AI-1 Influences the Kinase Activity but Not the Phosphatase Activity of LuxN of Vibrio harveyi*

The Gram-negative bacterium Vibrio harveyi produces and responds to three autoinducers, AI-1, AI-2, and CAI-1 to regulate cell density dependent gene expression by a process referred to as quorum sensing. The concentration of the autoinducers is sensed by three cognate hybrid sensor kinases, and information is channeled via the HPt protein LuxU to the response regulator LuxO. Here, a detailed biochemical study on the enzymatic activities of the membrane-integrated hybrid sensor kinase LuxN, the sensor for N-(d-3-hydroxybutanoyl)homoserine lactone (AI-1), is provided. LuxN was heterologously overproduced as the full-length protein in Escherichia coli. LuxN activities were characterized in vitro and are an autophosphorylation activity with an unusually high ATP turnover rate, stable LuxU phosphorylation, and a slow phosphatase activity with LuxU∼P as substrate. The presence of AI-1 affected the kinase but not the phosphatase activity of LuxN. The influence of AI-1 on the LuxN→ LuxU signaling step was monitored, and in the presence of AI-1, the kinase activity of LuxN, and hence the amount of LuxU∼P produced, were significantly reduced. Half-maximal inhibition of kinase activity by AI-1 occurred at 20 μm. Together, these results indicate that AI-1 directly interacts with LuxN to down-regulate its autokinase activity and suggest that the key regulatory step of the AI-1 quorum sensing system of Vibrio harveyi is AI-1-mediated repression of the LuxN kinase activity.

Vibrio harveyi is a free-living marine bacterium that controls bioluminescence, exopolysaccharide and siderophore production, type III secretion, and other processes through quorum sensing. V. harveyi has three quorum sensing systems. AI-1, an acyl homoserine lactone, sensed by LuxN is employed for intraspecies signaling (1). AI-2, a furanosyl borate diester, sensed by LuxPQ is involved in inter-species signaling (2). CAI-1, whose chemical nature is unknown, is sensed by CqsS and is proposed to be responsible for Vibrio-specific signaling (3). Information from all three hybrid sensor kinases is transduced via phos-phorelay to the response regulator LuxO with the HPt protein LuxU acting as intermediate (Fig. 1). Phosphorylated LuxO, a 54 -dependent response regulator, activates transcription of genes encoding four small regulatory RNAs, which together with Hfq, destabilize the transcript for the LuxR protein, the master transcriptional regulator required for quorum sensing gene expression (4 -6). Bioluminescence is used as the canonical readout for V. harveyi quorum sensing controlled gene expression. Genetic studies indicated that LuxO is phosphorylated at low cell density, causing the luciferase operon to be repressed, and the cells are dark. When a threshold concentration of autoinducer is reached, LuxO is dephosphorylated, leading to induction of the luciferase operon and a bright phenotype (Fig. 1). It has been postulated that the hybrid sensor kinases switch from kinases to phosphatases at high autoinducer concentration and hence high cell density (3)(4)(5)7).
Despite numerous genetic studies of quorum sensing in V. harveyi, biochemical characterization of the corresponding proteins is rare. The solution structure of LuxU was solved by NMR. LuxU contains a four-helix bundle with the active-site histidine residue (His 58 ) exposed to the solution (8). Co-crystallization of LuxP and the periplasmic domain of LuxQ showed that LuxP is bound to LuxQ independent of the presence/absence of AI-2 (9). Amino acid replacements of residues located in the LuxP::LuxQ interface sensitize V. harveyi to AI-2, implying that binding of AI-2 to LuxP causes a conformational change within LuxQ that is transmitted across the membrane, presumably affecting LuxQ enzymatic activities (9).
Here, we characterized LuxNЈs enzymatic activities and its phosphotransfer properties to LuxU. LuxN possesses an autokinase activity that has an exceptionally high ATP turnover rate. Moreover, AI-1 has an inhibitory effect on LuxN kinase activity, whereas the LuxN phosphatase activity is unaffected by AI-1.

MATERIALS AND METHODS
Bacterial Strains and Growth Conditions-E. coli strains TKR2000 (10) and JM109 (11) were used throughout this study. Cells were grown in KML medium (1% tryptone, 1% KCl, 0.5% yeast extract) and incubated aerobically in Erlenmeyer flasks in a rotary shaker at 37°C. When appropriate, the medium was supplemented with ampicillin (100 g⅐ml Ϫ1 ). Overexpression of luxN or luxU was induced by addition of 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h.
Cloning of luxN and luxU-The luxN gene was amplified by PCR from purified cosmid DNA using the primers LuxN-BamHI sense (5Ј-AAAGGGGGATCCATGTTTGATTT-TAGCCTAGAG-3Ј) and LuxN-PstI antisense (5Ј-AGGTG-CCTGCAGTTCTCTCTCAGCTTCACAAGC-3Ј). The PCR fragment obtained was digested with BamHI/PstI, gel-purified, and ligated with appropriately digested pTCaiTH (12) to obtain pTLuxN-6His in which luxN is under control of the lac promoter/operator. For cloning of luxN into pKK223-3 (GE Healthcare), a HindIII site in luxN had to be removed. This was accomplished by introducing a silent mutation with concomitant formation of a DraI site using oligonucleotides LuxN-HindIII sense (5Ј-GAGGATTTTAAACTTTCTATT-3Ј) and LuxN-600 sense (5Ј-TATTTCTCATGCCAGTTGGC-3Ј) and applying the PCR overlap extension method (13). The product was digested with XhoI and NsiI, and the resulting 420-bp fragment was introduced into appropriately digested pTLuxN6His resulting in plasmid pTLuxNb6His. The mutation was verified by sequencing and detection of a new restriction site for DraI. For construction of pKKLuxNb6His a 2646bp fragment of luxN obtained by digestion of pTLuxNb6His with XmaI and HindIII was introduced into pKK223-3. Similarly, individual substitution of the amino acid residues at positions His 471 3 Gln and Asp 771 3 Asn was achieved by the PCR overlap extension method using synthetic oligonucleotide primers with the corresponding codons exchanged. PCR products were purified, digested with the appropriate restriction enzymes, and cloned into plasmids pTLuxN6His or pKKLuxNb6His, respectively. luxU was amplified by PCR from cosmid DNA using the primers LuxU-NdeI sense (5Ј-ACAAAACATATGAATAC-GGACGTATTAAAT-3Ј)/LuxU-BamHI antisense (5Ј-TTG-CTCGGATCCTTAGTTTGTCCAAGAACGGTA-3Ј). The fragment was digested with NdeI and BamHI and cloned into appropriately digested pET16b (Novagen) to construct pETLuxU10His. luxU was also amplified by PCR with the primers LuxU-BamHI sense (5Ј-AAAAGGATCCAATA-CGGACGTATTAAATCAG-3Ј) and LuxU-HindIII antisense (5Ј-TTGCTCGAAAGCTTAGTTTGTCCAAGAACG-3Ј) and cloned into plasmid pQE30 (Qiagen) resulting in plasmid pQELuxU6His. As described above, oligonucleotide-directed specific mutagenesis using the PCR overlap extension method was used to introduce codon exchanges corresponding to a replacement of His 58 3 Gln (13). PCR-generated DNA fragments were verified by sequencing.
Purification of LuxU-LuxU-His 6 was purified from the cytosolic fraction of E. coli JM109/pQELuxU6His. Cells were grown aerobically in KML medium at 37°C and induced with isopropyl 1-thio-␤-D-galactopyranoside (0.5 mM) during exponential growth (A 600 ϭ 0.8) for 3 h. Cells were harvested and lysed as described above. The His-tagged LuxU was purified from the cytosolic fraction by affinity chromatography using nickel-nitrilotriacetic acid-agarose (Qiagen) equilibrated with buffer E (10 mM imidazole, 50 mM Tris/HCl, pH 8.0, 10% (v/v) glycerol, 10 mM 2-mercaptoethanol, 0.4 mM Pefablock SC). Histagged protein was eluted with buffer E containing 250 mM imidazole. During the entire purification process the temperature was maintained at 4°C to minimize proteolysis.
Phosphorylation and Dephosphorylation Assays-Autophosphorylation was tested with LuxN in membrane vesicles. A typical reaction mixture (total volume 180 l) contained 8 mg/ml membrane proteins with LuxN-His 6 , and when indicated 0.1 mg/ml purified His 6 -LuxU, in phosphorylation buffer (50 mM Tris/HCl, pH 8.0, 10% (v/v) glycerol, 500 mM KCl, 2 mM DTT). The reaction was started by addition of radiolabled Mg 2ϩ -ATP, typically 100 M [␥-32 P]ATP (0.94 Ci/mmol) and 110 M MgCl 2 . The reaction mixture was incubated at room temperature, and at various time points the reaction was terminated by the addition of SDS loading buffer followed by separation on SDS-polyacrylamide gels. Gels were dried at 80°C on filter paper or blotted onto nitrocellulose membranes. Dried gels or membranes were exposed to a phosphoscreen for at least 24 h and subsequently scanned using a PhosphoimagerSI (Amersham Biosciences). Signal intensities of phosphorylated proteins were quantified in comparison to an [␥-32 P]ATP standard using ImageQuant V5.0.
For dephosphorylation assays, first, LuxU was phosphorylated using LuxN. In this case, the phosphorylation buffer contained 10 mM CaCl 2 instead of MgCl 2 and twice the amount of LuxN and LuxU. After 10-min incubation at room temperature, membrane vesicles were removed by centrifugation (100,000 ϫ g, 15 min, 4°C), and ATP and CaCl 2 were removed by gel filtration (Sephadex G25 columns, GE Healthcare). Dephosphorylation of LuxUϳP (0.6 mg/ml) was initiated by addition of 110 M MgCl 2 , 100 M ATP␥S, 2 and LuxN in membrane vesicles (8 mg/ml). As described above, samples were subjected to SDS-PAGE and exposed to a phosphoscreen.
Analytical Procedures-Protein was assayed by the method of Peterson (14) using bovine serum albumin as standard. Proteins were separated by SDS-PAGE (15) using 7.5 or 15% acrylamide gels. His-tagged Lux proteins were detected after immunoblotting with monoclonal anti-mouse antibodies against the His-tag (Qiagen) followed by incubation with alkaline phosphatase-conjugated anti-mouse IgG (Rockland) according to the manufacturer's instructions.

RESULTS
Characterization of the in Vitro Autophosphorylation Activity of LuxN-Full-length LuxN was overproduced in the F 1 /F 0 -ATPase negative E. coli strain TKR2000. In vitro phosphorylation of LuxN was examined in inverted membrane vesicles. Incubation of these vesicles in the presence of [␥-32 P]ATP led to very rapid, but short-lived autophosphorylation of LuxN (Fig. 2, A and E). The maximal phosphorylation level was reached in less than 1 min, after which the amount of LuxNϳP decreased. LuxN autophosphorylation activity was maximal in the presence of K ϩ ions (0.5 M) and under reducing conditions, whereas NaCl, even at low concentration, prevented autophosphorylation completely (data not shown). The turnover rate of LuxNϳP was unusually high. No phosphorylation of LuxN was detected using low concentrations of radiolabeled [␥-32 P]ATP with the highest specific radioactivity (3000 Ci/mmol) commercially available (data not shown). An increase of the ATP concentration led to an increase in LuxN autophosphorylation, which was maximal at about 2.5 mM ATP. Thus, for the kinase activity, a V max of 27 pmol/min/mg of protein, and an apparent K m for ATP of 555 M were determined. It should be noted that all phosphorylation experiments were performed with 20 -100 M [␥-32 P]ATP, which enabled significant labeling with reasonable levels of radioactivity. Consequently, the rates of LuxN phosphorylation measured were far below the V max determined for the autokinase activity. LuxN derivatives containing amino acid replacements of the two predicted phosphorylation sites, LuxN-H471Q (Fig. 2B), and LuxN-D771N (Fig. 2C) were tested for autokinase activity. Replacement of the conserved histidine residue (His 471 ) abolished autophosphorylation activity (Fig.  2B). The weakly phosphorylated protein observed in the figure is unrelated to LuxN. It runs at a slightly higher molecular mass on the gel than does LuxN. Moreover, in control vesicles lacking LuxN, a weakly phosphorylated protein of the same size was observed (Fig. 2D). In contrast, LuxN-D771N exhibited 2-3fold higher phosphorylation than wild-type LuxN (Fig. 2, C and  E). Western blots show that this difference cannot be attributed to higher production of the mutant LuxN protein (Fig. 2F), so we interpret this result to be a consequence of the increased stability of the phosphoamidate (HisϳP) than the mixed acid ester (AspϳP) under the alkaline conditions of an SDS gel (16,17). Furthermore, an anhydride bond has a higher energy state compared with a phosphoamidate bond. LuxN-D771N lacks this high energy phosphorylation site, and thus phosphorylation at His 471 seems to have increased stability. The autophosphorylation activity of LuxN-D771N combined with the lack of activity of LuxN-H471Q indicated that His 471 is phosphorylated by ATP in the transmitter domain before the phosphoryl group is transferred to Asp 771 located in the receiver domain.
The Hybrid Sensor Kinase LuxN Transfers the Phosphoryl Group to the Histidine Phosphotransfer Protein LuxU-Incubation of purified His 6 -LuxU with radiolabeled ATP did not result in phosphorylation of the protein (data not shown). In contrast, the phosphoryl group was transferred very fast from phosphorylated LuxN to LuxU resulting in accumulation of LuxUϳP (Fig. 3, A and C). As judged by Western blot (Fig. 3B), 50 times 2 The abbreviation used is: ATP␥S, adenosine 5Ј-O-(thiotriphosphate). more LuxU was present than LuxN in this experiment, which is consistent with the typical low levels of integral membrane histidine kinases found in cells compared with their soluble partners (18). Whereas phosphorylated LuxN was barely detectable, LuxUϳP was stable and increased constantly for 10 min. This result indicates that LuxU can be regarded as a sink for phosphate from LuxN. Using 100 M [␥-32 P]ATP as substrate, about 1% of the LuxU molecules were labeled at a rate of 0.83 pmol/min/mg protein in a LuxN-specific manner. No phosphotransfer was observed when LuxN-H471Q or LuxN-D771N was used (data not shown). Furthermore, replacement of His 58 with Gln in LuxU also abolished the phosphorylation of the protein (data not shown), confirming the proposed phosphorylation sites in both LuxN and LuxU.
Reconstitution of LuxN3LuxU Signal Transduction in Vitro-To mimic the in vivo situation, LuxN and LuxU were incubated together, and the reaction was initiated by addition of radiolabeled ATP. The time course of LuxU phosphorylation was followed for 60 min (Fig. 4, A and C). Following addition of ATP, phosphorylated LuxU was detected within 1 min, and it reached a maximum level after 10 min and declined thereafter. Phosphorylated LuxN was hardly detectable (data not shown). The identical experiment was performed following preincubation of the membrane vesicles containing LuxN with synthetically prepared AI-1 (50 M). 3 H-Labeled homoserine lactone is freely diffusible through the cytoplasmic membrane, and following exogenous addition it is found within 20 s in the cytoplasm (19). Nevertheless, to ensure that AI-1 reached the periplasmic domain in inside-out membrane vesicles, vesicles were preincubated (5 min) with AI-1. Preliminary tests revealed that an additional sonication step did not improve the results. The time course for LuxN-dependent phosphorylation of LuxU was essentially identical in the presence and absence of AI-1, how- Incubation was continued, and samples were withdrawn at the indicated times. Radiolabeled proteins were separated by SDS-PAGE followed by Western immunoblotting using antibodies against the His-tag (B). The dried membranes were exposed to a phosphoscreen for autoradiography (A). The arrows mark LuxU and LuxN. Phosphorylated LuxU was quantified with ImageQuant using [␥-32 P]ATP as standard (C ). ever, strikingly lower LuxUϳP was produced (60% reduction) when AI-1 was present compared with when it was absent (Fig.  4, B and C). Thus, the presence of AI-1 does not prevent LuxN autophosphorylation or phosphotransfer from LuxN to LuxU. Rather, AI-1 significantly reduces LuxUϳP production. Addition of AI-1-containing conditioned cell-free culture fluids produced comparable effects (data not shown). The concentrationdependent influence of AI-1 on LuxN3 LuxU signaling is shown in Fig. 4D. Half-maximal inhibition of LuxU phosphorylation was determined with 5 M AI-1. At concentrations higher than 20 M, a constant level of phosphorylated LuxUϳP was observed.
Autoinducer-1 Inhibits the LuxN Kinase Activity-We considered two possibilities for the effects of AI-1 on the levels of LuxU phosphorylation; AI-1 could affect LuxN kinase activity or LuxN phosphatase activity. To examine the effects of AI-1 on LuxN kinase activity, various concentrations of AI-1 were incubated with LuxN and the initial rates of autophosphorylation determined. Because of the weak phosphorylation signal obtained with wild-type LuxN (see Fig. 2), which is likely further reduced in the presence of AI-1, the LuxN-D771N derivative was used for this series of experiments to provide us a method with which to reliably quantify our data. The kinase activity of LuxN-D771N was significantly affected by AI-1 (Fig. 5). Increasing AI-1 concentrations successively decreased the rate of autophosphorylation of LuxN-D771N. The concentrationdependent influence of AI-1 followed the same course as was observed for the LuxN3 LuxU phosphorylation. The halfmaximal inhibition was determined to be 20 M AI-1. Despite the difficulties associated with the wild-type protein, we could in fact observe that AI-1 affected the wild-type LuxN kinase activity (Fig. 5). The phosphorylation signal for the wild-type was only slightly above background, which could account for the observed weak effects of AI-1 on wild-type autophosphorylation activity. Nevertheless, even the highest AI-1 concentration did not completely abolish the autophosphorylation activity of LuxN wild-type or LuxN-D771N.
LuxN Phosphatase Activity Is Not Regulated by AI-1-To determine whether the decline of LuxUϳP observed in the reconstituted cascade (see Fig. 4A) was due to effects of AI-1 on phosphatase activity of LuxN, we examined this activity. Purified LuxUϳP was incubated with LuxN at a ratio 50:1, and the time-dependent LuxUϳP dephosphorylation rate was 0.33 pmol/min/mg protein (Fig. 6, B and G). This result shows that indeed, LuxN has phosphatase activity on LuxU-P because LuxU-P alone was stable for at least 2 h (Fig. 6, A and G). LuxN phosphatase activity was independent of the presence of ATP (data not shown). The significance of the phosphorylation sites in LuxN were examined in the phosphatase assay. The LuxN-H471Q phosphatase activity was indistinguishable from the wild type (Fig. 6, D and G). However, no LuxUϳP dephosphorylation was observed when LuxN-D771N was used (Fig. 6, E and F) indicating that this aspartate residue is essential for the phosphatase activity of LuxN. Surprisingly, AI-1 had no effect on the phosphatase activity of LuxN. In all cases, the dephosphorylation rates in the presence of AI-1 were nearly identical to the rate determined in its absence (Fig. 6, B, C, E, and F).

DISCUSSION
LuxN belongs to the family of unorthodox sensor kinases that are equipped with two phosphorylation sites. Sequence alignment studies suggested His 471 and Asp 771 as the phosphorylation sites of LuxN. Consistent with this, genetic analyses of LuxN derivatives with point mutations at the putative sites of phosphorylation provided evidence for the important roles of the conserved amino acid residues His 471 in the transmitter domain and Asp 771 in the receiver domain (20). Moreover, investigation of the bioluminescence phenotypes of these and other lux mutants of V. harveyi led to the conclusion that LuxN catalyzes three enzymatic activities within a phosphorelay: (i) autophosphorylation and intramolecular phosphotransfer from His 471 to Asp 771 ; (ii) phosphotransfer from LuxN to LuxU, a histidine phosphotransfer protein; and (iii) dephosphorylation of LuxU. In the present report these proposed enzymatic activities of LuxN and the effect of AI-1 on them were investigated in vitro.
luxN was cloned and overexpressed in E. coli as a full-length protein including the N-terminal membrane-spanning domain. For in vitro phosphorylation assays, inverted membrane vesicles containing LuxN were incubated with radiolabeled [␥-32 P]ATP. Autophosphorylation of LuxN was found to be an unusually fast reaction and the labeling of LuxN was extremely short-lived. Moreover, autophosphorylation was only detectable at relatively high ATP concentrations (Ͼ20 M). All of these results indicate that the equilibrium favors the reactants ATP and unlabeled LuxN and not the products ADP and phosphorylated LuxN. No sensor kinase or hybrid sensor kinase characterized to date (21-23) exhibits such unusual kinetic behavior. The unusual LuxN autophosphorylation kinetics might be a specific requirement for executing the all-or-nothing quorum sensing transition in V. harveyi, whereas other behaviors controlled by sensor kinases do not require such rapid/extreme kinetics. In vitro studies with derivatives bearing amino acid replacements at the sites proposed for phosphorylation from genetic studies, His 471 and Asp 771 in LuxN and His 58 in LuxU (20,24), confirmed the phosphorelay consisting of LuxN-His 471 3 LuxN-Asp 771 3 LuxU-His 58 .
In contrast to the weak and unstable autophosphorylation of LuxN, the equilibrium is shifted toward the phosphorylated proteins in the presence of LuxU. Although the measured phosphorylation degree of LuxU was determined to be only 1%, this is in the range of the values observed for other sensor kinase/ response regulator proteins (18). The low degree of phosphorylation is largely due to the low ATP concentration used in in vitro experiments. For example, similar phosphorylation levels were obtained in vitro for the KdpD/KdpE system of E. coli, however, mathematical simulations predicted that a much higher phosphorylation level of the proteins would occur at physiological ATP concentrations (18). LuxU seems to act as a phosphate sink scavenging phosphoryl groups from LuxN (see further discussion below). Although we made numerous attempts to obtain LuxO (the final member of the Lux phosphorelay) as a soluble protein, we were unsuccessful. Therefore, it remains unclear whether LuxU retains its exquisite ability to scavenge phosphoryl groups in the presence of LuxO.
In the reconstituted system, in which reactions were initiated upon addition of ATP to LuxN and LuxU, rapid phosphorylation of LuxU was observed and declined afterward. The degree of LuxU phosphorylation in a reconstituted system is the result of all LuxN enzymatic activities, which are equilibrium reactions. Therefore, it is conceivable that the ratio of kinase/phosphatase activity changes over time due to the consumption of ATP and/or the accumulation of ADP. However, this rapid time course of phosphorylation is unusual in comparison with other two-component systems, like KdpD/KdpE (18) or EnvZ/ OmpR (22), which were similarly tested. However, one important difference between the LuxN/LuxU experiments and those with other two-component systems such as KdpD/KdpE and EnvZ/OmpR is that the latter were performed in the presence of the corresponding response regulators and the cognate DNA to which the response regulators bind. It has been shown that the presence of DNA shifts the equilibrium toward the phosphorylated response regulator (25). Despite the lack of a response regulator to test in the LuxN/LuxU signaling cascade, there are other explanations for the rapid decline of LuxUϳP. Because of the unusually high turnover rate of LuxN kinase activity, the substrate ATP might be rapidly exhausted in vitro. Alternatively, LuxN could have a high LuxUϳP phosphatase activity. Although our experiments revealed a phosphatase activity of LuxN with LuxUϳP as substrate, its activity was low compared with autokinase activity (see discussion below).
Tests of wild-type LuxN and the two derivatives of LuxN in which the amino acids corresponding to the phosphorylation sites are replaced showed that Asp 771 in LuxN is essential for the dephosphorylation of LuxUϳP. These biochemical data are in agreement with the phenotypic studies of the V. harveyi LuxN-D771A mutant (20). Phospho-LuxO represses bioluminescence expression in V. harveyi and V. harveyi luxN D771A produced bioluminescence in a density dependent manner, but no stimulation of light production was observed in the presence of AI-1 suggesting that the luxN D771A mutant lacks a phosphatase activity that can remove phosphate from LuxO (via LuxU) to induce bioluminescence expression (1,20). An important role for the analogous aspartate residues in both kinase and phosphatase activity was also reported for the hybrid sensor kinases ArcB and BvgS (26,27). These earlier studies provided the first evidence that the aspartate residue on sensor kinases is re-phosphorylated during dephosphorylation (17). Our investigations support this possibility, because we detected faint amounts of LuxNϳP in LuxUϳP dephosphorylation assays (data not shown).
The studies presented here were performed with full-length LuxN protein, which included all transmembrane helices and their corresponding intervening periplasmic regions. To mimic the endogenous environment of LuxN (the cytoplasmic membrane), the protein was embedded in vesicles for all of the assays in this work. This strategy allowed us to examine the different enzymatic activities of LuxN in vitro and most importantly to examine the influence the AI-1 signal molecule has on each of the signaling functions of LuxN. In contrast to earlier assumptions, we demonstrated here that the phosphatase activity of