The bioluminescent marine bacterium Vibrio harveyi controls light production using two parallel quorum-sensing systems. V. harveyi produces two autoinducers (AI-1 and AI-2), which are recognized by cognate membrane-bound two-component hybrid sensor kinases called LuxN and LuxQ respectively. Under conditions of low cell density, in the absence of autoinducer, the hybrid sensors are kinases, and under conditions of high cell density, in the presence of autoinducer, the sensors are phosphatases. These activities allow LuxN and LuxQ to modulate the level of phosphorylation of the response regulator protein LuxO. LuxO, in turn, controls the transcription of the genes encoding luciferase. The phosphorelay protein LuxU is required for signalling to LuxO. In this report, we present a genetic analysis of the activities of the AI-1 sensor LuxN. Point mutations and in frame deletions were constructed in luxN and recombined onto the chromosome of V. harveyi for in vivo phenotypic analysis. We show that the conserved histidine (H471) in the sensor kinase domain of LuxN is required for kinase activity but not for phosphatase activity. In contrast, the conserved aspartate (D771) in the response regulator domain of LuxN is required for both activities. Furthermore, the LuxN phosphatase activity is localized to the response regulator domain. Our results indicate that the LuxN kinase activity is regulated by the presence of AI-1, whereas the LuxN phosphatase activity is constitutive. We also show that signalling from the two V. harveyi quorum-sensing systems is not equivalent. AI-1 and LuxN have a much greater effect on the level of LuxO phosphate and therefore Lux expression than do AI-2 and LuxQ.
Bacteria have evolved elaborate mechanisms for sensing and adapting to changes in their environment. One widespread sensory transduction mechanism is the ‘two-component’ adaptive response system (Stock et al., 1989; Parkinson and Kofoid, 1992). Two-component systems are typically composed of a membrane-bound sensor kinase protein and a downstream response regulator protein. The sensor kinase protein is responsible for detecting a particular environmental stimulus and transducing a signal to the response regulator protein. The response regulator in turn is responsible for carrying out some appropriate action in response to this stimulus. In a prototypical two-component system, after interaction with a particular stimulus, the sensor kinase protein autophosphorylates at a conserved histidine residue. The phosphoryl group is subsequently transferred to a conserved aspartate residue in the response regulator protein. Phosphorylation of the response regulator protein alters its activity. Typically, response regulators are DNA-binding proteins that directly affect gene expression in response to a cognate stimulus. In addition to signal detection, autokinase activity and the ability to transfer phosphate to the response regulator protein, many sensor kinase proteins also dephosphorylate response regulator proteins in vitro and in vivo (Parkinson and Kofoid, 1992).
Increasingly complex two-component phosphorelay circuits are being identified in many different bacteria and lower eukaryotes (Arico et al., 1989; Iuchi et al., 1990; Burbulys et al., 1991; Appleby et al., 1996; Posas et al., 1996). These two-component phosphorelay circuits contain at least four modules. These modules can occur as discrete proteins or as domains of hybrid proteins. The components are a sensor kinase, a response regulator, a phosphorelay domain and, finally, a second response regulator. Signalling occurs as follows: the sensor kinase module autophosphorylates at a conserved histidine (H1) and the phosphoryl group is transferred to a conserved aspartate (D1) in the first response regulator module. The phosphoryl group is next passed to a conserved histidine (H2) in the phosphorelay module and, finally, the phosphate is transferred to a conserved aspartate (D2) in the second response regulator module. In circuits involving the four-step mechanism, only the kinase activity of hybrid sensor kinase proteins has been demonstrated in vivo, although in some cases phosphatase activity has been established in vitro (Uhl and Miller, 1994, 1996; Georgellis et al., 1997, 1998).
Quorum sensing, or the control of gene expression in response to cell density, involves the production and detection of extracellular signalling molecules called autoinducers. Genetic analysis of Vibrio harveyi shows that a complex, multichannel two-component phosphorelay circuit regulates quorum sensing in this luminous marine bacterium (for a review, see Bassler, 1999a, b). V. harveyi produces and responds to two autoinducers, AI-1 (N-3-hydroxybutanoyl-l-homoserine lactone; Cao and Meighen, 1989) and AI-2 (unknown structure), which are recognized by cognate sensors called LuxN and LuxPQ respectively. LuxN and LuxQ are membrane-bound hybrid sensor kinase proteins, and each contains both a sensor kinase module and a response regulator module (Bassler et al., 1993, 1994a). LuxP is a periplasmic protein bearing homology to the ribose and galactose binding proteins of Escherichia coli. LuxP and LuxQ are proposed to function together as the AI-2 sensor. Sensory signals emanating from both LuxN and LuxQ are channelled to the shared phosphorelay protein LuxU, and subsequently from LuxU to the response regulator protein LuxO (Bassler et al., 1994b; Freeman and Bassler, 1999a). LuxO phosphate negatively regulates the expression of the luciferase structural operon luxCDABE (Freeman and Bassler, 1999b).
Our genetic analysis suggests that under conditions of low cell density, in the absence of their cognate autoinducers, the LuxN and LuxQ sensors possess kinase activity (Freeman and Bassler, 1999b). Analogous to other two-component systems, we proposed that under these conditions the sensors autophosphorylate on their conserved histidine residues (H1). The phosphoryl group is next transferred intramolecularly to the conserved aspartate (D1) in the attached response regulator module. Next, the phosphoryl group is transferred to the conserved histidine (H2) in the phosphorelay protein LuxU. Finally, the phosphate is transferred to the conserved aspartate (D2) of the response regulator protein LuxO (Freeman and Bassler, 1999a). Phosphorylation of LuxO leads to repression of the expression of luminescence. In contrast, under conditions of high cell density, in the presence of the autoinducers, we proposed that the sensors switch activities from kinases to phosphatases. In this mode, LuxN and LuxQ promote the dephosphorylation of LuxO. Dephosphorylation of LuxO inactivates it and leads to light production (Freeman and Bassler, 1999b).
This report presents a genetic analysis of the functions of LuxN. We show that the interplay between the kinase and phosphatase activities of LuxN dictates the exact level of phosphorylation of LuxO and allows precise modulation of the lux structural operon. As mentioned, analogous to LuxN, the AI-2 sensor LuxQ has both kinase and phosphatase activity. In the present work, we use LuxQ either as a LuxO kinase or as a LuxO phosphatase. By driving LuxO into the phosphorylated or dephosphorylated form via LuxQ, we can subsequently test what effects different mutations have on the various activities of LuxN. These activities include AI-1 binding, kinase and phosphatase. Importantly, the ability to use the second sensor LuxQ to modulate the level of phosphate in the circuit allows us to examine specifically LuxN phosphatase activity in vivo.
Construction of in vivo luxN mutations
A genetic system was designed to facilitate recombination of mutant luxN alleles onto the V. harveyi chromosome. The strategy used was similar to one we reported previously to examine the signalling role of the LuxO protein in V. harveyi (Freeman and Bassler, 1999b). The strategy consists of two steps: first, the construction of a marked deletion of the lux gene of interest on the V. harveyi chromosome; and, second, the exchange of the deletion for a differently marked mutant allele of the same gene.
As described in Experimental procedures, a chloramphenicol-resistant (Cmr) marked deletion of luxN was constructed (called ΔluxN-Cmr) and used to replace the wild-type luxN gene in the V. harveyi chromosome. This strain is called BNL63, and the V. harveyi DNA encompassed by the deletion is shown in Fig. 1. A kanamycin resistance cassette (Knr) was inserted downstream of the cloned wild-type luxN gene, and this construction (luxN-Knr) was subsequently subjected to random and site-directed mutagenesis (below). Mutant luxN alleles were introduced onto the V. harveyi chromosome by allelic exchange for the ΔluxN-Cmr region of BNL63. Using the ΔluxN-CmrV. harveyi strain BNL63 as the recipient in these procedures ensured that the mutations of interest in luxN were recombined onto the chromosome along with the linked Knr marker.
To test the method, the ΔluxN-Cmr locus in strain BNL63 was replaced by the wild-type luxN-Knr allele. Figure 2 shows the Lux phenotypes of the wild-type V. harveyi strain (BB120), the ΔluxN-Cmr strain (BNL63) and the wild-type luxN-Knr replacement strain (JAF551). The presence of a functioning LuxN was tested by the addition of exogenous AI-1. In this experiment, overnight cultures of the V. harveyi strains to be tested were diluted, and cell density and light emission were measured during the subsequent growth of the cultures. Supernatant (10% cell-free) from V. harveyi strain MM30 was added as the source of AI-1 to specified cultures at the time of the dilution. V. harveyi MM30 does not produce AI-2 but produces wild-type levels of AI-1 (Surette et al., 1999). Figure 2 shows that light production in wild-type cells, in the absence of exogenously supplied autoinducer, decreased rapidly after dilution (open squares). The decline in light production after dilution occurs because the existing luciferase is diluted out as the cells double. However, during the growth of the culture, the wild-type strain produced endogenous autoinducers that accumulated in the medium and induced the expression of luciferase. Light emission then increased exponentially by more than 1000-fold. This is the characteristic quorum-sensing behaviour of V. harveyi. Figure 2 also shows that addition of exogenous AI-1 at the time of dilution caused the wild-type strain to induce light production at a much lower cell density (closed squares). Both the ΔluxN-Cmr strain, BNL63, and the luxN-Knr strain, JAF551, have phenotypes identical to the wild type in the absence of exogenously added autoinducer (open circles and open triangles respectively). However, the ΔluxN-Cmr deletion strain BNL63 did not respond to AI-1. Figure 2 also shows that this strain failed to prematurely induce light production after the addition of MM30 cell-free supernatant (closed circles). In contrast, after addition of AI-1, the luxN-Knr strain JAF551 did induce premature lux expression, similar to the wild-type strain BB120 (closed triangles). These data indicate that our strategy for the introduction of the luxN-Knr for the ΔluxN-Cmr was successful and that linkage of the downstream Knr cassette did not adversely affect LuxN function. We therefore assume that the analogous introduction of mutant luxN alleles linked to the downstream Knr marker into the chromosome of V. harveyi would result in informative in vivo phenotypes.
LuxN H471 is required for kinase activity but not for phosphatase activity
We have previously reported that both of the autoinducer sensors, LuxN and LuxQ, possess kinase and phosphatase activity (Freeman and Bassler, 1999b). To assess the role of the conserved histidine (H471) of LuxN in its kinase and phosphatase activities, we substituted glutamine for the histidine and subsequently recombined the mutant allele onto the V. harveyi chromosome, using the approach described above. The luxN H471→Q strain is called BNL104.
The Lux phenotype of V. harveyi BNL104 is shown in Fig. 3. As in the previous experiment, the mutant and wild-type V. harveyi strains were diluted into fresh medium and light emission was measured as the cultures grew. AI-1 was added by supplying 10% cell-free supernatants from V. harveyi strain MM30. Also, as in Fig. 2, the wild-type strain displayed the characteristic density-dependent Lux expression in the absence of AI-1 and was stimulated to produce light when exogenous AI-1 was added (open and closed squares respectively). In contrast, the luxN H471→Q mutant showed maximal constitutive luminescence in both the absence and the presence of AI-1 (open and closed triangles respectively). The phenotype of the luxN H471→Q mutant is consistent with the specific elimination of the kinase activity that is responsible for the phosphorylation of LuxO and the repression of light production, and the retention of the phosphatase activity that is responsible for the dephosphorylation of LuxO and the induction of light production.
Hsing and Silhavy (1997) demonstrated that amino acid substitutions at the conserved histidine of EnvZ had varying effects on phosphatase activity. All substitutions at this residue eliminated kinase activity; however, some substitutions (glutamine, asparagine and aspartate) greatly reduced phosphatase activity, whereas other substitutions (tyrosine, serine and alanine) had very little or no effect on phosphatase activity. The authors concluded that these differences were a result of the ability or lack thereof of the substituted amino acid to interact with a water molecule to serve as the phosphate acceptor in the phosphatase reaction (Hsing and Silhavy, 1997). We reasoned that perhaps substitutions other than glutamine at H471 of LuxN might affect phosphatase activity. We therefore constructed several other substitutions of LuxN H471, and we also made a number of site-directed mutations in the amino acids in the vicinity of H471 (Table 1). Our rationale for these changes was that the analogous substitutions were shown to affect phosphatase activity in EnvZ (Brissette et al., 1991; Tokishita et al., 1992). The Lux phenotypes of the various LuxN missense mutants are listed in Table 1. Unlike in EnvZ, no mutation at or near H471 of LuxN altered the LuxN phosphatase activity. The majority of mutations resulted in constitutive luminescence similar to H471→Q, i.e. these missense mutant LuxN proteins lacked kinase activity but retained phosphatase activity. The luxN S468→E mutation resulted in a wild-type phenotype. Apparently, this mutation did not affect either kinase or phosphatase activity.
Table 1. . Phenotypes of LuxN sensor kinase mutants.
Loss of AI-1 recognition results in constitutive kinase activity
Our results suggest that H471 of LuxN is required for kinase activity, but is dispensable for phosphatase activity. To identify mutations in LuxN that did not affect kinase activity, a random mutagenesis of luxN was performed followed by a screen for Lux− phenotypes in V. harveyi. Because LuxN kinase activity leads to the phosphorylation of LuxO and repression of light production, we predicted that two different types of dominant luxN mutants could cause a dark phenotype: first, LuxN mutations that specifically affected phosphatase activity; and, second, LuxN mutations that affected the switch from kinase to phosphatase.
The wild type luxN-Knr clone pJAF836 was transformed into the mutD5 strain CC130 (Manoil and Beckwith, 1985). The transformants were grown overnight and plasmid DNA was prepared, transformed into JM109 and subsequently conjugated into the wild-type V. harveyi strain, BB120. Exconjugants were screened for a Lux− phenotype. One dark exconjugant was identified, and the luxN-Knr cosmid was isolated and reintroduced into V. harveyi BB120 to verify that the dark phenotype was associated with the cosmid. Cosmid DNA containing this luxN-Knr allele was prepared and sequenced. The only mutation in the luxN gene was a transversion of a G to a C in codon 166, which alters a leucine to an arginine residue (L166→R).
The luxN L166→R allele was used to replace the ΔluxN-Cmr allele on the V. harveyi chromosome to generate strain JAF549. The phenotype of JAF549 is shown in Fig. 4. In this experiment, overnight cultures of wild-type and mutant V. harveyi strains were grown to high cell densities and light emission per cell was measured. Figure 4 shows that at high cell densities the wild-type strain BB120 is maximally bright and produces > 105 relative light units (RLU). The luxN L166→R strain JAF549, in contrast, emits 50 000-fold less light than wild type at ≈ 5 RLU. The LuxN L166→R mutation lies in the portion of the protein predicted to encode the transmembrane segments and periplasmic loops. We hypothesize that this is the sensory domain involved in AI-1 recognition. We further hypothesize that the dark phenotype associated with the LuxN L166→R mutation is a result of the inability of this protein to interact productively with AI-1.
This result suggests that the luxN L166→R mutation locks the LuxN protein into the low cell density/kinase mode. Furthermore, we have already reported that the luxN L166→R phenotype is dependent on an intact phosphorylation cascade. Specifically, the dark phenotype conferred by the luxN L166→R mutation requires the presence of the phosphorelay protein LuxU and the response regulator protein LuxO (Freeman and Bassler, 1999a). In this report, we have shown that the luxN H471→Q mutation eliminates kinase activity and locks the LuxN protein into the high cell density/phosphatase mode. If the luxN L166→R mutation results in constitutive kinase activity, then we predict that this activity should be dependent on H471. To test this hypothesis, we combined the luxN L166→R and H471→Q mutations and incorporated the double mutation onto the V. harveyi chromosome. The phenotype of this strain (JAF627) is shown in Fig. 4. The double luxN L166→R/H471→Q mutant is maximally bright, indicating that the luxN H471→Q mutation suppresses the dark phenotype conferred by the luxN L166→R mutation. Furthermore, in a density-dependent luminescence assay, the luxN L166R→/H471→Q strain displayed constitutive luminescence, a phenotype identical to the single luxN H471→Q mutation (data not shown).
Our results indicate that LuxN acts as a constitutive kinase and confers a dark phenotype when it cannot recognize AI-1. Therefore, we predict that a V. harveyi mutant incapable of AI-1 synthesis should have a phenotype identical to that conferred by the luxN L166→R mutation. We constructed two in frame deletions: one disrupting luxM and a second disrupting both luxL and luxM (see Experimental procedures). The luxLM genes are responsible for production of AI-1 and they lie immediately upstream of luxN in the V. harveyi chromosome (Fig. 1). The luxM and luxLM deletions were introduced onto the chromosome of V. harveyi by allelic exchange. The phenotypes of the ΔluxM strain JAF633 and the ΔluxLM strain JAF634 are shown in Fig. 4. Both deletions conferred a dark phenotype similar to that conferred by the luxN L166→R mutation. As shown in Fig. 4, addition of AI-1 to the deletion strains restored light production.
In a previous report, we showed that transposon insertions in the two genes responsible for AI-1 synthesis, luxL and luxM, did not result in a dark phenotype (Bassler et al., 1993). Because our former results are contradictory to the ones presented here for the ΔluxL and ΔluxLM mutants, we investigated the transposon insertion mutants further. Examination of the genetic organization of luxLMN suggests that the three genes are arranged in an operon. Additionally, assays to test the LuxN response to AI-1 in the luxL and luxM transposon insertion mutants indicated that only residual LuxN kinase activity remains. We interpret these data to mean that the luxL and luxM transposon insertions are partially polar on the expression of luxN.
LuxN phosphatase activity is localized to the response regulator module
Because mutations in the N-terminus and in the sensor kinase module of LuxN appear not to affect its phosphatase activity, we hypothesized that this activity could be localized to the response regulator module. To test this prediction, we constructed a single amino acid substitution in which the conserved aspartate (D771), predicted to be the site of phosphorylation, was altered to an alanine. This mutation was incorporated onto the V. harveyi chromosome by replacement of the ΔluxN-Cmr allele to make the luxN D771→A strain, BNL107. The phenotype of the luxN D771→A strain is shown in Fig. 5 along with the wild-type strain BB120. Similar to the description given for Fig. 2, the wild-type strain shows density-dependent expression of luminescence in the absence of AI-1, but is stimulated to produce high levels of light after the addition of AI-1 (open and closed squares respectively). Strain BNL107 also displays density-dependent luminescence in the absence of added AI-1 (open circles). However, when AI-1 was added to BNL107, no stimulation of light production was observed (closed circles). This is the luxN null phenotype.
The observation that a null phenotype results from the alteration of D771 suggests that this residue is involved in both the kinase and phosphatase activities of LuxN. In order to show that D771 is required in the kinase reaction, we constructed a LuxN double mutant. We combined the luxN L166→R mutation with the luxN D771→A mutation. In the preceding section, we have argued that the luxN L166→R substitution results in a dark phenotype because it blocks signal recognition and subsequently locks the sensor into the kinase mode, and this phenotype requires a complete signalling circuit. We therefore reasoned that, if D771 in the response regulator module of LuxN is required for phosphotransfer to LuxO, the dark phenotype conferred by the luxN L166→R mutation should also depend on D771. The phenotype of the double luxN L166→R/D771→A mutant is shown in Fig. 4. Whereas the single luxN L166→R mutant is dark, the double mutant produces wild-type levels of light, indicating that the D771→A mutation is epistatic to the L166→R mutation and that D771 is required for phosphotransfer to LuxO.
Because D771 was required for both kinase and phosphatase activity, it is possible that the LuxN phosphatase activity is localized to the response regulator module and that the sensor kinase module has no role in this activity. To test this hypothesis, we constructed an in frame deletion in the N-terminus of luxN to generate a luxN allele encoding only the response regulator domain. The response regulator construction, referred to as luxN RR, was incorporated onto the chromosome of V. harveyi and the corresponding strain is called JAF636. The phenotype of strain JAF636 is shown in Fig. 5. This figure shows that expression of the LuxN response regulator module alone results in maximal constitutive luminescence (open triangles). This phenotype is similar to that observed for the luxN H471 mutants and is consistent with a lack of kinase activity and the presence of phosphatase activity.
To examine the role of D771 in the phosphatase activity, we next altered D771 to an alanine in the luxN RR construction and incorporated this allele onto the V. harveyi chromosome. The V. harveyi strain carrying this mutation (luxN RR D771→A) is called JAF635. The phenotype of JAF635 is also shown in Fig. 5. Strain JAF635 does not display constitutive light production (closed triangles). Rather, JAF635 shows density-dependent luminescence, indicating that without D771 the response regulator module has no phosphatase activity.
In a previous report, we presented genetic data suggesting that the AI sensors LuxN and LuxQ possess both kinase and phosphatase activity (Freeman and Bassler, 1999b). We have suggested that under conditions of low cell density, in the absence of autoinducer, the sensors act as kinases, phosphorylating the response regulator protein LuxO. We have demonstrated that phosphorylation of LuxO leads to the repression of light production. In contrast, under conditions of high cell density, the two sensors act as phosphatases, dephosphorylating LuxO. Once dephosphorylated, the LuxO protein is inactivated and the expression of luminescence occurs. Furthermore, we have shown that the phosphorylation of LuxO is dependent upon the phosphorelay protein LuxU (Freeman and Bassler, 1999a). In the present report, we have used a genetic approach to examine the activities of the AI-1 sensor LuxN in order to understand how the LuxN sensor switches between its kinase and phosphatase activities.
We first examined the role that the conserved histidine in the sensor kinase module of LuxN has on its various activities. Because H471 is the proposed site of autophosphorylation, we reasoned that it should be required for kinase activity. We made several different substitutions for H471 and recombined each onto the chromosome of V. harveyi. The results in Fig. 3 and Table 1 show that mutation of H471 resulted in constitutive luminescence, suggesting that the LuxN sensor had lost kinase activity but retained phosphatase activity. Furthermore, the phosphatase activity of LuxN in the H471 mutants functions constitutively. We therefore conclude that LuxN H471 is not required for phosphatase activity. Our results are in contrast to those reported for the two-component sensor kinase EnvZ. In the case of EnvZ, the conserved histidine is not absolutely required for phosphatase activity; however, certain substitutions at this residue did have severe effects on phosphatase activity (Hsing and Silhavy, 1997). These results indicated that the conserved histidine of EnvZ is involved in the phosphatase reaction.
Although in vivo phosphatase activity has not been reported for any four-step phosphorelay besides Lux, in vitro phosphatase activity has been demonstrated in the hybrid sensors ArcB and BvgS (Uhl and Miller, 1996; Georgellis et al., 1998). The ArcB and BvgS proteins each contain a sensor kinase module, a response regulator module and a phosphorelay module. The phosphatase activities of ArcB and BvgS have been shown to be independent of their sensor kinase modules, but dependent on their response regulator and phosphorelay domains. In addition, the ArcB and BvgS phosphatase activities were dependent on the conserved aspartate in the response regulator modules. In the case of LuxN, Figs 4 and 5 show that mutation of D771 abolished both LuxN kinase and phosphatase, indicating that D771 is required for both activities. Furthermore, in vivo expression of the LuxN response regulator module alone resulted in constitutive phosphatase activity, and mutation of D771 eliminated this activity (Fig. 5). Collectively, the results with the luxN D771→A, luxN RR and luxN RR D771→A mutants suggest that the LuxN phosphatase activity is localized to the response regulator module and that this activity is dependent on the conserved aspartate D771. Therefore, we propose that the phosphatase mechanism of LuxN is more similar to that of the ArcB and BvgS sensors than to that of the EnvZ sensor. Perhaps localization of phosphatase activity to the response regulator module of these complex proteins may be a trait common to four-step phosphorelays.
We wanted to understand how the LuxN protein switches between the kinase and phosphatase modes. Using a random mutagenesis followed by a screen for dark phenotypes, we identified the luxN L166→R mutation (Fig. 4). The luxN L166→R mutation is in the N-terminal region of LuxN proposed to be involved in AI-1 recognition. We hypothesize that this mutation results in a ‘signal blind’ phenotype in which the LuxN sensor cannot recognize AI-1 and is therefore locked in the kinase mode. Consistent with this hypothesis, two mutants defective in AI-1 synthesis had the identical phenotype (Fig. 4). Because the phenotype of the luxN L166→R mutant indicates that in the absence of AI-1 the LuxN kinase is constitutive, we conclude that interaction with AI-1 leads to the inactivation of the LuxN kinase.
In contrast to the dark luxN L166→R mutant, the luxN H471→Q mutant displays constitutive phosphatase activity, indicating that when the LuxN kinase activity is abolished, in this case by mutation, the phosphatase activity is unmasked. Together, these results indicate that AI-1 exerts its effect at the level of regulation of the kinase activity, and that the phosphatase activity is constitutive. If so, the LuxN kinase activity must be dominant to the LuxN phosphatase activity. In further support of this hypothesis, we combined the luxN L166→R and the luxN H471→Q mutations. The L166→R mutation blocks signal recognition and locks the sensor into the low cell density mode, resulting in constitutive kinase activity. The H471→Q mutation, on the contrary, eliminates kinase activity and locks the sensor into the high cell density mode, resulting in constitutive phosphatase activity. The double mutant displayed the luxN H471→Q phenotype, i.e. constitutive light production. This result suggests that the phosphatase activity is present even when the sensor is locked into the low cell density state. From these results, we conclude that the LuxN kinase activity is regulated by the presence or absence of autoinducer and that the LuxN phosphatase activity is not regulated.
In Fig. 6, we present a model depicting the two states of the AI-1 sensor LuxN. Under conditions of low cell density (Fig. 6A), when no or little AI-1 is present, the LuxN sensor kinase autophosphorylates on the conserved histidine (H1) in the sensor kinase domain. This phosphoryl group is subsequently transferred to the conserved aspartate (D1) in the response regulator domain. Phosphate is next transferred from LuxN to the conserved histidine (H2) in LuxU then to the conserved aspartate (D2) in the response regulator domain of LuxO. Phosphorylation of LuxO then activates its repressor function and no light is produced. Because we propose that LuxN phosphatase activity is constitutive, the LuxN kinase activity must be greater than the phosphatase activity, leading to the phosphorylation of LuxO. Under conditions of high cell density, in the presence of AI-1 (Fig. 6B), the kinase activity of LuxN is inactivated. However, the phosphatase activity of the response regulator domain remains. Dephosphorylation of LuxN serves to drain phosphate from LuxU and LuxO. Based on the ArcB and BvgS results in which the aspartate equivalent to D771 of LuxN is phosphorylated in the reverse reaction (Uhl and Miller, 1996; Georgellis et al., 1998), we suggest that D771 of LuxN is the point of hydrolysis in the phosphatase reaction. Furthermore, we propose that LuxU couples phosphotransfer from LuxO to LuxN via the histidine residue at position 58. Ultimately, the concentration of LuxO phosphate is reduced and dephosphorylation of LuxO inactivates it, allowing the production of light.
We hypothesize that the function of the LuxU phosphorelay protein is to establish an equilibrium between the aspartyl-phosphate of LuxN and the aspartyl-phosphate of LuxO. LuxN kinase results in a net influx of phosphate and leads to the phosphorylation of the LuxO protein, whereas LuxN phosphatase results in a net efflux of phosphate and leads to the dephosphorylation of LuxO. We postulate that LuxN does not interact directly with LuxO; that function belongs to LuxU.
Although this study focused on the AI-1 sensor LuxN, the AI-2 sensor LuxQ appears to function in an analogous manner. We already know that LuxQ acts as a kinase in the absence of AI-2 and as a phosphatase in the presence of AI-2 (Bassler et al., 1994a; Freeman and Bassler, 1999b); however, the amino acids responsible for these activities have not been identified. Most certainly, there are situations in which the two sensors have opposing activities. For example, when exogenous AI-1 is added to diluted wild-type cells, the LuxN sensor acts as a phosphatase, whereas the LuxQ sensor acts as a kinase. The ultimate level of LuxO phosphate then becomes dependent on the relative activities of the two sensors as well as the relative abundance of the sensors. Furthermore, the inputs from the two sensors are not equivalent (Freeman and Bassler, 1999b). Results described here indicate that the LuxN sensor plays a more important role than the LuxQ sensor in dictating the level of LuxO phosphate. For example, we showed that the luxN H471→Q mutant constitutively produces light. In this mutant, under conditions of low cell density, the LuxQ sensor is acting as a kinase, but the phosphatase activity of LuxN H471→Q overrides this input, resulting in no LuxO phosphate-mediated repression of light production. In an opposite example, the luxN L166→R mutant has a dark phenotype. In this case, at high cell density, the LuxN sensor is a constitutive kinase and this activity overrides the LuxQ phosphatase activity.
In its natural environment in the ocean, V. harveyi lives in a mixed population with other species of bacteria. We have reported that V. harveyi uses the AI-1/LuxN system for intraspecies communication and the AI-2/LuxQ system for interspecies cell–cell signalling (Bassler et al., 1997). Results presented here indicate that the impact of these two systems on the expression of luminescence is not equivalent. Indeed, it appears that the AI-1/LuxN system has a greater influence on light production than the AI-2/LuxQ system. Perhaps in the wild, this interesting and complex signal integration process, in which intraspecies signals result in a different output from interspecies signals, provides some specific survival benefit to V. harveyi.
Bacterial strains and media
The bacterial strains and plasmids used in this study are listed in Table 2. V. harveyi strains were grown in liquid HI medium at 30°C with aeration for genetic manipulation, chromosomal DNA preparation and polymerase chain reaction (PCR) analysis (Bassler et al., 1993). Lux assays were performed in AB medium as has been previously described (Bassler et al., 1993). The recipe for AB has been reported previously (Greenberg et al., 1979). Plasmids used for mutagenesis and sequencing were maintained in E. coli JM109 [supE, Δ(lac-proAB), hsdR17, recA1, F′traD36, proAB+, lacI q, lacZΔM15]. The techniques used for conjugation of recombinant clones into and out of V. harveyi and for allelic replacement onto the V. harveyi chromosome have been described previously (Bassler et al., 1993; Freeman and Bassler, 1999b). The plasmids used for mobilizing V. harveyi genes and for integrating lux mutations onto the chromosome were pRK2013 (Ditta et al., 1980) and pPH1JI (Beringer et al., 1978). These plasmids were maintained in E. coli CC118 (araD139, Δara, leu76a7, ΔlacX74, ΔphoA20, galE, galK, thi, rpsE, rpoB, argE, Am, recA1). All E. coli strains were grown at 37°C in liquid Luria–Bertani (LB) medium that contained 10 g l−1 bactotryptone, 5 g l−1 bactoyeast extract, 10 g l−1 NaCl. Antibiotics were used at the following concentrations; 100 mg l−1 ampicillin, 10 mg l−1 tetracycline, 100 mg l−1 kanamycin, 10 mg l−1 chloramphenicol, 100 mg l−1 gentamicin and 100 mg l−1 rifampicin.
All DNA isolations, manipulations and transformations were performed as described by Sambrook et al. (1989). Restriction endonucleases and T4 DNA ligase were purchased from New England Biolabs. Taq DNA polymerase, calf intestinal alkaline phosphatase and lysozyme were obtained from Boehringer Mannheim Biochemicals. Pfu DNA polymerase and ligase were purchased from Stratagene. All enzymes were used under the conditions recommended by the manufacturer. Oligonucleotides were synthesized by the Midland Certified Reagent Company. Site-directed mutagenesis was performed using one of three methods: the Altered Sites II strategy (Promega), the Quik Change system (Stratagene) or a PCR-based method adapted from Michael (1994). Sequencing was performed by the Princeton University sequencing facility. PCR analyses of luxN allelic replacement strains were performed on boiled V. harveyi cells, using the following primers that flank luxN: 5′-GCGCTGCAGATGTTT GATTTTAGCCTAGAGGC-3′ and 5′-GCGGTACCCTATTC TCTCTCAGCTTCACAAGCG-3′. Southern blot analysis was performed to verify all chromosomal mutations. DNA probes were radiolabelled by the incorporation of [32P]-dATP (Dupont, NEN), using the multiprime DNA labelling procedure (Amersham). The probe for luxN was the 2500 bp PCR product produced using the above primers. The probe for the Cmr marker was a 3.6 kb BamHI DNA fragment isolated from plasmid pHP45Ω (Fellay et al., 1987).
A V. harveyi PstI genomic DNA fragment containing the luxLMN genes (Fig. 1) was inserted into a Tetr, Ampr version of plasmid pALTER called pJAF409. This construction was subsequently digested with EcoRI followed by religation to delete the DNA between two endogenous EcoRI sites that flank luxN. The resulting ΔluxN construction is called pJAF416. The EcoRI fragment that was deleted from pJAF416 was cloned into a version of the pALTER vector in which the BamHI site in the multiple cloning site had been removed. Next, a BamHI restriction site was introduced 65 bp 3′ to luxN using the primer 5′-CACCGTGATGAATCAAGCCC-3′. This construction is called pJAF573. A second BamHI restriction site was introduced 225 bp 5′ to luxN using the primer 5′-CAGGTTACAAAATTGGATCCACTATTATCGAGC-3′. The DNA between the two BamHI sites was removed by restriction digestion and a BamHI DNA fragment containing a chloramphenicol resistance cassette (Cmr) from plasmid pPHP45Ω was inserted. The EcoRI DNA fragment containing the ΔluxN-Cmr construction was subsequently cloned into the EcoRI site of pJAF416 in order to increase the flanking V. harveyi DNA. In a final step, the ΔluxN-Cmr construct was cloned into the broad-host-range cosmid pLAFR2 (Friedman et al., 1982) to create the ΔluxN-Cmr construct pBNL067.
Construction of Knr marked mutant luxN alleles
Using the primers 5′-GCGGGATCCCACTCCGCTAGCGCT GAGGTCTG-3′ and 5′-GCGGGATTCCTGCCAGTGTTACA ACCAATTAAC-3′, the kanamycin resistance (Knr) gene flanked with BamHI sites was amplified from plasmid pACYC177. The Knr was cloned into the BamHI restriction site introduced 3′ to luxN in plasmid pJAF573, resulting in pJAF747. To construct the wild-type Knr marked luxN allele, the flanking DNA was increased by cloning into pJAF416, followed by insertion into pLAFR2. This construct is called pJAF836. luxN missense mutations were generated in plasmid pJAF573 using the method of Michael (1994). In these experiments, the luxN DNA was amplified using the primers 5′-CTCGAACGAGGAAATCTCAGC-3′ and 5′-CCCATTCAT AACAGGCATTTGTACATCCATCAGAATCAAATC-3′ in combination with one of the luxN mutagenic primers listed in Table 3. In order to construct random mutations at H471, the degenerate primer in Table 3 was used. All mutations were verified by sequencing. The Knr cassette was next cloned into the BamHI site downstream of each mutant luxN allele. Subsequently, the flanking V. harveyi DNA was increased by cloning into pJAF416 followed by cloning into pLAFR2.
Table 3. . Oligonucleotides used to construct luxN missense mutations.
Expression of our mutant luxN constructions in V. harveyi resulted in measurable Lux phenotypes, i.e. the luxN alleles encode proteins retaining kinase and/or phosphatase activity. We interpret this to mean that the stability of the LuxN proteins has not been affected by the various mutations.
Construction of the luxN RR and luxN RR D771→A alleles
Two MluI restriction sites were introduced into the luxN open reading frame using the Stratagene Quik Change SDM system. The 5′ site was introduced at amino acid 6 using the primers 5′-GGGATGAGCATGTTTGATACGCGTCTAGAGG CTATCGTC-3′ and 5′-GACGATAGCCTCTAGACGCGTATC AAACATGCTCATCCC-3′. The 3′ site was introduced at amino acid 719 using the primers 5′-GTACAAATAAATAACA CGCGTCCAACAGTGCTTATCGTC-3′ and 5′-GACGATA AGCACTGTTGGACGCGTGTTATTTATTTGTAC-3′. Plasmid pJAF573 was used as the parent for luxN RR, and the pJAF573 derivative containing the luxN D771→A mutation was used as the parent for luxN RR D771→A. After the introduction of the two MluI sites, the DNA between the sites was removed by restriction digestion to create the N-terminal in frame deletions. The Knr cassette was subsequently inserted downstream of each deletion. The flanking V. harveyi DNA arms were then re-established followed by cloning into pLAFR2.
Construction of the ΔluxM and ΔluxLM alleles
An EcoRI restriction site was introduced 38 codons upstream of the termination codon of luxM using the primers 5′-GGCAAACAGACTTACCGTGAATTCTGGAATTTTGAAAT G-3′ and 5′-CATTTCAAAATTCCAGAATTCACGGTAAGT CTGTTTGCC-3′. The DNA between this site and the endogenous EcoRI site located within luxM at codon 89 was next deleted by restriction digestion to create the ΔluxM construction. After engineering of the first EcoRI site, a second EcoRI site was introduced five codons downstream of the translational start codon of luxL using the primers 5′-CTGTTAGCCGACAGGAATTCTAATGACAACATTAAT TTC-3′ and 5′-GAAATTAATGTTGTCATTAGAATTCCTGT CGGCTAACAG-3′. The DNA between the two engineered EcoRI sites was removed to create the ΔluxLM construction. The two deletion constructions were then inserted into pLAFR2 for allelic exchange in the V. harveyi chromosome.
Random mutagenesis of the luxN-Knr construct
The wild-type luxN-Knr clone pJAF836 was transformed into the E. coli mutD5 strain CC130. The transformants were grown overnight in liquid culture and cosmid DNA was subsequently prepared. This DNA was next transformed into E. coli JM109, and the transformants were grown in liquid culture to stationary phase. The pooled transformants were conjugated into wild-type V. harveyi strain BB120 and screened for a dark (Lux−) phenotype. One exconjugant displayed a dark phenotype and the cosmid containing the luxN-Knr allele (luxN L166→R) responsible for the phenotype was isolated and sequenced. The mutant cosmid was also reintroduced into wild-type V. harveyi strain BB120 to verify the phenotype.
This work was supported by the National Science Foundation grant no. MCB-9506033 and The Office of Naval Research grant no. N00014-99-0767. We thank Dr T. J. Silhavy for helpful discussions and for reading this manuscript.