Strains and plasmids used in this study are listed in Table 1. The Escherichia coli strains were aerobically grown in LB medium (10 g/l NaCl, 10 g/l tryptone, 5 g/l yeast extract) at 37°C in a rotary shaker. The V. harveyi strains were cultivated in autoinducer bioassay (AB) medium (Greenberg et al., 1979) or Luria marine (LM) medium (20 g/l NaCl, 10 g/l tryptone, 5 g/l yeast extract) and were grown aerobically in a rotary shaker at 30°C. When required, media were solidified by using 1.5% (w/v) agar. If necessary, media were supplemented with 50 μg/ml kanamycin sulfate and/or 100 μg/ml ampicillin sodium salt. The conjugation strain E. coli WM3064 was grown in the presence of 300 μM meso-diaminopimelic acid (DAP). For microscopy and in-gel fluorescence experiments V. harveyi cells were grown overnight in LM medium and afterwards inoculated 1:5,000 in fresh AB medium. For the in-gel fluorescence approach cells were harvested in the early or late exponential phase (OD₆₀₀ ~0.08 and ~0.7, respectively).

Molecular methods were carried out according to standard protocols (Sambrook, 1989) or according to manufacturer's instructions. Kits for the isolation of plasmids and purification of PCR products were purchased from Südlabor (Gauting, Germany). Enzymes were purchased from New England Biolabs (Frankfurt, Germany) and Fermentas (St. Leon-Rot, Germany). Chemically competent cells of E. coli were transformed with the corresponding plasmids (Inoue et al., 1990). mNeonGreen (Shaner et al., 2013) (Allele Biotechnology, San Diego) was used to replace GFP in the vector pNPTS138-R6KT-GFP after restriction with SpeI and PspOMI. For construction of genes coding for the full-length receptors tagged with mNeonGreen (P_luxN-luxN-mNeonGreen, P_luxQ-luxQ-mNeonGreen, and P_cqsS-cqsS-mNeonGreen) the coding sequence as well as promoter sequence of luxN, luxQ, and cqsS were ligated into pNPTS138-R6KT-mNeonGreen after restriction with EcoRI/BamHI (in case of luxN) or BamHI/PspOMI (for luxQ and cqsS) (Table S1). Chromosomal insertions of these gene fusions into V. harveyi were achieved by integrating the resultant suicide vectors via RecA dependent single homologous recombination as described previously (Fried et al., 2012). The conjugative plasmid transfer from donor strain E. coli WM3064 containing the required plasmid into V. harveyi was performed as described before. For this purpose, the donor and the recipient strain were cultivated in LB medium up to an OD₆₀₀ of 0.8–1.0, supplemented with 300 μM meso-diaminopimelic acid (DAP) for the growth of E. coli WM3064. Single colonies were checked for chromosomal integration via performance of a PCR with the genomic DNA.

Vibrio harveyi ATCC BAA-1116 was cultivated as described above, and samples were harvested by centrifugation every hour. RNA was isolated as described before (Fritz et al., 2009). The RNA was used as template for random-primed first-strand cDNA synthesis according to the manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) (iQ5 real-time PCR detection system, Biorad) was performed using the synthesized cDNA, a SYBR-green detection system (Biorad) and specific internal primers for recA, luxN, luxQ, cqsS, luxO, and luxU (Table S1). The CT value (cycle threshold) was determined after 40 cycles using the iQ software (Biorad). Values were normalized with reference to recA and relative changes in transcript levels were calculated using the comparative CT method (Schmittgen and Livak, 2008).

Escherichia coli Rosetta (DE3) pLysS was transformed with plasmid pNKN, pNKQ, or pKK223-3-cqsS-F175C encoding wild type LuxN, LuxQ, or CqsS-F175C respectively. For overproduction the cells were induced using 1 mM IPTG. Membrane vesicles were prepared as described before and extensively washed to remove peripheral ATP (Timmen et al., 2006).

Escherichia coli MDAI-2 was transformed with the plasmid pGEX_LuxP, and LuxP was purified as described elsewhere (Neiditch et al., 2005). LuxU was overproduced using E. coli JM109 transformed with pQE30LuxU-6His, and purified as described in Timmen et al. (2006). All proteins were stored at −80°C.

Phosphorylation reactions were performed in phosphorylation buffer (50 mM Tris/HCl pH 8.0, 10% (v/v) glycerol, 500 mM KCl, 2 mM DTT) at room temperature. The hybrid histidine kinases LuxQ, CqsS-F175C, and LuxN were used as full-length membrane integrated proteins in membrane vesicles. LuxQ containing membrane vesicles were added at final concentrations of 1 mg/ml. Via comparative Western Blot the amounts of LuxN and CqsS-F175C membrane vesicles needed were calculated as ratio based on LuxQ. The reaction mixture contained 0.36 mg/ml LuxU and 0.1 mg/ml LuxP unless otherwise indicated. To incorporate LuxP into LuxQ containing membrane vesicles, three cycles of freezing and thawing were performed. Unless otherwise indicated, AI-2 (synthesized by Rita Ventura, Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa), HAI-1 (purchased from the University of Nottingham) and V. cholerae CAI-1 [(S) 3-hydroxytridecan-4-one) (kindly provided by Dr. Joachim Schultz, University of Tübingen) were added in a final concentration of 10 μM. The phosphorylation reaction was started by adding radiolabeled Mg²⁺-ATP, typically 100 μM [γ-³²P] ATP (0.94 Ci/mmol; Perkin-Elmer, Rodgau-Jügesheim, Germany) and 110 μM MgCl₂, and stopped at various time points by the addition of SDS loading buffer (Jung et al., 1997), followed by separation of the proteins on a SDS-PAGE (Laemmli, 1970). Gels were dried at 80°C on filter paper, exposed to a phosphoscreen for at least 24 h and scanned using a Typhoon Trio variable mode imager (GE Healthcare, München, Germany). Quantification of the bands was performed using ImageQuant (GE Healthcare, München, Germany).

For dephosphorylation assays, LuxU was phosphorylated using Lux(P)Q. In this case the phosphorylation buffer contained 10 mM CaCl₂ instead of MgCl₂ and twice the amount of Lux(P)Q and LuxU. After 10 min incubation at room temperature, membrane vesicles were removed by centrifugation (100,000 × g, 15 min, 4°C), and ATP and CaCl₂ were removed by gel filtration (Sephadex G25 columns, GE Healthcare). Dephosphorylation of P-LuxU was started by the addition of 110 μM MgCl₂, 100 μM ATP-γ-S (adenosine-5′-O-thiophosphate), and membrane vesicles containing the respective receptor. As described above, the reaction was stopped at the indicated time points, samples were subjected to SDS-PAGE and exposed to a phosphoscreen.

Cells were cultivated as described above. In the early and late exponential growth phase (OD₆₀₀ ~0.08 and ~0.7, respectively), cells were harvested by centrifugation (7,000 × g, 15 min, 4°C). The pellet was resuspended in TG buffer (50 mM Tris/HCl pH 8.0, 10% (v/v) glycerol) supplemented with lysozyme, DNase and PMSF (0.5 mM) and thereby concentrated 200 fold. After incubation at 37°C for 15 min, membranes were collected by ultracentrifugation (100,000 × g, 15 min, 4°C). The pellet was resuspended in TG buffer and diluted when indicated. Samples were mixed with buffer [200 mM Tris/HCl (pH 8.8), 20% (v/v) glycerol, 5 mM EDTA (pH 8.0), 0.02% (w/v) bromphenol blue (aliquots of 700 μl)], and before use 200 μl 20% (w/v) SDS and 100 μl 0.5 M DTT (Drew et al., 2006) was added. SDS-PAGE was performed according to Laemmli (Laemmli, 1970) and run at 150 V for 2 h in the dark to avoid bleaching of the fluorophore. In gel-fluorescence was analyzed using the Typhoon Trio scanner (Amersham Biosciences) with 488 nm laser and 526 nm emission filter. Quantification of fluorescence was performed using ImageQuant (GE Healthcare, München, Germany).

Cells were grown in AB medium until the late exponential growth phase and fluorescence microscopy was conducted. Three microliter of cells were spotted on 1% (w/v) agarose pad containing AB medium and imaged at 30°C. A DeltaVision Elite microscope (GE Healthcare, Applied Precision) equipped with a CoolSnap HQ2 CCD camera was used. Images were taken with 100x oil PSF U-Plan S-Apo 1.4 NA with an exposure time of 2 s using 475/28 nm excitation and 525/48 nm emission wavelengths. Analysis was performed using ImageJ.

In vitro phosphorylation assays were performed to determine the enzymatic activities of each QS receptor. For this purpose, genes of the full-length receptors were overexpressed; membrane vesicles were prepared and used for phosphorylation experiments. Previous characterizations of the kinase and phosphatase activities for LuxN and Lux(P)Q suggested that their kinase activities were in the same range and that the presence of AIs only influences the kinase but not the phosphatase activity (Timmen et al., 2006; Anetzberger et al., 2012). However, the AIs concentration dependent inhibition on the kinase activities was incomplete, indicating that even at high HAI-1 and AI-2 concentrations the phosphorylated LuxU was still detectable (Timmen et al., 2006; Anetzberger et al., 2012). Nonetheless, to quantitatively compare enzymatic activities we believe it is crucial to perform the assays for all QS receptors at the same time and under exactly same conditions. Since V. harveyi CAI-1 can only be enriched in dichloromethane, which interferes with our phosphorylation assay and CAI-1 is not commercially available, we introduced an amino acid substitution at the position 175 (Phe to Cys). In vivo studies have shown that the V. harveyi mutant CqsS-F175C exhibits relaxed specificity and detects both Ea-C8-CAI-1, the natural V. harveyi CAI-1, and the CAI-1 from V. cholerae (Ng et al., 2011). To ensure that equal protein amounts are used for the phosphorylation assays, we quantified the amount of the receptors in the membrane vesicles via Western blot.

We have previously shown that autophosphorylation activity of LuxN is low and unstable, but the phosphorylation of the subsequent LuxU is stable (LuxU seems to act as a phosphate sink; Timmen et al., 2006). Therefore, we used the degree of phosphorylation of LuxU as a measure for the kinase activities of the QS receptors. LuxU cannot be phosphorylated by ATP (Timmen et al., 2006). The phosphorylation tube contained the corresponding QS-receptor and LuxU, and at time zero radiolabeled ATP was added to start the reaction. All three QS receptors autophosphorylated and phosphorylated LuxU in a time dependent manner (Figures 2A,B). The degree of increase of P-LuxU depends on the receptor present in the reaction (Figure 2B). After 10 min, the highest amount of P-LuxU was obtained in the presence of LuxN, indicating that LuxN presents the strongest kinase activity. Membrane vesicles containing CqsS-F175C showed a comparable initial rate of LuxU phosphorylation, however the saturation level of P-LuxU was already reached after 1 min (Figure 2B). Altogether the capacity of the receptors to phosphorylate LuxU showed the following order LuxN>Lux(P)Q>CqsS-F175C. Furthermore, the addition of each AI inhibited the respective kinase activity. It is interesting to note that despite the fact that LuxN shows the strongest kinase activity, all kinase activities were inhibited to the same degree upon AI addition (Figure 2C). However, the inhibitory effects of all three AIs were incomplete and residual kinase was still detected, which is in agreement with previous studies (Anetzberger et al., 2012).

In order to investigate the corresponding phosphatase activities, LuxU was phosphorylated with [γ³²-P] ATP via Lux(P)Q, purified and employed for subsequent dephosphorylation analysis. All three receptors were able to dephosphorylate P-LuxU in a time-dependent manner (Figure 2D). P-LuxU is stable for at least 40 min (Timmen et al., 2006), excluding the possibility that the decrease of P-LuxU is due to the instability of the protein and/or the phosphorylation (Figure S1). The dephosphorylation of P-LuxU was calculated as number of fold-reduction based on P-LuxU incubated without the receptors. All three receptors dephosphorylated the P-LuxU to a comparable level in a similar timeframe (Figure 2D). However, none of the hybrid histidine kinases was able to dephosphorylate P-LuxU completely.

To investigate the potential differences in the quantity of the QS components, we performed qRT-PCR and compared transcript levels of the QS receptors. Transcript levels were measured from the V. harveyi wild type by taking samples of the cell culture every hour (Figure 3C). Overall, luxN was the most transcribed kinase with its highest transcript levels in the late exponential growth phase (Figure 3A). The transcription of luxQ and cqsS did not differ significantly and, except for a slight peak for luxQ after 3 h, remained constant (Figure 3A). Moreover, we could show that the transcript levels of the kinases were overall lower compared to luxU coding for the HPt protein, or luxO coding for the response regulator at any time (Figures 3A,B). The transcript levels of luxU were 2-fold higher than the mRNA levels of luxO (Figure 3B).

In many bacteria, chemotaxis receptors form clusters at the cell poles to detect chemical signals from the environment with high sensitivity over a wide range of concentrations (Briegel et al., 2009). In order to elucidate the potential cluster formation of the QS receptors we performed electron cryotomography (ECT) with the wild type V. harveyi and a triple kinase mutant (V. harveyi ΔluxNΔluxQΔcqsS). This ECT method was successfully used to identify cluster formation for many bacterial chemoreceptors (Briegel et al., 2009). However, we could not find any differences between the wild type and the triple mutant (Ariane Briegel and Grant Jensen, data not shown). This could be either due to the low receptor expression levels or to the lack of cluster formation.

Subsequently, we constructed reporter strains encoding full-length QS receptors tagged with a C-terminal fluorophore to investigate their subcellular localization as well as the corresponding protein levels. All genes encoding these fluorophore hybrids were integrated at the native locus in the chromosome. Our first attempts to visualize the receptors using GFP or mCherry fusion protein were unsuccessful presumably due to the low expression levels. Contrarily, the receptors fused to the mNeonGreen (NG) protein showed fluorescence signal at the membrane, most likely due to the higher brightness compared to e.g., GFP (Shaner et al., 2013). Cells were grown until late exponential growth phase (see Figure 3C t₂) and subsequently employed for fluorescence imaging on agar pads.

The non-tagged V. harveyi wild type was used as control for background fluorescence to ensure that the observed signals are not artifacts (Figure S2). The signal in the control was significantly lower compared to the reporter strains harboring the receptor fluorophore hybrids. LuxN-NG clustered in the membrane and exhibited the strongest fluorescence signal (Figure 4 upper panel). LuxQ-NG also formed clusters, however, LuxQ-NG localized preferentially at the cell poles (Figure 4 second panel from top). Contrary to LuxN and LuxQ proteins, the fluorescence of CqsS-NG was rather homogeneously distributed in the membrane (Figure 4 third panel from top). The negative control with the non-tagged wild type using the same imaging settings confirmed that the clusters seen in the reporter strains were not artifacts (Figure 4 bottom panel).

To determine the abundance of each of the receptors, we performed in-gel fluorescence assays. For this assay, we used the same reporter strains as for the microscopy approach. Cells were harvested in the early and late exponential growth phase, respectively (for time points see Figure 3C t₁ and t₂), lysed, the membrane fractions were prepared and subsequently analyzed by SDS-PAGE. For each reporter strain (LuxN-NG, LuxQ-NG and CqsS-NG) a specific band (corresponding to the size of the hybrid receptor) could be observed, which was absent in the non-tagged wild type control (Figure 5A). In the early exponential growth phase LuxN was the most abundant kinase, followed by LuxQ and CqsS (Figure 5A). Quantification of the band intensities using ImageQuant and normalized to the LuxQ band showed that the amount of LuxN was three times higher than LuxQ. Whereas CqsS showed the lowest copy number and was two-fold lower compared to LuxQ (Figure 5B). In the late exponential growth phase, specific bands for each of the fusions could be observed as well. However, the distinct expression pattern differed from the one in the early exponential phase. LuxN was still the predominantly expressed kinase, but its ratio relative to the other kinases was further increased: LuxN was more abundant than LuxQ by a factor of 8 ± 3. LuxQ and CqsS showed comparable protein amounts with CqsS being slightly more abundant than LuxQ (factor 1.24 vs. 1; Figure 5C). It is interesting to note, that the same differences on the abundance of the receptors were observed also at the transcript levels (compare Figures 3, 5).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.