The quorum-hindered transcription factor YenR of Yersinia enterocolitica inhibits pheromone production and promotes motility via a small non-coding RNA

Summary

The YenR and YenI proteins of Yersinia enterocolitica resemble the quorum sensing proteins LuxR and LuxI of Vibrio fischeri. Apo-YenR activated a gene, designated yenS, that lies adjacent to and divergent from yenR. YenR-dependent expression of yenS was inhibited by endogenous or exogenous 3-oxohexanoylhomoserine lactone (OHHL) a pheromone made by YenI. Purified apo-YenR bound non-cooperatively to two 20-nucleotide sites that lie upstream of yenS. Binding occurred in the absence of (OHHL), and YenR was largely released from the DNA by this pheromone. yenS encoded two non-translated RNAs 169 and 105 nucleotides long that share the same 5′ end but have different 3′ ends. One or both RNAs inhibited the translation and accumulation of the yenI mRNA by binding to a region that overlaps the YenI start codon. A mutation in yenI strongly stimulated swarming motility on the surface of semi-solid agar, while exogenous OHHL completely suppressed this phenotype. Hypermotility in yenI mutants was also suppressed by mutations in yenR or yenS, suggesting that YenS plays a direct, stimulatory role in swarming motility.

Introduction

In recent years, it has become clear that many groups of bacteria can release and detect diffusible chemical signals and can use these signals to co-ordinate a wide variety of behaviours (Fuqua et al., 1994; Dunlap, 1999; White and Winans, 2007; Irie and Parsek, 2008). In some pathogenic bacteria, these signals alter the expression of proteins required for pathogenicity (Bassler, 1999; Winzer and Williams, 2001; Anand and Griffiths, 2003; Eberl, 2006; Girard and Bloemberg, 2008). Proteobacteria generally use acylhomoserine lactones (AHLs), which are normally synthesized by proteins that resemble LuxI of Vibrio fischeri, and are detected by receptor proteins that are similar to the V. fischeri LuxR transcription factor (Whitehead et al., 2001; Novick and Geisinger, 2008; von Bodman et al., 2008).

Most LuxR-type proteins function only in the presence of their cognate AHL (Whitehead et al., 2001). Structural studies of three members of this family have shown that AHLs bind deeply within these receptors and contribute to the overall hydrophobicity of the protein core (Vannini et al., 2002; Zhang et al., 2002; Yao et al., 2006; Bottomley et al., 2007). One such receptor, TraR of Agrobacterium tumefaciens, requires its cognate AHL as a scaffold for its folding. In the absence of this signal, TraR is rapidly degraded by the cellular proteases Clp and Lon (Zhu and Winans, 1999; 2001). Some members appear to require the cognate AHL for folding, yet once folded, can bind and release the AHL, and must not require it continuously to remain folded (Urbanowski et al., 2004; Yang et al., 2009).

Yersinia enterocolitica is a gammaproteobacterium that colonizes the small intestine and can cause gastrointestinal distress and can also cause septicemia in immunocompromised patients (Matsumoto and Young, 2009). Its YenI protein synthesizes primarily 3-oxohexanoylhomoserine lactone (OHHL) and lesser amounts of hexanoylhomoserine lactone (HHL) (Throup et al., 1995; Atkinson et al., 2006), which were presumed to regulate the activity of its YenR protein. YenI and YenR orthologues are found in other species of Yersinia, including Y. pestis, the causative agent of Bubonic plague. A yenI mutation of Y. enterocolitica caused several changes in the organism's proteome that were detected by 2D gel electrophoresis (Throup et al., 1995), although the altered proteins were not identified. A yenI mutation also caused a delay in swimming motility and abolished swarming motility (Atkinson et al., 2006). This mutation also abolished the accumulation of the major flagellin protein and the corresponding mRNA. However, exogenous addition of OHHL or HHL did not suppress these phenotypes.

A small number of LuxR homologues are active only in the absence of their cognate AHLs. The EsaR protein of Pantoea stewartii is a repressor of a gene required for exopolysaccharide biosynthesis, and also autorepresses its own synthesis (Minogue et al., 2002; 2005; von Bodman et al., 2003). EsaR can also activate the esaS promoter (Schu et al., 2009) and the heterologous luxI promoter (von Bodman et al., 1998). In all four cases, EsaR is active as an apoprotein and its activities are blocked by the cognate AHL. ExpR of Pectobacterium spp., SmaR of Serratia sp. and EanR of Erwinia ananatis have similar properties (Cui et al., 2005; Fineran et al., 2005; Castang et al., 2006; Sjoblom et al., 2006; Morohoshi et al., 2007). EsaR, ExpR and SmaR are part of a monophyletic clade within the larger family of LuxR-type proteins (Andersson et al., 2000; Tsai and Winans, 2010), suggesting that the ability of these proteins to function only as apoproteins may have evolved just once.

YenR closely resembles the EsaR, EanR, ExpR and SmaR proteins described above (46%, 46%, 50% and 40% identical respectively). This similarity suggested to us that YenR might function only as an apoprotein. The current study was initiated in an effort to identify promoters that are regulated directly by YenR, and led to the unexpected discoveries that (i) YenR activates expression of a small non-translated RNA-designated YenS, (ii) that YenS inhibits the production of the cognate AHL synthase, and (iii) that YenS could have one or more additional targets involved in motility.

Results

Immunoprecipitation of YenR–DNA complexes

The primary goal of this study was to identify YenR-regulated promoters. As described above, the closest YenR homologues include EsaR, EanR, ExpR and SmaR, and evidence has been published indicating or suggesting that these proteins bind DNA only as apoproteins. We therefore hypothesized that YenR might have similar properties (this hypothesis was later confirmed, see below). Accordingly, we performed chromatin immunoprecipitation experiments in a strain expressing YenR and lacking YenI. The yenI null mutation effectively locked this quorum sensing system into a low-cell-density state regardless of the actual cell density.

We cultured strain YW003 (a yenI derivative of Y. enterocolitica JB580) to an OD₆₀₀ of 0.3, cross-linked all DNA bound proteins to DNA using formaldehyde, and immunoprecipitated DNA-YenR complexes with an anti-YenR antiserum. The precipitated DNA was PCR amplified, cloned into a plasmid vector, and 28 clones were sequenced (Table S1). One such clone consisted of an intergenic region extending from 132 nucleotides upstream of the yenR translation start site to 335 nucleotides upstream. The present study focuses on this intergenic region.

We performed a reconstruction experiment, in which immunoprecipitated chromatin was PCR amplified using primers that amplify a region upstream of yenR. This DNA was found primarily in the precipitated fraction in YenR⁺ strain (Fig. S1A, lane 7), while only trace amounts were detected in a YenR^- strain (lane 5). This fragment was also not detected in mock immunoprecipitations in which the antiserum had been omitted (Fig. S1A, lanes 4 and 6).

To demonstrate a direct interaction between YenR and this DNA sequence, we performed electrophoretic mobility shift assays (EMSA) using a radiolabelled DNA fragment and purified apo-YenR. YenR formed two major complexes with this DNA fragment (Fig. S1B), suggesting that it may bind to two sites (confirmed below), and lower amounts of a third complex that migrated slightly faster than complex 2. Detection of this third complex was variable and was not pursued.

YenR binds to two sites upstream of yenR

To localize the YenR binding sites, we conducted DNase I protection assays using a fluorescently labelled DNA fragment and purified apo-YenR. YenR strongly altered the protection patterns of a 55 nucleotide region that lies 123–178 nucleotides upstream of the yenR translation start site (Fig. 1A). Four hypersensitive sites were identified within this region, similar to those found in footprints of the TraR and CepR proteins (Zhu and Winans, 1999; Weingart et al., 2005). These non-protected sites suggest that YenR binds to one face of the DNA and does not protect the opposite face. Similar conclusions were reached with the TraR protein (Zhu and Winans, 1999), and were later confirmed by X-ray crystallography (Vannini et al., 2002; Zhang et al., 2002).

Figure 1. Identification of the sites bound by YenR. A. DNase I footprinting of the region upstream of yenR. A DNA fragment extending from 82 nucleotides downstream of the yenR translation start site to 327 nucleotides upstream was synthesized by PCR amplification using oligonucleotides o262 and o263, the former of which contains the fluorophore 6XFAM. This fragment was subjected to DNase I digestion in the absence (top) or presence (bottom) of purified YenR. Fragments were analysed by automated capillary gel electrophoresis. B. Ten duplex DNA fragments (1–10) were radiolabelled, combined with purified apo-YenR, and tested for binding using EMSA (lower panels). YenR was added at the following concentrations in each EMSA assay: 5.44, 16.3, 48.9, 146, 440, 1322 and 0 nM.

We noticed two 20-nucleotide sequences within the footprinted region that show imperfect dyad symmetry and that are similar to each other in sequence. To determine whether these sequences served as YenR binding sites, we synthesized 10 duplex oligonucleotides, each 24 nucleotides in length that tiled the footprinted region (Fig. 1B). Of these 10 fragments, fragment 3 fully contains the left sequence, while fragment 8 fully contains the right sequence. These two fragments bound YenR with high affinity (the K_d's for DNA binding were 40 and 150 nM respectively), while the remaining eight fragments bound very weakly or not at all (Fig. 1B). These sites were designated yen box I and yen box II. The presence of two independent YenR binding sites supports our detection of two YenR–DNA complexes in Fig. S1B.

Several other LuxR-type transcription factors bind to sites that show dyad symmetry (Minogue et al., 2002; Vannini et al., 2002; Zhang et al., 2002; Schuster et al., 2004; Urbanowski et al., 2004; Weingart et al., 2005; Castang et al., 2006). To further characterize these putative binding sites, we constructed a total of four artificial fragments that were perfectly symmetric. The first replaced the right half of yen box I with the inverse complement of the left half (Fig. S2, Box I L-L′). The second replaced the left half of this site with the inverse complement of the right half (Fig. S2, Box I R′-R). The remaining two artificial sequences were similar to the first two but using yen box II as the template sequence (Box II, L-L′, and Box II, R′-R). The two wild-type and four symmetric sequences were tested for YenR affinity. All six fragments bound YenR with high affinity. Interestingly, all three sequences based upon Box I showed similar affinities, while all three sequences based upon Box II also bound with similar affinities that were slightly lower than the first three. We have aligned the four half-sites and derived a consensus (Fig. S2B). Of the four synthetic binding sites, sequence 5 (box II L-L′) most closely resembles the consensus, yet is not bound with higher affinity than the other sites. More work will need to be done to identify the optimal YenR binding site.

The two yen boxes are centred 25 nucleotides apart, or 2.5 helical turns. In the DNase I footprint, there were no sensitive sites between the two binding sites, indicating that the two YenR dimers must lie very close together, and could contact one another. It therefore seemed plausible that YenR might bind yen box I and yen box II cooperatively. This was tested using quantitative EMSA and a fragment containing both binding sites. The fragment was shifted by YenR, forming one well defined complex (Fig. 2A, complex C2) and a fainter complex that migrated farther down the gel (Fig. 2A, complex C1). The dissociation constant (K_d) was about 25 nM. Binding did not show detectable cooperativity, as the Hill coefficient was approximately 1 (Fig. 2B). This indicates that YenR bound to one site does not affect the binding affinity at the other site.

Figure 2. YenR activity in vitro and in vivo in the presence and absence of AHLs. A. Gel mobility shift assays using a DNA fragment containing both yen boxes (made by hybridizing oligonucleotides o308 and o309). The protein concentration for each lane is: 0, 0.09, 0.28, 0.87, 2.7, 8.4, 26, 81, 250 and 774 nM. Abbreviations: C2, complex 2; C1, complex 1; F, free DNA. B. Hill plot of the EMSA data shown in (A). The slope is approximately 1, indicating that binding is non-cooperative. C. Disruption of YenR–DNA complexes by OHHL. Complexes were formed between apo-YenR and a DNA fragment containing both binding sites (the same fragment as in part A). OHHL was then added at various concentrations, incubated for 30 m, and analysed by EMSA. OHHL caused some complexes to dissociate so that the DNA fragment migrated as free DNA, while the remaining complexes migrated more slowly than in the absence of OOHL, and some did not enter the gel, suggesting that the complexes had aggregated. The highest OHHL concentration was 3 µM and was serially diluted in twofold increments. Abbreviations: O, gel origin; C2, complex 2; C1, complex 1; F, free DNA; NC, negative control DNA. D. Inhibition of yenS expression in vivo by AHLs. β-galactosidase activities of the yenS–lacZ transcriptional fusion in strain CST216 (yenI^- yenR⁺lacZ^-) containing plasmid pCST2395 cultured in the presence of the indicated concentrations of OHHL (circles) or HHL (triangles). All assays were done in triplicate in AB defined medium.

Binding of YenR to DNA is altered by addition of OHHL

As described above, YenR binding was detected in vivo and in vitro in the absence of AHLs. We tested whether OHHL could influence binding of purified YenR to DNA. Complexes were made using apo-YenR and a DNA fragment containing two yen boxes (the same fragment as used in Fig. 2A). OHHL was then added at a range of concentrations and incubated for 30 min prior to size-fractionation by EMSA. In the absence of OHHL (Fig. 2C, lane 2) most of the DNA was in complex 2, while far smaller amounts of DNA were in complex 1 or unbound (Fig. 2C). As OHHL was titrated in, both complexes became less abundant, and increasing amounts of the DNA became unbound. In addition, some DNA molecules migrated progressively more slowly, or formed aggregates that failed to enter the gel. Apparently, OHHL caused some complexes to dissociate, and caused other complexes to aggregate. In either case, OHHL disrupted soluble YenR–DNA complexes. Similar results were obtained by adding OHHL to YenR prior to addition of DNA (data not shown).

YenR does not significantly autoregulate

The finding that YenR binds to a site close to the yenR promoter suggested that this protein might regulate one or more promoters in this region. We constructed a single-copy yenR–lacZ fusion (in pCST2399) containing 248 nucleotides of yenR extending from 179 nucleotides upstream of the translation start site to 69 nucleotides downstream. This fusion was introduced into the ara chromosomal locus by double homologous recombination (Maxson and Darwin, 2005), using strains CST216 (yenR⁺, yenI^-, lacZ^-) or CST214 (yenR^-, yenI^-, lacZ^-). Synthesis of β-galactosidase by this fusion was not affected by the yenR or yenI status of the cell or by addition of OHHL (Table 1). YenR therefore does not regulate its own synthesis.

Table 1.  The role of YenR, OHHL and yen boxes in the expression of yenS and yenR.

Integrated plasmid	Fusion	yen box I	yen box II	CST216 yenR^+, yenI^-, lacZ^-	CST216	CST214 yenR^-, yenI^-, lacZ^-	CST214
a.  β-Galactosidase-specific activity. Data shown in the average of triplicate experiments with standard deviations as shown. All assays were done using AB defined medium (Cangelosi et al., 1991). b.  yen box II was altered from AACTAGACCTAAGGCTACGT to AAGAAGACCTAAGGCTACGT.
				No OHHL	1 µM OHHL	No OHHL	1 µM OHHL
pCST2399	yenR–lacZ	+	+	385 ± 70^a	317 ± 35	343 ± 50	373 ± 37
pCST2395	yenS–lacZ	+	+	1895 ± 65	618 ± 25	75 ± 11	50 ± 8
pCST2396	yenS–lacZ	Half	+	964 ± 159	195 ± 30	88 ± 8	64 ± 16
pCST2397	yenS–lacZ	−	+	942 ± 59	30 ± 5	68 ± 6	40 ± 12
pCST2398	yenS–lacZ	−	half	44 ± 12	48 ± 12	55 ± 21	35 ± 4
pCST2410	yenS–lacZ	+	altered^b	1008 ± 112	169 ± 35	55 ± 3	23 ± 2
pCST2411	yenS–lacZ	−	altered^b	60 ± 3	48 ± 2	60 ± 8	31 ± 11

Identification of two conserved DNA sequences upstream of yenR

In an effort to determine whether the yen boxes were conserved across Yersinia spp., we aligned these regions using BLASTN. As expected, these two sites were conserved (Fig. S3). To our surprise, we found that the conservation extended far further upstream. The two yen boxes lie at the yenR-proximal end of a 250-nucleotide conserved region (Fig. 3A, grey boxes and Fig. S3). Upstream of this sequence is a 210-nucleotide non-conserved region, followed by a second conserved region of approximately 160 nucleotides (Fig. S3). The non-conserved regions suggest that in the absence of genetic selection, genetic drift between these species has been extensive. This in turn indicates that the observed conservation is probably attributable to some conserved and adaptive function. In this study, we focus on the 250 nucleotide conserved region.

Figure 3. Genetic map of yenI, yenR and yenS. A. The DNA fragment recovered by chromatin immunoprecipitation is indicated (CHIP). The two regions conserved in five Yersinia spp. are indicated with two grey bars. B. A yenR–lacZ (pCST2399) and six yenS–lacZ fusions containing progressive resections or other alterations of yen box I and yen box II. The transcriptional start sites and promoter motifs of both genes are indicated using open squares.

Apo-YenR activates a gene divergent from yenR

To determine whether the 250-nucleotide region is transcribed, we constructed a transcriptional fusion between it and lacZ, oriented so as to measure transcription divergent from yenR (pCST2395, Table 1). This fusion was integrated into the chromosome at the ara locus, using strains CST216 (yenR⁺, yenI^-, lacZ^-), and CST214 (yenR^-, yenI^-, lacZ^-). The resulting two strains were cultured in the presence or absence of 1 µM OHHL. In the absence of OHHL, the yenR⁺ strain synthesized about 25-fold more β-galactosidase than the yenR^- strain. Addition of OHHL reduced expression about threefold in the yenR⁺ strain, but had little or no effect on the yenR mutant. Evidently, apo-YenR activated expression of a gene, designated yenS, while OHHL antagonized YenR. We used the same fusion to measure the inhibitory effects of varying amounts of OHHL and HHL. Both AHLs inhibited expression of this fusion (Fig. 2D). OHHL seemed to inhibit YenR activity at slightly lower concentrations than HHL.

yen boxes I and II act cooperatively to activate yenS

To determine whether yen boxes I and II are required for induction of yenS, we compared six suicide plasmids that contained or lacked these sites (Fig. 3B). These plasmids were integrated into the ara locus by double recombination, using CST216 (yenR⁺, yenI^-, lacZ^-) and CST214 (yenR^-, yenI^-, lacZ^-). Four of these plasmids make stepwise resections of the two yen boxes, while two contain point mutations in yen box II (from AACTAGACCTAAGGCTACGT to AAGAAGACCTAAGGCTACGT). All six plasmids had similar levels of transcription in the absence of YenR. In the absence of YenR, the addition of OHHL may have caused a slight decrease in expression. The significance of this is unknown. It could be due to a second LuxR homologue encoded by this strain (YE1026A, 43% identical to YenR), but if so, such a protein would have to act in the absence of the two yen boxes.

In the strain synthesizing YenR but not YenI, several fusions were expressed at elevated levels (Table 1). As described above, the fusion containing yen box I and yen box II (pCST2395) was induced about 25-fold. The fusion containing half of yen box I and all of yen box II (pCST2396) was induced about 11-fold, while the fusion that contained just yen box II (pCST2397) was induced about 14-fold. Apparently, YenR can activate this promoter by binding to just yen box II. The fusion containing only half of yen box II (pCST2398) was not affected by YenR (Table 1).

Of the two fusions with point mutations in yen box II, one lacked yen box I, and was not induced by YenR (Table 1, fusion pCST2411). The other of these two fusions contained yen box I (pCST2410); this fusion was inducible. These data indicate that YenR can activate yenS by binding to just yen box I. Conversely, as seen above, YenR can also activate the yenS promoter by binding to just yen box II.

All fusions that were induced by YenR were also inhibited by OHHL. However, in some cases, inhibition was rather mild, while in other cases it was more severe. For example, in the fusion containing both yen boxes (pCST2395), OHHL reduced expression only about threefold (Table 1). In the strain lacking half of yen box I (pCST2396), yenS expression was inhibited about 5-fold by OHHL, while the fusion containing yen box I and an altered yen box II (pCST2410) was inhibited about sixfold. In contrast, induction of the fusion containing just yen box II (pCST2397) was completely blocked by OHHL.

Expression of yenS is enhanced at low cell population density

All the in vivo experiments described above were done using a yenI mutant, which locks the system into a low population density state. Addition of exogenous AHLs mimicked conditions of high cell density. Our findings predict that in a YenI-proficient strain, yenS would be expressed preferentially at low cell density. To test this, we introduced a plasmid containing a PyenS–gfpmut3.1 (LAA) fusion (pCST2406) into strain JB580 (wild-type), CST201 (yenR^-yenI^-) and CST206 (yenR⁺, yenI^-). These three strains were diluted 10⁸-fold and cultured to stationary phase. As predicted, strain CST201 expressed the fusion at low levels during all stages of growth, while CST206 expressed the fusion at high levels (Fig. 4). In contrast, the wild-type strain expressed the fusion at moderately high levels at low population densities, and at lower levels as the culture density increased (Fig. 4, centre panels).

Figure 4. Preferential expression of yenS at low cell population densities. Strains containing a plasmid borne PyenS–gfpmut3.1 (LAA) fusion was diluted to 10 cells per ml, and cultured to stationary phase. At the indicated optical densities, samples were withdrawn and 5000 cells were assayed for the distribution of fluorescence.

yenS encodes two transcripts with identical 5′ ends and differing 3′ ends

Northern blotting experiments were conducted to determine the size of the yenS transcript (Fig. 5). We used four different oligonucleotide probes in order to give an approximate location of the upstream and downstream ends of this transcript. RNA was prepared from two strains, one yenR⁺, yenI^- (CST206) and the other yenR^-, yenI^- (CST201). The former strain was cultured in the presence and absence of OHHL. Using probe o451 (second panel in Fig. 5), two RNA species were detected, approximately 100 and 170 nucleotides in length. They were detected in the yenI mutant cultured in the absence of OHHL (lane 1), but not from the same strain cultured in the presence of OHHL (lane 2) or in the yenR^-yenI^- double mutant (lane 3). Probe o673 hybridized to the larger RNA but not the shorter one (Fig. 5, third panel), while probes o698 and o699 did not hybridize to either RNA (Fig. 5, first and last panel). These data indicate that yenS lies fully between probes o698 and o699, and that it encodes two RNA molecules that differ at their 3′ ends. The longer transcript was designated YenS while the shorter one was designated YenS'.

Figure 5. Identification of two RNA molecules encoded by yenS. Northern hybridizations of YenS transcripts using oligonucleotide o698 (panel 1), o451 (panel 2), o673 (panel 3) and o699 (panel 4). In each panel, lane 1: CST206 (yenR⁺, yenI^-); lane 2: CST206 cultured in the presence of 100 nM OHHL; lane 3: CST201 (yenR^-, yenI^-). Bacteria were cultured using LB medium.

The yenR and yenS transcription start sites were localized using primer extension and 5′ RACE assays (Figs S4 and S5). The yenS start site lies 44.5 nucleotides downstream of the centre of yen box II and 69.5 nucleotides downstream of yen box I (Fig. 3B). In contrast, the two yen boxes are centred over 110 nucleotides upstream of the yenR start site. This may help to explain why YenR had no effect on yenR expression.

The prediction that YenS and YenS' share a single 5′ end but have two different 3′ ends was tested by using 3′ RACE assays. As predicted, two different 3′ ends were identified. One was located 105 nucleotides downstream from the start site, while the other was located 169 nucleotides downstream (Fig. S3). YenS and YenS′ transcripts are therefore 169 and 105 nucleotides long, respectively, in close agreement with the Northern assays described above.

yenS encodes a non-translated RNA

We scanned the yenS gene for open reading frames (Fig. S6). There are no ATG or GTG codons in any reading frame, and the only TTG codon is followed by two sense codons and a stop codon. Near the 3′ end of yenS, there are three short overlapping reading frames beginning with ATG (Fig. S6). Two of these are extremely short (22 codons each), while the third is somewhat longer (66 codons). For three reasons, it seemed highly unlikely that any of these ORFs was translated. First and most important, all three extend beyond the 3′ end of the yenS transcript. Second, none of these ORFs has an apparent ribosome binding site. Third, none is conserved among different species of Yersinia. We nevertheless determined whether any of these ORFs is translated, by constructing plasmid-based lacZ fusions, one transcriptional and three translational (plasmids pCST2053-56) and introducing them into a yenR⁺, yenI^- strain and a yenR^-, yenI^- strain. Each fragment contained the yenS promoter and both yen boxes. None of the fusions was significantly expressed in either strain (data not shown). We conclude that the yenS transcript is not translated.

YenS inhibits the expression of yenI and blocks production of AHLs

Many small, non-translated RNAs regulate gene expression by binding to complementary mRNAs, often blocking their translation and/or altering their stability to cellular RNases (Majdalani et al., 2005). We used the algorithm TargetRNA (Tjaden, 2008) to try to predict possible mRNAs that might interact with YenS. Surprisingly, the yenI gene appeared at the top of a ranked list of possible targets (Fig. S7). In a similar search using the putative YenS orthologue of Y. pestis, the gene corresponding to yenI also received the highest score (Fig. S8). The predicted alignment between the YenS and YenI RNAs is shown in Fig. 6A.

Figure 6. YenS targets yenI mRNA. A. Proposed duplex formation between YenS RNA and YenI mRNA, as predicted by the TargetRNA algorithm. The YenI start codon is underlined. B. Decrease in the abundance of YenI mRNA by YenS. In each panel, lane 1 is the radioactively radiolabelled DNA probe without nuclease digestion. The remaining lanes show the abundance of oligonucleotides that were resistant to nuclease S1 due to hybridization with Yenl mRNA (upper panel) or 23S rRNA (lower panel); (2) CST224 (yenR yenS double mutant); (3) CST226 (wild-type); (4) CST232 (yenS mutant); (5) CST234 (yenR mutant); (6) CST206 (yenI deletion mutant). Ratio is a measure of [YenI]/[23S rRNA] for each strain/[YenI]/[23S rRNA] for the wild-type strain.

To determine whether YenS alters the expression of YenI, we constructed a yenI–lacZ translational fusion in strains containing or lacking yenS. Expression of the fusion was lower in a strain expressing YenS than in one lacking YenS (Table 2). Inhibition of the fusion by YenS required YenR, as a yenR mutant expressed the fusion as strongly as a yenS mutant. This was expected, as yenS expression requires YenR. S1 nuclease protection assays and Q-PCR assays also showed that yenI mRNA is less abundant in strains containing YenR and YenS than in any single or double mutants (Fig. 6B, Table 2, right column).

Table 2.  Negative regulation of the expression of a yenI–lacZ translational fusion^a by YenS.

Strain	Genotype	β-gal activity^b	Relative expression	YenI–lacZ mRNA^c
a.  All strains contained pCST2058, which encodes a yenI–lacZ translational fusion. b.  Miller units (Miller, 1983). c.  Relative YenI abundance was calculated from the 2^−ΔΔCT method by using 23S rRNA as a reference gene and CST226 as a calibrator (Livak and Schmittgen, 2001). The PCR primers used are listed in Table S5. ND, not determined.
CST226	yenR^+, yenS^+	52 (6)	(1)	(1)
CST224	yenR^-, yenS^-	573 (77)	11.0	7.7 (1.7)
CST232	yenR^+, yenS^-	545 (17)	10.5	6 (2.3)
CST234	yenR^-, yenS^+	418 (101)	8.0	6.8 (2.1)
CST224 (pPZP100)	yenR^-, yenS^- (vector)	584 (37)	11.2	ND
CST224 (pCST2099)	yenR^-, yenS^- (yenR, yenS)	61 (4)	1.2	ND

The finding that YenS inhibits the expression of YenI predicts that YenS should also inhibit the production of extracellular AHLs. To test this, stationary phase cultures of a wild-type strain and a yenS mutant were diluted 5000-fold into fresh medium and cultured to mid-log phase. At intervals, samples were removed and bioassayed for AHLs. As predicted, the wild-type strain produced only low levels of AHL until it had reached an OD(600) of approximately 0.1, while the yenS mutant produced readily detectable levels of AHLs at far lower optical densities (Fig. 7).

Figure 7. Inhibition of AHL synthesis by YenS. A wild-type strain (triangles) and a yenS deletion mutant (circles) were diluted 5000-fold into fresh growth medium and cultured to mid-log phase. At intervals, samples were removed and bioassayed for extracellular AHLs. A standard dose–response curve was constructed using synthetic OHHL.

The hypothesis that YenS binds YenI mRNA near the latter's translation start site predicts that inhibition of a yenI–lacZ fusion by YenS would require this site. We constructed a series of transcriptional yenI–lacZ fusions that contain or lack this region. Fusions lacking the putative YenS binding site (Table 3, plasmids pCST2123-2124) were not affected by YenS, while expression of fusions containing all or part of the YenS binding site (plasmids pCST2127, pCST2142 and pCST2143) was decreased by YenS. We conclude that inhibition requires part of the region of YenI predicted to bind YenS, although evidently the entire binding site is not essential.

Table 3.  Activity of yenI–lacZ fusions containing or lacking the putative YenS targeting site in strains containing or lacking YenS.

	Sequences in each fusion^a	CST213 (yenR^+yenS^+)	CST230 (yenR^+yenS^-)	CST217 (yenR^-yenS^-)
a.  sequence numbering is relative to the first base of the yenI translation start codon. YenS is predicted to bind to nucleotides −7 to +13 of the YenI mRNA (Fig. 6A). b.  Miller units (Miller, 1983). The data are representative of three experiments.
Transcriptional fusions		β-Galactosidase activity^b
pCST2123	−412 to −41	27	20	21
pCST2124	−412 to −22	203	220	213
pCST2127	−412 to +6	58	219	176
pCST2142	−412 to +9	48	95	104
pCST2143	−412 to +30	55	91	119
Translational fusions		β-Galactosidase activity
pCST2132	−412 to +6	60	321	321
pCST2144	−412 to +9	3	21	25
pCST2145	−412 to +30	3	16	17

The inhibition of transcriptional yenI–lacZ fusions by YenS described above would presumably have occurred by changes in the abundance of mRNA. The effects of YenS on these fusions were somewhat subtle, about twofold. We hypothesized that if YenS can block the translation of the remaining YenI mRNA, then translational fusions would be affected more strongly. This was indeed the case (Table 3, last three lines). All three translational fusions were inhibited by YenI about sixfold to eightfold. Taken together, these data indicate that YenS decreases the abundance of the yenI–lacZ message about twofold, and inhibits the translation of the residual mRNA approximately threefold to fourfold.

Mutations in yenR or yenS suppress hypermotility caused by a yenI mutation

A report from another lab showed that a mutation in yenI inhibited swimming and swarming motility of Y. enterocolitica strain 90/54 (which belongs to a different serotype that the strains used in the present study (Atkinson et al., 2006). Unexpectedly, that mutation (made using an antibiotic resistance cassette) was not suppressed by exogenous OHHL, suggesting that it may have affected expression of other genes, including the adjacent yenR gene. The medium used in that study included NaCl, which inhibits motility of the parent strain (JB580) used in the present study (Young et al., 1999).

We assayed for swimming and swarming motility of the wild-type strain and various mutants lacking YenR, YenI or YenS. After testing several media, we found that medium containing 1% tryptone and 10 mM glucose and 0.2% Bacto agar provided a sensitive assay for swimming motility, while a similar medium containing 0.3% Acros agar facilitated the detection of swarming motility. On the former medium all strains showed similar levels of motility (Fig. 8, top row). On the latter medium, neither the wild-type strain nor the yenR mutant was motile (Fig. 8, second row). In contrast, the yenI mutant was highly motile. Importantly, this motility was fully suppressed using medium supplemented with 1 µM OHHL, indicating that the defect was due entirely to the lack of OHHL production. Hypermotility was also blocked by an additional mutation in yenR or yenS. The fact that YenR and YenS were needed for this phenotype suggests strongly that YenS plays a direct, positive role.

Figure 8. Motility of wild-type and mutant strains of Y. enterocolitica. The strains used were JB580 (WT), CST206 (yenR^-), CST206 (yenI^-), HGB001 (yenS^-), CST201 (yenI^-, yenR^-) and HGB002 (yenI^-, yenS^-). Where indicated, OHHL was incorporated into the medium at a final concentration of 1 µM. Upper row, assays of swimming motility, using 0.2% Bacto agar; lower row, assays of swarming motility, using 0.3% Acros agar. Both media also contained 1% tryptone and 10 mM glucose.

Discussion

This study was initiated by a search for promoters that are regulated by YenR, and resulted in the unexpected finding that YenR activates the expression of a non-translated RNA that regulates the production of the cognate pheromone. Our work does not tell us whether YenR can directly regulate additional promoters. The ChIP experiments yielded other potential YenR binding fragments (Table S1), although some of these fragments appear to lack promoters and could be false positives. A preliminary search for other YenR binding sites yielded several possible candidates. It is also plausible that YenR may regulate only yenS.

It is far from clear why YenR binds to two sites in the yenR–yenS intergenic region rather than one, or why it can activate yenS from either site. The distance between yen box I and the yenS transcription start site is similar to that of class I activators, while the position of yen box II is similar to that of class II promoters. Class I activators are generally thought to interact with the C-terminal domain of the α-subunit of RNA polymerase, while Class II activators are thought to interact with the alpha-CTD, alpha-NTD, and/or sigma subunits. Most characterized LuxR homologues bind DNA non-cooperatively as dimers (Zhu and Winans, 1999; 2001; Urbanowski et al., 2004; Weingart et al., 2005), and usually bind approximately 42 nucleotides upstream of their transcription start sites. One LuxR protein (LasR) can bind some promoters cooperatively, forming a probable dimer of dimers (Schuster et al., 2004). No member of the family other than YenR is known to bind two adjacent sites non-cooperatively, or to function from either of two binding sites.

We can cautiously speculate about possible advantages of having two binding sites rather than one. The yenS–lacZ fusion that contained both yen boxes was induced 25-fold by apo-YenR and inhibited only about threefold by OHHL. In contrast, a similar fusion containing just yen box II was induced 14-fold by YenR, and induction was completely blocked by OHHL. Similarly, a fusion containing yen box I and an altered yen box II was induced 18-fold and induction was inhibited sixfold by OHHL. It is possible that the altered yen box II in that construct still retained residual ability to bind YenR. It seems that YenR binding to the two sites act cooperatively to recruit RNA polymerase, despite the fact that YenR binding to these sites is not cooperative. These data also suggest either that some apo-YenR persists even in the presence of OHHL or that YenR-OHHL complexes retain residual activity. Either way, this residual YenR activity in the presence of OHHL is most effective in the presence of two yen boxes (pCST2395), works slightly in the presence of one wild-type yen box and a second truncated or altered yen box (pCST2396 and pCST2410), and does not work at all with just one yen box (pCST2397).

As described above, YenR preferentially bound to target DNA in the absence of its cognate auto-inducers, OHHL and HHL. A few other LuxR homologues have similar properties, including EsaR, ExpR and SmaR (Fineran et al., 2005; Minogue et al., 2005; Castang et al., 2006), all of which are close relatives of YenR (Tsai and Winans, 2010). Virtually all other members of this family require AHLs for activity and for DNA binding (Whitehead et al., 2001). AHLs are also required for at least some members of this family to fold into mature, soluble forms (Zhu and Winans, 1999; 2001; Schuster et al., 2004; Urbanowski et al., 2004; Weingart et al., 2005). AHLs are completely buried within TraR, LasR and SdiA and contribute to the hydrophobic core of these proteins (Vannini et al., 2002; Zhang et al., 2002; Yao et al., 2006; Bottomley et al., 2007). The fact that YenR and its relatives function only as apoproteins indicates that they must not need AHLs for folding. Moreover, these proteins must have a binding site for AHLs that is accessible in the fully folded protein. It will therefore be extremely interesting to use biochemical and structural approaches to learn how OHHL perturbs the properties of YenR or its close relatives.

Many LuxR–LuxI systems exhibit hysteresis, meaning that their responses to quorum sensing stimuli are conditioned by their recent history (von Bodman et al., 2008). Simply put, these systems, when in the inactive state tend to stay inactive despite environmental perturbations, and vice versa. For typical LuxR–LuxI systems, hysteresis results from positive regulation of the AHL synthase gene by the AHL receptor protein. The YenR–YenI system could also exhibit hysteresis, but by a completely different mechanism. There are at least two factors that may contribute to hysteresis, the first of which involves YenS. At low cell densities, AHLs accumulates very poorly, so apo-YenR accumulates and activates the yenS promoter (Fig. 9). YenS inhibits production of AHLs, and this inhibition ensures continued accumulation of active apo-YenR. At high cell densities, accumulated AHLs inactivate YenR, so the production of YenS falls. This stimulates AHL production, which ensures that YenR remains inactive. In essence, apo-YenR makes an inhibitor of itself, while YenR–AHL complexes cannot do so (Fig. 9).

Figure 9. A model describing the regulatory circuitry of YenI, YenR and YenS. According to this model, at low population densities, OHL is expected to accumulate very poorly, causing apo-YenR to accumulate and activate transcription of yenS. The YenS RNA is proposed to bind to YenI mRNA, blocking its transcription and translation. The inhibition of YenI production will inhibit the production of OHHL. This ensures that YenR will remain in the unliganded, active form. At high population densities, sufficient OHHL accumulates to form complexes with YenR, thereby inactivating it. This leads to decreased expression of YenS, and that decrease causes an increase in YenI production, and an increase in OHHL accumulation, which in turn helps to ensure that YenR will remain inactive. Additionally, the YenI and YenR mRNAs may antagonize each other's transcription or accumulation, either via RNA polymerase collision or by mRNA-mRNA duplexes. In either case, the more abundant mRNA could ensure the destruction of the less abundant mRNA. These regulatory features could ensure hysteresis, that is, the tendency for the system to remain stable in one state or the other state even in the face of sub-threshold changes in environmental conditions.

The second way that this system may exhibit hysteresis is more speculative, but highly plausible. The yenR and yenI genes are convergent and overlap by one codon, indicating that their mRNAs must overlap by at least 10–20 nucleotides and possibly much more. All the members of the EsaR/I subfamily are also encoded by convergent, overlapping genes (Tsai and Winans, 2010). The YenR and YenI mRNAs may therefore hybridize, which could destabilize either or both (Fig. 9). If so, then at low cell densities, apo-YenR would, through YenS, decrease levels of YenI mRNA. This potential excess of YenR mRNA could stoichiometrically degrade both species of mRNA, until only YenR mRNA remained. The lack of YenI mRNA would ensure continued accumulation of apo-YenR. At high cell densities, YenR–AHL complexes would fail to activate the yenS promoter, causing an increase in YenI mRNA. The potential excess of YenI mRNA could cause stoichiometric degradation of YenR mRNA, further decreasing the production of YenR.

We strongly suspect that additional genes are controlled by this system. Any such genes could be regulated directly by YenR, or by YenS, and could themselves regulate still other genes. YenR could act either as a repressor or as an activator, depending largely on the position of the binding site with respect to the target promoter (Ren et al., 2000). We have carried out a preliminary bioinformatic search for YenR-regulated promoters (data not shown), and will study several such promoters in future studies. YenS could also regulate other genes, probably by inhibiting the accumulation or translation of an mRNA, although it could have the opposite effect, for example, by inactivating an inhibitory site of an mRNA. In principle, YenS targets should be identifiable by their complementarity. TargetRNA has identified several other candidate genes (Fig. S7). Transcriptional profiling as well as proteome analysis would provide other approaches for finding members of this regulon.

In an effort to identify target genes regulated by YenR or YenS, we undertook assays of swimming and swarming motility on nutrient agar. The yenI, yenR or yenS status of the strains had little effect on swimming motility. In contrast, yenI mutants showed strong swarming motility, while all the other strains were non-motile on this medium. This is very unlike the phenotypes observed by another group (Atkinson et al., 2006), probably due to the differences in strains, mutations or experimental design. Our strain JB580 was derived from strain 8081v, an O:8 serotype, while the previous study was done using strain 90/54, an O:9 serotype. Both studies used 0.3% agar with 1% tryptone, but we also added 10 mM glucose, while the medium in previous study contained no glucose but contained 0.5% NaCl. NaCl inhibits swarming of JB580 (Young et al., 1999). Another important difference may be in the nature of the two yenI mutations. Our mutation was fully suppressed by exogenous OHHL, while the yenI mutation of the earlier study was not, suggesting that it could have impeded the expression of yenR. If so, such a mutant would be non-motile (Fig. 8).

Our model for quorum sensing shows a regulatory cycle, in which OHHL impacts activity of YenR, which impacts production of YenS, which impacts the translation of YenI, which impacts the abundance of OHHL. A mutation in yenI breaks this cycle, and allows us to ask which of these components is most likely to act upon motility directly. In principle, the lack of OHHL caused by a yenI mutation could alter the activity of an AHL receptor other than YenR. Y. enterocolitica encodes a second LuxR-type protein, YE1026 (Thomson et al., 2006), and phylogenetic analysis suggests that it could well be a receptor for OHHL. However, a yenI, yenR double mutant was defective in motility, arguing against a role for a second AHL receptor. YenR is known to directly regulate just one gene, yenS, but could have additional direct targets, including one or more that are involved in motility. However, a yenI, yenS mutant was defective in motility. Although it remains conceivable that YenR could directly regulate motility genes, these data argue that YenS has a direct role in regulating motility, possibly via posttranscriptional effects on protein synthesis.

As this manuscript was being prepared, a report appeared that Pantoea stewartii expresses a non-translated gene divergent from its esaR gene whose expression is stimulated by EsaR (Schu et al., 2009). Expression of this gene, esaS, in a heterologous host (E. coli) was activated by apo-EsaR and inhibited by OHHL. There is no apparent sequence conservation between yenS and esaS, and there is no obvious complementarity between EsaS and EsaI. However, it is tempting to speculate that these genes may have some functional similarity, and it will be interesting to learn more about their similarities and differences.

Experimental procedures

Strains, plasmids and reagents

Bacterial strains, chromosomal mutations, plasmids and oligonucleotides used in this study are described in Tables S2–S5. Oligonucleotides were purchased from Integrated DNA Technologies (Coralville, IA, USA). Antibiotics and Protein A-Sepharose beads were purchased from Sigma. Restriction enzymes, DNase I, T4 DNA kinase and terminal DNA transferase were purchased from New England Biolabs. PCR fragment purification kits and RNeasy kits were purchased from Qiagen. Radionucleotides were purchased from Perkin Elmer. Trizol reagent and Superscript III reverse transcriptase were purchased from Invitrogen. Polyclonal rabbit Anti-YenR antiserum was prepared by the Cornell Center for Research Animal Resources. OHHL and HHL were generously provided by A. Eberhard (Cornell University), and were chemically synthesized using published procedures (Eberhard et al., 1981).

Chromosomal genes were disrupted, singly and in combination, by a two-step Campbell-type integration and excision of a suicide plasmid carrying deletion alleles of the target genes. DNA fragments (approximately 500 nucleotides in length) upstream and downstream of the target genes were constructed by PCR amplification, ligated, and then cloned into plasmid pKNG101, a suicide plasmid carrying an R6K origin and a sacB counter-selectable gene. The resulting plasmids were introduced into E. coli SM10 λ-pir by transformation, and then introduced into Y. enterocolitica JB580 or its derivatives by conjugation. Transconjugants containing the suicide plasmid integrated into the chromosome were selected using streptomycin and nalidixic acid. Excision of the suicide plasmid was subsequently selected by plating cells on medium containing 10% sucrose and nalidixic acid. Sucrose-resistant clones were characterized using PCR amplification to confirm the deletion.

Integration of lacZ fusions into the ara locus

Suicide plasmids with various lacZ fusions were transformed into S17-1/λ-pir and selected on Luria–Bertani (LB) agar plates with chloramphenicol. The resulting strains were used as conjugative donors in overnight matings on LB agar with recipient strains CST214 and CST216. Transconjugants were selected on LB plates containing both chloramphenicol and nalidixic acid. Single colonies were streaked onto LB agar with nalidixic acid and 10% sucrose. Candidate colonies were confirmed using PCR reactions and by screening for chloramphenicol sensitivity.

Purification of YenR

YenR was overexpressed by constructing plasmid pRA504, which contains a PT7-yenR fusion. An overnight culture of BL21(DE3)(pRA504) was diluted 100-fold into LB broth with appropriate antibiotics. At an OD₆₀₀ of 0.5, IPTG was added (1 mM final concentration) and the cells were cultured for four additional hours. The cells were then collected and disrupted using a French press. The lysate was clarified by high speed centrifugation and applied to a heparin sepharose matrix that had been equilibrated with SEDG buffer (50 mM sodium phosphate, 0.1 mM EDTA, 5% glycerol and 1 mM DTT). YenR was eluted with the same buffer containing a gradient of NaCl. Fractions containing YenR were pooled and dialysed against the SEDG buffer with 100 mM NaCl and 50% glycerol and stored at −80°C.

Chromatin immunoprecipitation (ChIP)

Chromatin immunoprecipitation assays were performed using a published procedure (Molle et al., 2003). Strain YWY003 was cultured in LB medium to an OD₆₀₀ of 0.3, at which time formaldehyde was added (1% v/v) to cross-link DNA binding proteins to DNA. The cross-linking reaction was terminated by the addition of excess glycine. Cells were disrupted and DNA was sheared by sonication. YenR–DNA complexes were precipitated using anti-YenR antiserum and protein A-Sepharose beads.

Precipitates were heat-treated to reverse the cross-links, and DNA fragments were ligated to uni-directional linkers (oJW102 and oJW103) containing an EcoRI restriction site (Ren et al., 2000). The ligated fragments were then amplified by PCR using oligonucleotides that bind to the linkers, digested with EcoRI, and cloned into plasmid pBluescript II SK(+). Plasmids containing inserts were submitted for automated DNA sequencing.

Electrophoretic mobility shift assays

Electrophoretic mobility shift assays were conducted using double-stranded synthetic oligonucleotides that were treated with (-P32-ATP and T4 polynucleotide kinase. Purified YenR and 0.12 pmol of labelled DNA were combined at room temperature in a buffer containing 10 mM Tris·HCl (pH 7.9), 1 mM EDTA, 1 mM DTT, 60 mM potassium glutamate, 30 µg ml⁻¹ calf thymus DNA, 20 µg ml⁻¹ BSA and 10% glycerol, and incubated at room temperature for 30 min. Complexes were size-fractionated using 5% or 15% polyacrylamide gels containing TAE buffer (40 mM Tris·acetate/2 mM EDTA, pH 8.5). Gels were examined using a Storm B840 PhosphorImager (Molecular Dynamics). ImageJ software (Rasband, 2008) was used to calculate the ratio of bound to free DNA. In some experiments, auto-inducers in the indicated concentrations were added 30 min after YenR–DNA binding reactions were initiated, then the reactions were incubated for 30 min more and characterized by EMSA.

β-Galactosidase assay

Cells were cultured in LB medium overnight, then diluted 100-fold into AB defined medium (Cangelosi et al., 1991). When the cultures had reached an OD₆₀₀ of approximately 0.4, they were assayed for β-galactosidase-specific activity as described (Miller, 1983).

DNase I footprinting assays

DNA primer o262, which contains a 5′-6XFAM fluorescent dye, and o263 were used to PCR amplify a DNA fragment containing yen boxes I and II. Binding of YenR was carried out as described above. Protein–DNA complexes were digested using DNase I (NEB, diluted to 0.2 U µl⁻¹) for 40 s. The reactions were then stopped by the addition of SDS and EDTA. DNA was further purified using Qiaquick purification kits (Qiagen), and submitted to the Cornell Life Sciences Core Laboratories Center for analysis using an Applied BioSystems 3730xl DNA Analyzer.

RNA extraction, Northern blotting, primer extension, S1 nuclease assay, quantitative PCR and RACE assays

Cells were cultured in AB or LB medium and collected at an OD₆₀₀ of 0.3. A solution containing ethanol and phenol was added to kill the bacteria and preserve RNAs. Cells were centrifuged and frozen at −80°C. RNA was extracted using either Trizol reagent (Invitrogen) or a Qiagen RNeasy kit. The total RNA was quantified by absorbance at 260 nm and RNA integrity was checked by agarose gel electrophoresis. RNA was size-fractionated by electrophoresis in 8% polyacrylamide gels containing 8 M urea and 1× TBE, and then blotted electrophoretically onto an Amershan Hybond-XL membrane (GE Healthcare) in 1× TBE buffer. The membranes were probed with 5′-radiolabelled oligonucleotides o451, o673, o698 and o699.

Primer extension reactions were performed using an oligonucleotide containing a 6XFAM 5′ end (o268) and Superscript III reverse transcriptase. The reactions were stopped by heating and RNA was subjected to alkaline hydrolysis. Fluorescently labelled cDNA was purified and analysed using an Applied BioSystems 3730xl DNA Analyzer.

S1 nuclease assays were carried out as described previously (Cho and Winans, 2005), primers used are listed in Table S5. Quantitative reverse transcriptase PCR analysis was carried out using cDNA obtained with iScript cDNA synthesis kit (Bio-Rad). Oligonucleotide primers and iTaq SYBR Green Supermix with ROX (Bio-Rad) were added to cDNA samples and the mixture was amplified and analysed by an ABI 7300 Real-time PCR system.

5′ RACE reactions were carried out by synthesizing cDNA using a primer specific for each gene. Terminal DNA transferase and dCTP were added to produce homopolymers at the 3′ ends of each cDNA. A poly-dG primer and a second gene specific primer were used to PCR amplify these cDNA fragments. The resulting fragments were cloned into pBluescript SK(+) and submitted for automated DNA sequencing.

3′ RACE reactions were carried out using RNA treated with CIP (Calf Intestinal alkaline phosphatase, NEB). CIP was then removed by phenol/chloroform extraction. RNA adapter E1 (5′Phos-UUCACUGUUCUUAGCGGCCGCAUGCUC-3′ InvdT) was denatured and chilled on ice. The CIP treated RNA, E1 adapter and T4 RNA ligase were then combined overnight at 16°C. RNA ligase was then removed by phenol extraction. Primer extension with a DNA oligonucleotide complementary to the E1 adapter was performed using Superscript III reverse transcriptase (Invitrogen). The resulting cDNA was amplified using primer o024 (which is complementary to the cDNA) and an oligonucleotide complementary to E1 adapter. The PCR product was cloned into pBluescript SK(+) and sequenced.

Analysis of yenS gene expression by flow cytometry

Strains CST201(pCST2406), CST206(pCST2406) and JB580(pCST2406) were cultured in LB medium overnight, then washed and diluted 10⁸-fold into AB medium containing 1% LB. Cells were collected at intervals and analysed for GFP-mediated fluorescence using a Coulter Epics XL-MCL flow cytometer. Data were collected for 5000 cells in each sample, and analysed using WinMDI software.

Bioassays for extracellular AHLs

Strains producing AHLs were diluted 5000-fold into fresh defined AB growth medium supplemented with 2.5% LB. At intervals, 1 ml samples were removed and stored at −80°C. At the end of the growth interval, all samples were thawed, centrifuged to remove bacterial cells, and 10 µl of each supernatant was added to early-log phase cultures of the A. tumefaciens AHL bioassay strain WCF47(pCF218)(pCF372) (Zhu et al., 1998). Cultures of this strain were assayed for β-galactosidase-specific activity after overnight culturing.

Motility on semi-solid medium

Swimming and swarming motility of wild-type and mutant strains was tested using growth medium containing 1% tryptone, 10 mM glucose and 0.2% Bacto agar (for swimming) and the same medium containing 0.3% Acros agar (for swarming). For swarming test, a volume of 2 µl of log phase cultures was spotted in the centre of each plate without penetrating the agar surface. For swimming assays, 2 µl of log phase cells was stabbed into the medium, while for swarming assays, 2 µl of cells was spotted onto the agar surface. The plates were incubated at 22°C for 16–20 h and photographed.

Acknowledgements

We thank Dr Heidi Goodrich-Blair, Dr Reiko Akakura and Yuping Wei for help with strain construction. We are grateful to Dr Dorothy Debbie (Cornell) and Dr Andrew Darwin (New York University) for strains and advice, and to the members of our laboratory for critical review of this manuscript. We thank Carol Bayles and the Cornell Microscopy and Imaging Facility for the help of flow cytometry experiments, and Peter Schweitzer for help with DNase I protection experiments. We also thank Dr Anatol Eberhard for providing auto-inducers. This study was supported by a grant from the National Institute of General Medical Sciences (GM041892).