LuxR-type Quorum Sensing Regulators That Are Detached from Common Scents

Abstract

The ability of LuxR-type proteins to regulate transcription is controlled by bacterial pheromones, N-acylhomoserine lactones (AHLs). Most LuxR-family proteins require their cognate AHLs for activity, and at least some of them require AHLs for folding and protease-resistance. However, a few members of this family are able to fold, dimerize, bind DNA, and regulate transcription in the absence of AHLs; moreover, these proteins are antagonized by their cognate AHLs. Complexes between some of these proteins and their DNA binding sites are disrupted by AHLs in vitro. All such proteins are fairly closely related within the larger LuxR family, indicating that they share a relatively recent common ancestor. The 3′ ends of the genes encoding these receptors invariably overlap with the 3′ ends of the cognate AHL synthase genes, suggesting additional antagonism at the level of mRNA synthesis, stability, or translation.

Bacteria are “scentient” beings. Collectively, they synthesize and detect a rich variety of diffusible chemical signals that impact diverse behaviours, including bioluminescence, horizontal DNA transfer, biofilm formation, pathogenesis, and the production of antimicrobials and other secondary metabolites (Whitehead et al., 2001). New signals, signalling systems, and signal-controlled behaviours are continually being discovered. Chemical communication within populations of bacteria enable them to estimate their population density, a process sometimes referred to as quorum sensing. The ability of bacterial populations to coordinate their behaviour might reflect the difficulty of an individual bacterium to impact its environment. Bacteria give new poignancy to the aphorism “To be is to be perceived” (Berkeley, 1710).

Two general classes of diffusible signals, or pheromones, have been described in bacteria. Communication among Gram-positive bacteria often involves oligopeptides that are detected by membrane-spanning kinases (Dunny and Leonard, 1997) or by cytoplasmic receptors (Dunny, 2007; Perego and Brannigan, 2001). In contrast, signalling among Proteobacteria often involves N-acyl-homoserine lactones (AHLs), which have identical polar head groups and a variety of hydrophobic acyl groups that differ in length, oxidation, and desaturation. AHL signal molecules are often referred to as autoinducers. A number of other bacterial pheromones have also been described, including coumarylhomoserine lactone (Schaefer et al., 2008), γ-butyrolactones (Chater and Horinouchi, 2003), unsaturated fatty acids (Ryan et al., 2009; Vojnov et al., 2001; Wang et al., 2004), a fatty acid methyl ester (Flavier et al., 1997), a quinolone (Dubern and Diggle, 2008), and a substituted alkane (Ng and Bassler, 2009). One signal, denoted autoinducer-2, is synthesized by many species of bacteria, and might therefore serve as an intergeneric signal, though this conclusion remains somewhat controversial (Ng and Bassler, 2009).

This review focuses on members of the LuxR family of transcription factors. Virtually all known members of this family are controlled by AHLs, which are synthesized by LuxI-type AHL synthases. Most LuxR-type receptors require AHLs for function and in at least some cases, AHLs are required for protein folding and protease resistance (Zhu and Winans, 1999, 2001). This review focuses primarily on several atypical members of this family that are active only in the absence of cognate AHLs and whose activity they inhibit. We will briefly highlight recent biochemical and structural advances of three proteins that require AHLs for activity, LuxR, LasR, and TraR. We will then review in greater depth the genetic and biochemical literature on LuxR-type proteins that function only as apoproteins.

LuxR, LasR, and TraR: three representative AHL-dependent transcription factors

LuxR and LuxI are encoded by the bioluminescent bacterium Vibrio fischeri (Engebrecht et al., 1983; Engebrecht and Silverman, 1984). LuxI synthesizes N-3-oxohexanoyl-L-homoserine lactone (OHHL) (Eberhard et al., 1981) while LuxR is an OHHL sensor and an OHHL-dependent transcriptional activator of the luciferase operon (Engebrecht et al., 1983; Engebrecht and Silverman, 1984, 1987). As a population of V. fischeri cells grows in density, the concentration of external OHHL increases. When the concentration of this signal reaches the nanomolar range, its passive efflux from the cells becomes balanced by an influx, so that it can interact with LuxR. LuxR-OHHL complexes bind the promoter of the luxICDABEG operon and activate its transcription (Urbanowski et al., 2004) (Fig. 1).

Fig. 1.

Gene induction at high cell density can be achieved via a diffusible signal that activates a transcription activator (top panel, for example LuxR, LasR, and TraR) or that inhibits a repressor (middle panel, for example, EasR, ExpR1, ExpR2, SmaR). Gene activation at low cell density can be achieved via a signal that activates a repressor (not shown) or by a signal that inhibits an activator (bottom panel, EasR family proteins).

LuxR has two domains, an N-terminal domain that binds pheromone, and a C-terminal domain that binds DNA (Choi and Greenberg, 1991; Hanzelka and Greenberg, 1995). A strain producing the N-terminal domain can sequester exogenous OHHL (Hanzelka and Greenberg, 1995). Such a fragment can also block the activity of full length LuxR by forming inactive heterodimers (Choi and Greenberg, 1992). Taken together, these studies suggest that AHLs stimulate dimerization of the N-terminal domains. A LuxR fragment containing only the C-terminal domain is constitutively active and unaffected by AHLs (Choi and Greenberg, 1991). This indicates that the C-terminal domain contains all the sites essential for LuxR-DNA and LuxR-RNA polymerase interactions. Three amino acids clustered in the C-terminal domain of LuxR are required for positive control of transcription, and are presumed to make direct contacts with RNA polymerase (Egland and Greenberg, 2001; Trott and Stevens, 2001). Experiments using RNA polymerase mutants suggest that LuxR contacts the alpha and sigma subunits (Finney et al., 2002; Johnson et al., 2003; Stevens et al., 1999).

LuxR binds to a DNA binding site called a lux box, that is 20 nucleotides in length and centred 42.5 nucleotides upstream of the transcription start site (Devine et al., 1989; Urbanowski et al., 2004). This sequence has imperfect dyad symmetry, suggesting that the DNA binding domains are multimeric and have a corresponding two-fold rotational symmetry. Bases located 6–8 nucleotides away from the dyad axis were essential for wild type activity, while other bases were less critical (Antunes et al., 2008).

LuxR was purified in a soluble, active form after high-level production of the protein in the presence of OHHL (Urbanowski et al., 2004), and bound to lux box DNA, stimulating the functional binding of RNA polymerase to an adjacent promotor. An important property of LuxR that distinguishes it from LasR and TraR (see below) is that binding of OHHL is reversible (Urbanowski et al., 2004). When LuxR was diluted to sub-micromolar concentrations for gel shift assays, it bound to lux box DNA only when additional OHHL was provided. LuxR bound OHHL non-cooperatively with a dissociation constant of 100 nM (Urbanowski et al., 2004).

The LasR protein of Pseudomonas aeruginosa is a central component of a regulatory web that controls the expression of hundreds of genes, some of which play direct roles in disease. This organism produces the cognate AHL synthase LasI, which synthesizes 3-oxo-dodecanoyl-homoserine lactone (OdDHL), as well as another quorum sensing receptor, RhlR, and its cognate AHL synthase, RhlI (which synthesizes butanoyl-homoserine lactone (BHL), and the so-called orphan receptor QscR, which also detects OdDHL (Brint and Ohman, 1995; de Kievit and Iglewski, 2000; Parsek and Greenberg, 1999; Passador et al., 1993). The regulons of these three receptors are partially overlapping. The LasR/LasI system stimulates production of the RhlI/RhlR system, causing the two P. aeruginosa quorum-sensing circuits to initiate sequentially. Genes that are controlled by LasR or RhlR have been identified by Tn5lac mutagenesis and by transcriptional profiling (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003; Whiteley et al., 1999).

LasR was purified in a soluble form by producing it in the presence of OdDHL as described above for LuxR (Schuster et al., 2004). LasR is dimeric in solution, and binds one molecule OdDHL per subunit. Unlike LuxR, LasR does not detectably release its AHL. It bound to six LasR-dependent promoters in EMSA experiments and by DNase I footprinting (Schuster et al., 2004). It bound some promoters non-cooperatively, causing a DNase I footprint of approximately 30 nucleotides, and bound other promoters cooperatively, causing longer footprints The former promoters were presumably bound by one dimer, while the latter were bound by two. Chromatin immunoprecipitation experiments revealed additional target genes including some encoding transcription factors and others encoding secreted proteins and secretion machinery (Gilbert et al., 2009).

Two groups have described the crystal structure of the LasR N-terminal domain (Bottomley et al., 2007; Zou and Nair, 2009). In the first study, the protein was stabilized and folded during synthesis by adding OdDHL. This fragment was dimeric and each subunit consisted of an alpha-beta-alpha sandwich with the AHL lying between the beta sheet and three alpha helices. OdDHL was entirely buried within the protein, with no contact to bulk solvent. The homoserine lactone moiety and the 1 oxo group made hydrogen bonds with a number of conserved amino acids, while the acyl chain made van der Waals contacts with hydrophobic residues. The 12-carbon acyl moiety of the AHL forms a compact S-shape, which is essential for it to be fully buried within the protein. The second study crystalized LasR with OdDHL or with each of three different triphenyl mimic compounds (Zou and Nair, 2009). Despite the lack of structural similarity between these compounds and AHLs, they aided in the folding and stabilization of LasR, and occupied the AHL binding site.

The TraR and TraI proteins are encoded by Ti plasmids of Agrobacterium tumefaciens. TraI synthesizes primarily 3-oxo-octanoyl homoserine lactone (OOHL), while TraR is an OOHL-dependent activator of genes required for vegetative replication and conjugative transfer of the Ti plasmid (Fuqua and Winans, 1994; Hwang et al., 1994; Li and Farrand, 2000; Moré et al., 1996; Pappas and Winans, 2003; Piper et al., 1993). OOHL is essential for the folding of TraR in vivo, and for resistance to cellular proteases, as the half-life of TraR is increased over 20-fold by OOHL (Zhu and Winans, 1999, 2001). The protein binds to sequences called tra boxes that have dyad symmetry and are approximately 43 or 63 nucleotides upstream of the various transcription start sites (Pappas and Winans, 2003; Zhu and Winans, 1999). TraR binds these sites as a dimer and without cooperativity. Residues on both the N-terminal and C-terminal domains of TraR are essential for positive control and probably make direct contacts with RNA polymerase (Costa et al., 2009; Qin et al., 2009; White and Winans, 2005).

The structure of full-length TraR-OOHL-DNA complexes was solved by X-ray crystallography (Vannini et al., 2002; Zhang et al., 2002). TraR is dimeric and contains one molecule of OOHL per subunit. OOHL is fully embedded within the N-terminal domain of the protein, just as OdDHL is embedded within LasR. The C-terminal domain of TraR contains a bundle of four helices per subunit that binds to half of a tra box (Vannini et al., 2002; Zhang et al., 2002). The N-terminal and C-terminal domains of TraR are connected by a flexible twelve amino acid linker that, if fully extended, would allow the N-terminal domains and C-terminal domains to be separated by over 40 angstroms.

There are several surprising structural differences between LasR and TraR. First, although the 3-oxo groups of both AHLs make water-mediated hydrogen bonds to their respective receptor proteins, these 3-oxo groups and their bound water molecules are highly displaced in TraR relative to LasR (Fig. 2a and 2b). In TraR, the water is hydrogen-bonded by Thr129, while in LasR the water is bonded by Arg61, located on the opposite face of the binding pocket. The remainder of the acyl moiety is also oriented quite differently in LasR compared to TraR (Fig. 2c and 2d). Another surprising difference between LasR and TraR is in their subunit interfaces. The entire N-terminal domains are rotated by 90 degrees in TraR relative to those in LasR. Both proteins have long helices at the C-terminal end of these domains. In TraR, these helices are almost parallel, forming a coiled coil that is the primary dimerization determinant. In contrast, the corresponding helices of LasR are perpendicular and make fewer contributions to dimerization. In TraR, the C-terminal residues of the two NTDs are 10 angstroms apart, while their counterparts in LasR are separated by 30 angstroms.

Fig. 2. Structural comparisons of LasR and TraR N-terminal domains.

A and B. The conformation of OdDHL bound to LasR (A) and OOHL bound to TraR (B). The homoserine lactone group and residue W60 (LasR) and W57 (TraR) are aligned for reference. The 3-oxo group of OdDHL makes a water-mediated hydrogen bond to residue R61 of LasR, while that of OOHL makes a water mediated hydrogen bond to residue T129 of TraR. The remainders of the two acyl chains are located in different parts of the two proteins.

C and D. Dimerization of the two N-terminal domains of LasR (C) and of TraR (D). The left subunits of each dimer are aligned for reference. The C-terminal helices of the two proteins are in red, and are oriented approximately 90° apart in TraR relative to LasR (black arrows).

The structure of a third LuxR homolog was reported in 2006, when the ligand-binding domain of SdiA of E. coli was solved using NMR (Yao et al., 2006). The solubility of the SdiA fragment was enhanced by providing an AHL (octanoyl-HSL) during protein synthesis, as was done for LasR and TraR. The resolution of this structure was insufficient to identify protein-AHL contacts. This fragment was reported to be monomeric in solution, possibly due to the harsh conditions that were required to maintain solubility.

EsaR of Pantoea stewartii is inhibited by its cognate AHL

As described above, a small number of LuxR homologues appear to function only as apoproteins, in that they are fully active in the absence of any AHL, and their activity is blocked by cognate AHLs. Among the best characterized of these systems is the EsaR/EsaI system of Pantoea stewartii (formerly Erwinia stewartii), a pathogen of maize. In 1995, von Bodman and Farrand reported that P. stewartii produced OHHL, the same AHL as made by V. fischeri, and isolated the esaI and esaR genes from a cosmid library (Beck von Bodman and Farrand, 1995). The two genes are transcribed convergently and their reading frames overlap by 21 nucleotides at their 3′ ends (Fig. 3). EsaR represses transcription of its own gene, but does not affect expression of esaI. Disruption of esaI caused a sharp decrease in exopolysaccharide accumulation, and production was restored by adding OHHL (Beck von Bodman and Farrand, 1995). In a subsequent study, it was reported that EsaR mutants overproduce the same exopolysaccharide (von Bodman et al., 1998), indicating that null mutations in esaR and esaI have opposite phenotypes. Apparently, whatever EsaR does to regulate EPS was antagonized by OHHL. An esaR, esaI double mutant had the same phenotype as an esaR mutant.

Fig. 3.

The esaR and esaI genes are expressed convergently (A) and overlap by eight codons (B). The expR1 and expI genes, the smaR and smaI genes, and the yenR and yenI genes have similar orientations and overlaps (B). This convergent and overlapping arrangement of all four gene pairs suggests that the expression of one member of a gene pair might be antagonized by the expression of the other member, either via RNA polymerase collisions, or by hybridization of the two complementary mRNAs.

It was later discovered that apo-EsaR directly represses the promoter of rcsA, a gene required for exopolysaccharide synthesis (Minogue et al., 2005). The fact that EsaR repressed rcsA and esaR was unusual among LuxR proteins, which generally activate transcription. However, many transcription factors can activate some promoters while repressing others, often depending on whether they bind upstream of the target promoter, or within or downstream of the promoter. Both LuxR and TraR have been converted to repressors simply by moving their binding sites, even though they are not known to repress any natural promoters (Egland and Greenberg, 2000; Luo and Farrand, 1999). The fact that EsaR was a repressor could therefore be explained easily. In contrast, the fact that OHHL antagonized EsaR function was completely unprecedented, and in stark contrast to the properties of LuxR, LasR, TraR, and most related proteins. This finding shows that EsaR can fold and bind DNA in the absence of any AHL. It is not clear yet whether fully folded, active apo-EsaR can bind its AHL and if so, whether binding is reversible, as it is for LuxR, or irreversible, as it is for LasR and TraR.

EsaR negative autoregulation was reconstituted in E. coli, where OHHL again blocked repression (Minogue et al., 2002), providing more evidence for a direct interaction between EsaR and the esaR promoter. As might be predicted, EsaR does not require any AHL to remain soluble. Purified EsaR bound to the esaR and rcsA promoters fragments in gel shift assays (Minogue et al., 2002; Minogue et al., 2005). At the esaR promoter, EsaR bound to a DNA sequence that contained an imperfect rotational symmetry. Strangely, the binding site in the rcsA promoter showed little or no symmetry and little or no similarity to the EsaR binding site in the esaR promoter. EsaR bound the esaR promoter as a dimer and without cooperativity (Minogue et al., 2002). Surprisingly, OHHL did not have any effect on binding affinity in gel shift assays, though it did inhibit EsaR binding to its binding site in assays using surface plasmon resonance (Minogue et al., 2002). OHHL also altered intrinsic fluorescence of EsaR tryptophan residues and increased EsaR thermostability, providing additional evidence for a direct interaction (Minogue et al., 2002).

As described above, EsaR was first characterized as a repressor. However data published in 2003 showed that EsaR can also serve as a transcriptional activator, albeit in an artificially reconstituted system. EsaR was produced in an E. coli strain that also expressed the luxICDABEG operon of V. fischeri. EsaR activated transcription of this operon, but only in the absence of OHHL (von Bodman et al., 2003). The related protein ExpR of Pectobacterium carotovorum (see below) had similar properties in this assay. Activation of the lux operon was further enhanced by replacing the LuxR binding site with an EsaR binding site. Although this system was artificial, it proved that EsaR can in principle recruit RNA polymerase to a promoter, which would seem highly unlikely unless it had evolved to do so. A native promoter of P. stewartii was recently discovered that is activated by EsaR (Schu et al., 2009). This promoter expresses a small, noncoding RNA that is divergent from the esaR gene. OHHL blocked the activation of this gene by EsaR. An EsaR binding site located between esaR and esaS is required for repression of the former gene and is thought to be required for activation of the latter. This binding site is centred 60 nucleotides upstream of a possible transcription start site. Therefore, as described above, EsaR can activate or repress gene expression, depending largely on the position of the binding site, but in all cases, OHHL antagonizes EsaR function. The properties of EsaR and related proteins (see below) are summarized in Table 1.

Table 1.

Functions of the EsaR clade of LuxR receptors

Protein	Organism	Cognate AHL	Repressed Genes/Functions	Activated Genes/Functions	References
EsaR	P. stewartii	OHHL	esaR, rcsA	esaS, lux	(Beck von Bodman and Farrand, 1995; Minogue et al., 2005; Schu et al., 2009; von Bodman et al., 2003)
ExpR	P. carotovora	OHHL, OOHL	hydrolytic exoenzymes, pehA, pelC, celV,	rmsA	(Andersson et al., 2000; Cui et al., 2005; Pirhonen et al., 1993)
ExpR	E. chrysanthemi	OHHL	expR	rmsA	(Chatterjee et al., 2005; Cui et al., 2005)
VirR (ExpR2)	P. carotovora	OHHL, OOHL	pehA, pelC, celV		(Burr et al., 2006; Cui et al., 2005; Sjoblom et al., 2006)
SmaR	Serratia sp 39006	BHL	carR, carA, smaR		(Fineran et al., 2005; Slater et al., 2000; Thomson et al., 2000)
SpnR	S. marcescens	OHHL	pig, nucA, surfactant		(Horng et al., 2002)
YenR	Y. enterocolitica	OHHL		yenR, yenS	Tsai and Winans, unpublished

The exoenzyme regulators ExpR of Pectobacterium carotovorum and Erwinia chrysanthemi are inhibited by OHHL and OOHL

P. carotovorum (formerly Erwinia carotovora) and E. chrysanthemi are plant pathogens that macerate plant tissues by releasing enzymes capable of degrading pectate, cellulose, and protein. Expression of the corresponding genes is controlled by several master regulatory proteins. Mutations in the gene for one such regulator, ExpI, abolished expression of all such genes (Pirhonen et al., 1993). ExpI is related to members of the LuxI family. This gene is linked to expR, and two are expressed convergently, and overlap at their 3′ ends (Andersson et al., 2000), reminiscent of esaR and esaI (Fig. 3). Null mutations in expR cause a modest increase in plant tissue maceration and pectate lyase activity, while over-production of ExpR caused a decrease (Andersson et al., 2000).

Three groups identified a second possible copy of expR in the genome sequence of a related bacterium, and used this sequence to identify the orthologous gene in P. carotovorum (Burr et al., 2006; Cui et al., 2005; Sjoblom et al., 2006). In one study, mutants lacking just one of these two receptors still produced exoenzymes at slightly elevated levels. Expression of target genes increased when AHL was added. A strain lacking both receptor proteins produced exoenzymes at high constitutive levels (Sjoblom et al., 2006). This suggests that there are two receptor proteins with overlapping functions, and that both are able to repress transcription of exoenzyme genes. The newly identified receptor (ExpR2) detected a broader variety of AHLs than ExpR1. In a different study using a different strain of P. carotovorum, this second receptor (referred to as VirR) was solely responsible for exoenzyme production, in that a virR mutation fully restored exoenzyme production in an expI mutant. An expR, virR, expI triple mutant was a phenocopy of the virR, expI double mutant, suggesting that ExpR does not play any role in regulating these genes. Surprisingly, exoenzyme production in a virR, expI mutant was still induced at high cell density, rather than constitutive. This observation was unexplained.

E. chrysanthemi has orthologous expI and expR genes, once again transcribed convergently and overlapping at their 3′ ends (Nasser et al., 1998). Mutation of expI abolished production of two AHLs, but did not affect the production of a third one, suggesting the existence of at least one more AHL synthase gene. Mutations of expI and expR had little effect on pectate lyase synthesis, which remained quorum-regulated. This provided additional evidence for a second quorum sensing system. Despite this genetic evidence that ExpR does not (at least solely) regulate pectate lyase genes, the protein was tested for binding to five different pel gene promoters. Apo-ExpR caused a gel shift at all five promoters. For the one promoter tested, OHHL blocked DNA binding in gel shift assays. DNase I protection assays were performed using three of these promoters. In the absence of OHHL, footprints were detected at two promoters. Inexplicably (in light of the gel shift data), addition of OHHL caused far longer and more distinct footprints at these promoters.

In a subsequent study, apo-ExpR was shown to autorepress its synthesis, while OHHL almost fully blocked autorepression (Castang et al., 2006). Purified apo-ExpR bound the expR promoter without cooperativity, forming two high-affinity complexes (containing one or two ExpR dimers, respectively) and four additional slow migrating, low affinity complexes, possibly due to protein aggregation. OHHL largely disrupted these complexes. ExpR protected a region extending from −40 to +6 of the expR promoter from DNase I digestion. This 46 nucleotide region is sufficiently long to account for the two complexes observed in gel shift assays. Similar to the pectate lyase promoters described above, protection was more pronounced in the presence of OHHL than in its absence. This finding is extremely difficult to reconcile with the gel shift data. RNA polymerase bound an overlapping but longer region, probably displacing ExpR, although the authors of that study concluded that both proteins bound simultaneously. Apo-ExpR blocked open complex formation at the expR promoter, while OHHL restored open complex formation. Similar results were obtained using runoff transcription assays (Castang et al., 2006).

Although ExpR of these species represses most known target genes, it can also activate at least one gene, rsmA (orthologous to csrA of E. coli), whose product binds mRNAs and accelerates their degradation (Chatterjee et al., 2005; Cui et al., 2005). Even here, ExpR functions only as an apoprotein, as OHHL blocks activation in vivo and disrupts ExpR-DNA complexes in vitro. Both copies of ExpR are able to activate this promoter (Cui et al., 2005).

SmaR and SmaI act antagonistically to regulate antibiotic biosynthesis

SmaR of Serratia sp. 39006 is yet another example of a LuxR-type protein whose activity is blocked by the cognate AHL. The smaR gene and the cognate smaI gene are expressed convergently (Fig. 3). SmaI synthesizes predominantly butanoyl-HSL (BHL) and smaller amounts of hexanoyl-HSL (HHL). A smaI mutation abolishes the synthesis of the antibiotic carbapenem, the pigment prodigiosin, and several hydrolytic enzymes (Thomson et al., 2000), while a smaR smaI double mutant restores their production (Slater et al., 2003). Apo-SmaR is thought to bind to the promoters of the carR gene and the car (carbapenem) biosynthetic operon and repress their expression, and binding is blocked by BHL or HHL (Fineran et al., 2005; Slater et al., 2003). CarR is also a LuxR homolog and directly activats the car operon. CarR is rather closely related to members of the EsaR family (Fig. S1), which might suggest that its activity could be blocked by cognate AHLs. CarR is often referred to as “AHL-independent, ” as it was able to activate the car operon of P. carotovorum in a strain lacking AHLs (Barnard and Salmond, 2007). CarR is essential for transcription of the Serratia car operon and functions perfectly well in an AHL-defective strain. It still seems possible that CarR could be antagonized by AHLs, although it functions in strains that produce BHL. The question is complicated by some rather intricate regulatory circuits. It would be interesting to express CarR constitutively in a SmaI SmaR double mutant of Serratia and measure the expression of the carR or carA promoters in the presence and absense of exogenous BHL. What makes this question even more fascinating is that a close relative of CarR clearly requires AHLs for activity. The CarR protein of P. carotovorum requires OHHL (synthesized by ExpI) to activate that organism=s car operon (McGowan et al., 1995). The AHL binding domains of the two CarR proteins are 57% identical.

Serratia marcescens strain SS-1 is similar to Serratia sp. 39006 in that its SpnI/SpnR system controls production of prodigiosin, endonuclease, and a surfactant that affects motility (this strain does not synthesize carbapenem) (Horng et al., 2002). SpnR is thought to directly repress target promoters, while the AHL synthesized by SpnI antagonizes SpnR. SpnI and SpnR are not closely related to their counterparts in Serratia sp. 39006 (SmaI and SmaR), and instead are more closely related to EsaR and EsaI of P. stewartii. SpnI (like EsaI) synthesizes primarily OHHL, while SpnR is inhibited primarily by the same signal. SpnR activates the spnR promoter, providing positive autoregulation. Surprisingly, it was reported that SpnR does so more effectively in the presence of OHHL than in its absence. It is not terribly surprising that SpnR could activate some promoters and repress others, as several EsaR-type proteins can do both (see above). However, the finding that OHHL would block SpnR function at one promoter and enhance it at another is unprecedented and counterintuitive. No explanation was offered (Horng et al., 2002). The spnR/I genes are located on at mobile genetic element, and SpnR represses transcription of the Tn3-type transposase of this element (Wei et al., 2006). The idea that a quorum of bacteria would stimulate transposition provides an interesting example of Lamarkian evolution.

Yersinia spp and Pseudomonas syringae

At least two other gammaproteobacteria encode proteins that fall into the EsaR clade. Yersinia enterocolitica encodes YenR and YenI as well as one orphan LuxR homolog, while Y. pestis and Y. pseudotuberculosis (which have extremely similar chromosomes and differ largely in their plasmid content) each encode two LuxR/LuxI systems. The former organism encodes YpeRI and YspRI, while the latter encodes YpsRI and YtbRI. YenRI, YpeRI, and YpsRI are quite similar and are presumed to be orthologous (Fig. S1) while YtbRI and YspRI are also orthologous. All five pairs of genes are transcribed convergently and overlap at their 3′ ends. Unpublished work from the authors= laboratory indicate that YenR and YenI are similar to other members of the EsaR clade. Apo-YenR activates transcription of a non-coding RNA gene, yenS, which is divergent from yenR. Activation is blocked by OHHL. Purified apo-YenR binds to two sites between the two genes, and binding is reversed by addition of OHHL (Tsai and Winans, manuscript in preparation). Given that YpeR and YpsR are extremely similar in sequence to YenR, it is plausible that they have similar properties. The published literature on YpsRI and YtbRI is difficult to interpret, as the two systems may have redundant or overlapping functions and so the phenotypes of single mutants may be masked. It would be interesting to delete all four genes, and individually express YpsR and YtbR from constitutive promoters in the presence or absence of cognate AHLs, measuring expression of a target gene such as the small RNA homologous to yenS.

Many or perhaps all isolates of the plant pathogen Pseudomonas syringae encode members of the EsaR clade. The protein pairs are designated PsyRI in some strains and AhlRI in other strains. In all cases, the pairs of proteins are transcribed convergently and overlap at their 3′ ends. The AHL synthases of this group synthesize predominantly OHHL. Although a number of papers have appeared that describe these proteins, no published study shows whether or not the receptor proteins are antagonized by their cognate AHLs.

Members of this group form a monophyletic clade

A phylogenetic dendrogram of LuxR homologs shows that all the proteins whose activities are blocked by AHLs (indicated using an asterisk) form a single clade (Fig. S1). The closest relative that is known to require AHLs for function is CarR of E. carotovora (McGowan et al., 1995).

Members of the EsaR clade have several similarities in addition to their being antagonized by cognate AHLs. First, all confirmed members are found only within the Enterobacterialaes, which are themselves within the Gammaproteobacteriales. Second, all but one of these proteins preferentially bind OHHL. The exception is SmaR of P. carotovorum, which preferentially binds BHL. Third, while most LuxR-type proteins activate expression of their cognate AHL synthase genes, members of the EsaR subfamily generally neither activate nor repress these genes. The activation of AHL synthase genes is thought to play a role in the tendency of these systems to remain stably in one of the two possible states in the face of subthreshold variations in AHL concentrations, a phenomenon referred to as hysteresis. EsaR-type proteins might have other mechanisms to achieve this (see below).

A fourth similarity between these proteins is they can function as repressors (though some can also activate certain promoters), while most LuxR type proteins are thought to act solely as activators. Target genes that are repressed by a member of the EsaR subfamily would be induced at high population density, just as they are in more conventional quorum sensing systems, as repression is blocked by AHLs, whose concentration increases at high cell densities (Fig. 1, middle panel). On the other hand, genes that are activated by such proteins would be expressed preferentially at low cell densities, where quorum signals are limited (Fig. 1, bottom panel).

A fifth similarity between members of this family is that they are all encoded by genes that are adjacent to their cognate AHL synthase genes, and each pair of genes are convergently transcribed (Fig. 3). In every case, the coding sequences of these gene pairs overlap by a few nucleotides at their 3′ ends. Although the translation overlap is short, the transcription overlap would certainly be longer, possibly encompassing the entire length of the two genes. These genes lack any obvious transcriptional stop signals.

The convergent arrangement of these gene pairs suggests that expression of one gene might antagonize expression of the other. First, RNA polymerases transcribing the two genes could collide, possibly causing one to be dislodged and the other to continue transcription. Similarly, elongating RNA polymerases might dislodge convergently oriented RNA polymerases bound at the opposite promoter. This mechanism of transcriptional dominance, called “sitting duck” inhibition, has been reported previously (Callen et al., 2004). Third, the mRNAs of cognate receptor and synthase genes could form RNA-RNA duplexes. Such duplexes might inhibit the translation of one or both RNAs, or might enhance their degradation, or both (Waters and Storz, 2009). If all duplexed molecules are degraded, then only the more abundant mRNA will persist. Such a contest between the two genes or their mRNAs could contribute to hysteresis. If under some conditions the promoter of a synthase gene were more active than that of a receptor gene, these mechanisms could accentuate the imbalance, locking the system in the high cell density state. Conversely, if the receptor promoter were more active than the synthase promoter, these mechanisms would exacerbate the imbalance, locking the cells in the low cell density state. Given the antagonism between these receptor proteins and their cognate pheromones, it seems striking that the expression of the two convergent genes might also be antagonistic.

How are EsaR-type proteins inactivated by their cognate AHL pheromones?

Mechanistic explanations for inhibition by AHLs are only beginning to emerge. The ability of several EsaR-like proteins to bind DNA is blocked by AHLs (Castang et al., 2006; Cui et al., 2005; Minogue et al., 2002; Minogue et al., 2005). The most obvious way for AHLs to block DNA binding would be to inhibit dimerization of the NTDs, which in turn would inhibit dimerization of the CTDs by cooperative effects. Alternatively, AHLs might cause the two NTDs of a dimer to bind their respective CTDs in such a way that prevents DNA binding. It seems far less likely that a conformational change caused by AHL binding could be propagated through each subunit to its DNA binding domain, especially in light of the long, flexible linkers between the two domains. The predicted linkers of EsaR-type proteins are even longer than those of other LuxR-type proteins (Fig. S2), and probably impart some degree of functional autonomy to the CTDs. It would be extremely useful to determine experimentally whether AHLs can inhibit NTD dimerization. In vivo, AHLs could also alter the rate of proteolysis of these proteins, analogous to the decrease in proteolysis of TraR caused by OOHL (Zhu and Winans, 1999, 2001).

It is tempting to attempt to model the structure of the EsaR NTD, using the crystal structures of TraR and LasR, and to look for differences that could explain the antagonistic effects of AHLs. All members of the EsaR clade contain all the residues that are predicted to make hydrogen bonds with the homoserine lactone group of the pheromone (Fig. S2). Unfortunately, proteins in the EsaR subfamily are only distantly related to LasR and to TraR, and even the latter two proteins have surprising structural differences (see above). Therefore, it would be exceedingly difficult to make any meaningful predictions about the structures of these proteins. This fascinating group of proteins is therefore ripe for further genetic, biochemical, and structural studies.