logarithmic phase cultures of the respective strains grown in MOPS-buffered Lennox LB were diluted to an OD600 of 0.05 in the same medium and grown for 8 h, shaking, at 37°C. After 8 h, culture aliquots were withdrawn and β-galactosidase activity measured in a microplate reader (Infiniti M200, Tecan) using the Galacton Plus reagent (Applied Biosystems) as described (Whiteley et al., 1999). Reported values are based on three independent biological replicate experiments. Statistical significance was determined by a two-tailed Student’s t-Test. Prior to averaging, background expression from a strain containing pQF50 without a promoter was subtracted from the individual expression values obtained with each promoter fusion construct.
A two-plasmid system was used to assess heterologous gene expression in E. coli DH5α as described (Lee et al., 2006).
The program CONSENSUS (Hertz and Stormo, 1999) was used to identify a LasR consensus sequence within the sets of ChIP-chip sites indicated in Fig. 4. The entire intergenic region was used as input sequence for CONSENSUS if the precise binding site had not been mapped by footprinting. In those cases where footprinting data were available, the mapped sequence was chosen. To search for a consensus sequence within cooperatively bound sites, we seeded with the consensus identified for all ChIP-chip sites, because no seeding only yielded AT-rich regions of low complexity. Consensus motifs were visualized by WEBLOGO v.3 (Crooks et al., 2004). The program PATSER (Hertz and Stormo, 1999) was used to assign statistical significance to the sequences identified by CONSENSUS. The weight matrix generated with CONSENSUS was used as input.
This work was supported by start-up funds from Oregon State University (to M.S.), and by National Institutes of Health grants AI079454 (to M.S.) and GM059026 (to E.P.G.). T.H.K. is supported by the Rita Allen Foundation and the Sidney Kimmel Foundation for Cancer Research.