INTRODUCTION

Two-component signal-transduction systems, consisting of sensor-transmitter and receiver-regulator proteins, are ubiquitous in the bacterial world, where they mediate adaptive responses to diverse environmental stimuli (for reviews, see Refs. 1, 2, 3, 4). A general mechanism for signal transduction by two-component systems has been proposed, based in part on similarities in domain organization among family members (1, 2). The N-terminal domain of the sensor-transmitter component senses environmental signal(s) and modulates enzymatic activities of the C-terminal domain. As a result, a phosphate is transferred from a histidine residue in the C-terminal transmitter domain to the N-terminal domain of the receiver-regulator component. This, in turn, results in a regulatory (usually transcriptional) response by the C-terminal regulator domain. The sensor-transmitter is generally a transmembrane protein with an extracellular N-terminal sensor and an intracellular C-terminal transmitter, enabling the system to sense external stimuli and transduce information to the cytoplasm. Autokinase and phosphate transfer/phosphatase reactions involving the transmitter and receiver domains have been studied extensively in several systems (1, 2, 3, 4). By contrast, much less is known about how the sensor domain recognizes and responds to extracellular signals.

Signal detection and transduction by sensor-transmitters resemble similar processes in bacterial chemoreceptors (1, 2, 3, 4). Both share a common topology and the ability to mediate signal transduction through the regulation of a histidine kinase, although the kinase activity resides in a separate protein in the case of chemoreception. A fusion of the periplasmic domain of the Tar chemoreceptor and the cytoplasmic domain of the EnvZ osmosensor-transmitter results in a protein that activates transcription through the receiver-regulator of EnvZ in response to aspartate, the normal Tar ligand (5). This supports a similar mechanism for signal transduction for these two systems. The periplasmic, ligand-binding domain of Tar forms a dimeric, four--helix bundle (6). The periplasmic sensor domain of the Agrobacterium tumefaciens VirA sensor-transmitter has also been proposed to form a four-helix bundle (7).

Bacterial pathogens use two-component systems to sense host signals and respond by expressing virulence factors necessary for infection (8). phoP-phoQ is one such two-component regulatory pair in Salmonella typhimurium (9). Mutations in either gene raise the LD₅₀ for mouse infection (>10⁴-fold) and reduce bacterial survival within macrophages (9). Véscovi et al. (10) have proposed that the PhoQ sensor binds divalent cations, which act as signaling ligands to induce a conformational change that affects the kinase and/or phosphatase activities of the cytoplasmic transmitter (10). Presumably, high Mg²⁺ and Ca²⁺ levels (~1 mM) in mammalian extracellular fluids repress expression of virulence factors prior to engulfment by macrophages. Once inside the macrophage, however, S. typhimurium resides in phagosomes where low divalent cation levels (<100 µM) activate the PhoP-PhoQ system, resulting in the synthesis of proteins necessary for bacterial survival (10).

The PhoP-PhoQ system is also present in many non-pathogenic, Gram-negative bacteria including Escherichia coli, where a basic physiological role in response to Mg²⁺ starvation has been proposed (10). PhoP-PhoQ is an attractive system for studying ligand-induced signal transduction, since relatively few ligands have been identified in other two-component systems (1, 2). The PhoP and PhoQ proteins of E. coli and S. typhimurium are 93% and 86% identical, respectively (11), suggesting a large degree of structural and functional similarity. In this work, we describe the purification and characterization of the E. coli PhoQ sensor domain and a mutant domain, which is impaired in its response to divalent cations. The PhoQ sensor domain, unlike the corresponding domain of Tar, is monomeric and contains -sheet as well as -helices. The mutant, in which a cluster of acidic amino acids is substituted by isosteric, uncharged residues, is similar to wild type in its spectral properties and oligomeric state but is substantially more stable in urea denaturation experiments. The mutant, unlike wild type, is not stabilized by divalent cations and is defective in its response to divalent-cation starvation in vivo. These data support a model in which divalent cations bind to the acidic cluster of the wild-type sensor and stabilize an inactive conformation. Substitution of the acidic cluster with uncharged residues appears to stabilize the inactive conformation even in the absence of ligand.

EXPERIMENTAL PROCEDURES

Several plasmids were constructed for these studies. pNL3 (Fig. 1A) is a pBR322-derived reporter plasmid used to assay PhoP-mediated transcriptional activation. The regulatory region of the acid phosphatase gene (phoN; Ref. 12) of S. typhimurium strain MS1363 (13) was amplified from genomic DNA by polymerase chain reaction (PCR)¹ and fused to the lacZ structural gene from mini-Tn5 lacZ1 (14). pLPQ2 (Fig. 1A) is a pSC101-derived plasmid in which phoPphoQ expression is driven by the lacUV5 promoter. The lacUV5 promoter region from pJH370 (15) was fused to the phoPphoQ operon from Kohara phage 7F9 (16) at an NdeI site introduced by directed mutagenesis to overlap the initiating ATG codon of phoP. pPQts (Fig. 1A) is a pSC101-derived plasmid with a temperature-sensitive replicon and internal deletion in phoQ. Kohara phage 7F9 DNA was cut with EcoRI, and the 5 protruding ends were filled in with DNA polymerase Klenow fragment and dNTPs. The DNA was then cut with SalI, and the ~3.8-kilobase pair phoPphoQ fragment was cloned into the SalI-HindIII (filled in) backbone of pMAK705 (17). This plasmid was cut with ScaI and PaeR7I, filled in, and religated, creating an internal deletion in phoQ from the second base pair of codon 25 to the second base pair of codon 396. pAED4Q (Fig. 1A) is a pUC19-derived plasmid used to express the PhoQ sensor domain (residues 43-190) for purification. A DNA fragment encoding the periplasmic domain of phoQ was generated by PCR using primer A (5-CGGGAGGTGCATATGAGTTTCGATAAAACTACGTTTCGG-3), primer B (5-CGCGAATTCTTACATATAGGAACTTTTTAGCTCCACCGG-3), and chromosomal DNA isolated from E. coli strain X90 (18) as a template. The PCR product was cut with NdeI and EcoRI (generated by primers A and B, respectively) and cloned into the NdeI-EcoRI backbone of pAED4 (a gift from Don Doering; Ref. 19). The resulting plasmid (pAED4Q) has an initiating ATG codon (at the synthetic NdeI site) fused to codon 43 of phoQ and a translational termination codon after codon 190 of phoQ (Fig. 1B). In this plasmid, expression of the sensor domain (residues 43-190) is driven by a plasmid-borne T7 10 promoter and ribosome-binding site. The primary structure of the entire phoQ-SD gene was confirmed by DNA sequencing. Details of plasmid constructions are available upon request.

Sc/Pa

Construction of CSH26Q

An internal deletion in the chromosomal phoQ gene of E. coli strain CSH26 (20) was created by the gene replacement method of Hamilton et al. (17). E. coli strain CSH26/pPQts was grown on LB plates supplemented with 25 µg/ml chloramphenicol at 44 °C to select for integration of the temperature-sensitive plasmid into the chromosome. Cm^R colonies were inoculated into 25 ml of LB broth and grown overnight at 30 °C to allow resolution of cointegrates. Cultures were plated on LB-chloramphenicol plates and grown at 30 °C. Individual colonies were isolated and plasmid DNA extracted for restriction analysis. Colonies with plasmids containing an intact phoQ gene were assumed to have undergone a gene replacement event and were grown on LB plates at 44 °C to promote plasmid loss. Isolated candidate colonies were screened for Cm^S, indicating loss of the plasmid, and replacement of wild-type phoQ with the Q deletion was confirmed by PCR analysis of chromosomal DNA.

A variant of the PhoQ sensor domain in which residues 149-155 were changed from EDDDDAE  QNNNNAQ (Fig. 1B) was constructed as follows. A PCR fragment was generated using 5-CGGGAGGTGAGTTTCGATAAAACTACGTTTCGG-3 and 5-GGCTGCGGTCATCTTCGCATTGTTATTATTTTGCCGCAC-3 primers (mutant codons are italicized) and pAED4Q as a template. The PCR product was cut with NdeI and DraIII (underlined in primers) and cloned into the corresponding NdeI-DraIII sites of pAED4Q, which are present at the initiator ATG codon and within phoQ, respectively. The mutations were confirmed by DNA sequencing and then cloned into pLPQ2 (Fig. 1A) to assay their effects in vivo.

The PhoQ sensor domain (residues 43-190) was purified from E. coli strain X90(DE3) transformed with plasmid pAED4Q (or a variant containing the acidic-cluster mutations). Cells with phoQ-SD under control of the T7 10 promoter were treated with 0.5 mM IPTG, to induce expression of T7 RNA polymerase from the lacUV5 promoter on the DE3 prophage, harvested, and proteins were released by the freeze-thaw method (21). Cellular debris was removed by centrifugation. DNA was precipitated by the addition of polyethyleneimine (1% v/v) and removed by centrifugation. Ammonium sulfate was added (55% w/v), and following centrifugation, the pellet was resuspended and dialyzed against 20 mM Tris·Cl (pH 8.0). This solution was loaded onto a MonoQ (Pharmacia) anion-exchange column and eluted with a linear gradient from 0 to 1.0 M NaCl in 20 mM Tris·Cl (pH 8.0). Fractions containing the PhoQ sensor domain (as assayed by gel electrophoresis and Coomassie Blue staining) were pooled, dialyzed against 20 mM sodium phosphate (pH 8.0) and loaded onto an Affi-Gel Blue column (Bio-Rad). The column was washed with 20 mM sodium phosphate (pH 8.0) plus 0.1 M NaCl, and the protein was eluted with 20 mM sodium phosphate (pH 8.0) plus 2 M NaCl. The sensor domain was dialyzed against 10 mM sodium phosphate (pH 7.5) plus 0.2 M NaCl and judged to be greater than 95% pure by Coomassie Blue staining of a sample following electrophoresis on a 15% Tris-Tricine gel (22). Protein concentration was calculated using an extinction coefficient of 27,024 M¹ cm¹ at 280 nm based on the presence of 4 tryptophans and 4 tyrosines in the sequence (23). N-terminal sequencing and amino acid analysis were performed by the MIT Biopolymers lab.

Circular dichroism (CD) measurements were performed in SB buffer (10 mM sodium phosphate (pH 7.5) plus 0.2 M KCl) in the presence and absence of 10 mM MgCl₂ at 25 °C using an AVIV 60DS instrument equipped with a temperature-controlled sample holder. Thermal denaturation for the wild-type PhoQ sensor domain was monitored by changes in CD ellipticity at 218 nm and the T_m was calculated by nonlinear least-squares fitting of the data using a unimolecular unfolding model (24, 25). Fluorescence measurements were performed in SB buffer with or without divalent salts at 25 °C using an Hitachi F-4500 fluorescence spectrophotometer. The sample was excited at 280 nm, and spectra were recorded at 0.2-nm intervals from 300 to 420 nm. Urea denaturation was monitored by changes in the center of spectral mass (_p), which is defined as (F_i·_i)/F_i, where F_i is fluorescence intensity at wavelength  and _i is 1/ (26). For these assays, stocks of protein (2.4 µM) in 0 and 8 M urea (or 9.5 M for the mutant) were mixed to achieve the proper urea concentration and equilibrated for at least 1 min. The equilibrium constant for unfolding was calculated using the equation: K_u = f_u/(1  f_u), where f_u is the fraction of unfolded protein. G_u^H₂O and m were calculated from K_u using the following equations: G_u = RTlnK_u and G_u = G_u^H₂O + m·[urea] (27). The solution molecular weight was determined in 10 mM Tris·Cl (pH 7.5), 0.2 M KCl ± 10 mM MgCl₂ at 25 °C using a Beckman XL-A analytical ultracentrifuge as described (28, 29).

-Galactosidase Assays

Overnight cultures of strain CSH26Q/FlacI^Qkan/pLPQ2/pNL3 or an otherwise isogenic strain containing the PhoQ acidic-cluster mutant were grown in N medium (30) plus ampicillin (50 µg/ml), spectinomycin (50 µg/ml), kanamycin (25 µg/ml), and 1 mM MgCl₂ and were diluted 100-fold in N medium plus antibiotics and either 10 µM or 10 mM MgCl₂. Cultures were grown at 37 °C until mid-log phase, and -galactosidase activity was determined as described by Miller (20). Although lacI^Q was present on an F episome, IPTG was not required for induction of phoPphoQ and concomitant activation of phoN-lacZ. Presumably, the low level of phoPphoQ expression in the absence of IPTG is sufficient for activation of phoN-lacZ.

RESULTS

Based on similarities with transmembrane sensor-transmitters of known topology, the N-terminal sensor domain of PhoQ should reside in the periplasm and consist roughly of residues 45-190 (Fig. 1B). We cloned a slightly larger fragment of E. coli PhoQ consisting of residues 43-190 into a T7 expression vector and purified the domain using a combination of anion-exchange and Affi-Gel Blue column chromatography. N-terminal sequencing and amino acid analysis of the purified protein revealed that the initiator Met, which was added for expression purposes, had been removed. In addition to the wild-type sensor domain, we constructed and purified a variant in which residues 149-155 were changed from EDDDDAE to QNNNNAQ. The six acidic residues in the wild-type cluster have been predicted to form a site for divalent-cation binding (9, 10). If this prediction is correct, we reasoned that replacing these acidic residues with isosteric but uncharged residues should alter the mutant sensor domain's ability to bind divalent cations.

The wild-type sensor domain and the acidic-cluster mutant have extremely similar CD spectra, which possess a broad band of negative ellipticity with a single minimum near 208 nm (Fig. 2A). These spectra do not have minima near 222 nm, which would be expected if the secondary structure were predominantly -helical (31), and are suggestive of a mixed -helix/-sheet secondary structure. The PhoQ sensor domain is also predicted to possess a mixture of -helix and -sheet by the program SSCP, which predicts secondary structure content based on amino acid composition (32). The fluorescence spectra of the wild-type and mutant domains are essentially indistinguishable (Fig. 3A), with _max values near 340 nm, indicating partial burial of Trp residues (33). Denaturation of either domain in 6.9 M urea red-shifts _max to 347 nm, a value expected for solvent-exposed tryptophans, and causes an increase in fluorescence intensity (Fig. 3A). Neither the CD spectra (Fig. 2A) nor the fluorescence spectra (data not shown) of the wild-type or mutant proteins were changed significantly in the presence of 10 mM MgCl₂.

In equilibrium centrifugation experiments performed at concentrations of 10 and 75 µM, both the wild-type and mutant sensor domains sedimented with apparent molecular weights close to those expected for monomers (Table I). The presence of MgCl₂ did not alter the sedimentation behavior of the wild-type domain significantly. The slight increase in M_a at the higher protein concentration may indicate weak dimerization or simply be caused by aggregation. If the increase in M_a were due to dimerization, then monomers would still represent 90% of the protein species at 75 µM, allowing a lower limit of 600 µM to be placed on the equilibrium dimerization constant.

Table I. Oligomeric state of wild-type and mutant PhoQ sensor domains Equilibrium centrifugation experiments were performed at 25 °C in 10 mM Tris·Cl (pH 7.5) plus 0.2 M KCl. Protein Concentration Maa Ma/Mrb µM Wild-type sensor domain 10 17,657  (±126) 1.03 75 20,006  (±135) 1.17 Wild-type sensor domain 10 17,232  (±74) 1.01  (10 mM MgCl2 75 18,415  (±105) 1.08 Acidic-cluster mutant 10 17,154  (±74) 1.01 75 18,280  (±112) 1.07 a  Average molecular weight (Ma) was calculated using a single species function (28, 29). b  Ma divided by Mr, the monomer molecular weight calculated from the amino acid sequence. Mr values: wild-type sensor domain = 17,088; acidic-cluster mutant sensor domain = 17,094.

In thermal denaturation studies, the wild-type sensor domain melts cooperatively with a T_m (59.2 °C; Fig. 2B) that is independent of protein concentration from 2 to 50 µM (data not shown). These data are consistent with the sensor domain having a stable, monomeric, native structure. In the presence of 10 mM MgCl₂, the T_m is increased by approximately 6.2 °C (Fig. 2B), as would be expected if Mg²⁺ binds to and stabilizes the native protein. Similar studies of the acidic-cluster mutant could not be interpreted because the protein precipitated as it began to thermally denature (data not shown). Presumably, substitution of the charged residues affects solubility of this domain at elevated temperatures.

Urea denaturation experiments, monitored by changes in the center of fluorescence spectral mass (26), were also used to compare the stabilities of the wild-type sensor domain and the acidic-cluster mutant (Fig. 3B; Table II). As shown in Fig. 3B, the mutant domain is substantially more stable to urea denaturation than the wild-type domain (G_u = 3.6 kcal/mol). Presumably, electrostatic repulsion between the six negatively charged Glu and Asp residues that are replaced in the mutant destabilizes the wild-type protein. Addition of 10 mM MgCl₂ to the urea-denaturation buffer stabilized the wild-type sensor domain by 1 kcal/mol but had no significant effect on the stability of the mutant domain (Table II). These results are most easily explained if Mg²⁺ binds to the wild-type sensor domain but not to the mutant domain. Thus, the acidic residues that are changed in the mutant appear to be important determinants of Mg²⁺ binding.

Table II. Effect of divalent salts on the stability of wild-type and mutant PhoQ sensor domains These experiments were performed in SB buffer ± 10 mM divalent salts at a protein concentration of 2.4 µM. Unfolding was monitored by changes in p, and the data were fit to a unimolecular unfolding model by nonlinear least-squares methods (25) to derive Gu(H2O). Protein Divalent salt  Gu(H2O) m kcal/mol kcal/mol · M Wild-type sensor None 5.74  (±0.10) 1.60  (±0.03)  domain Na2SO4 5.97  (±0.09) 1.61  (±0.02) MgCl2 6.79  (±0.08) 1.73  (±0.02) MgSO4 7.00  (±0.07) 1.71  (±0.03) BaCl2 7.07  (±0.10) 1.73  (±0.03) CaCl2 7.69  (±0.08) 1.78  (±0.02) MnCl2 8.20  (±0.08) 1.86  (±0.02) Acidic-cluster mutant None 9.35  (±0.19) 1.82  (±0.04) MgCl2 9.41  (±0.17) 1.81  (±0.03)

Urea denaturation in the presence of different salts was used to examine binding preferences of the wild-type sensor domain for different ions. As shown in Table II, the presence of 10 mM Na₂SO₄ has only a small stabilizing effect, which is close to the error of the measurements. Thus, the marked stabilization seen with 10 mM MgCl₂ cannot be attributed to a general increase in ionic strength or to the simple presence of a divalent ion. Similarly, MgCl₂ and MgSO₄ stabilize the sensor domain to roughly the same degree, suggesting that the nature of the anion is not important. Ba²⁺, Ca²⁺, and Mn²⁺ also stabilize the sensor domain as much as or more than Mg²⁺, with Mn²⁺ showing the largest stabilization (Table II). These data indicate a strong preference of the sensor domain for divalent cations. Moreover, the differences in stabilization caused by different divalent cations argue strongly that these ions bind directly to the sensor domain, as binding should depend on the detailed molecular properties of different divalent cations whereas a general ionic effect on stability should not. The greater stabilization of the sensor domain by Ca²⁺ than Mg²⁺ is consistent with experiments that show that lower concentrations of CaCl₂ than MgCl₂ are required for half-maximal repression of PhoP-PhoQ signaling in vivo (10).

To examine the importance of the acidic-cluster residues in sensing divalent cation concentrations in vivo, we compared the abilities of the wild-type PhoQ protein and an intact PhoQ protein containing the acidic-cluster mutations to activate PhoP-mediated transcription of a phoN-lacZ reporter gene. As shown in Fig. 4, phoN-lacZ expression is induced about 7.5-fold by MgCl₂ deprivation when the wild-type PhoQ protein is present but is only induced about 2-fold in cells expressing the PhoQ mutant. This result shows that the acidic residues in the 149-155 cluster are important in sensing divalent cations in vivo and lends support to the significance of the mutant binding defects observed in vitro. The residual transcriptional activation in response to Mg²⁺ deprivation in cells expressing the mutant PhoQ protein may reflect weak Mg²⁺ binding at the primary site for divalent cation binding, the presence of a secondary site for divalent cation binding, or an indirect effect of MgCl₂ starvation.

DISCUSSION

Despite their central role in adaptive responses to a wide variety of environmental stimuli, relatively little is known about the mechanism by which sensor-transmitter proteins detect extracellular signals and transduce information into the cell. Here, we have begun to address these questions for the PhoQ sensor-transmitter by studying the structural and ligand-binding properties of the wild-type sensor domain and a mutant, which is defective in its ability to bind and respond to ligand. The PhoQ sensor domain forms a cooperatively folded, mixed / structure that is monomeric under the conditions tested. Divalent cations (Ba²⁺, Ca²⁺, Mg²⁺, and Mn²⁺) stabilize the wild-type domain against denaturation in a fashion consistent with direct binding of these ions to the protein. The mutant domain is not stabilized by Mg²⁺ in vitro and is defective in responding to Mg²⁺ starvation in vivo.

The best characterized sensor domain is that of Tar, which forms a dimeric, four-helix bundle with binding sites for aspartate located at subunit interfaces (6). The sensor domain of PhoQ differs from that of Tar in two ways. First, the PhoQ sensor domain is monomeric. Second, the CD spectrum of the PhoQ sensor domain indicates the presence of -sheet in addition to -helix. These results suggest that PhoQ uses a different structural motif for ligand binding and signal transduction. Our finding that the PhoQ sensor domain is monomeric in the absence and presence of divalent-cation ligands might appear to rule out models of signal transduction that involve ligand-mediated dimerization (34) or those postulated to involve conformational changes within a preformed dimer (35, 36, 37). We note, however, that dimerization of the PhoQ sensor domain may still occur in the context of the intact, membrane-inserted protein, either because the transmembrane and/or cytoplasmic portions of the protein provide additional dimerization determinants or because of the higher effective concentration of sensor domains in the membrane.

Based on differential trypsin-digestion patterns of PhoQ in spheroplasts, Véscovi et al. (10) have proposed that divalent cations bind to the sensor domain and induce a conformational change that mediates signal transduction. Our results strengthen this hypothesis by showing directly that divalent cations bind to and stabilize the PhoQ sensor domain. In addition, we have shown that a cluster of six acidic amino acids is required for strong binding of divalent cations to the sensor domain. Conservative substitution of these acidic residues by non-charged amino acids results in a protein that is unable to bind divalent cations in vitro and is impaired in its ability to respond to Mg²⁺ deprivation in vivo. Are the acidic-cluster residues needed directly to bind divalent cations, or are these residues needed only indirectly to maintain the native structure of the sensor domain? Our results support the former model since the biophysical properties of the acidic-cluster mutant indicate that it folds stably with a structure that is indistinguishable from wild-type by probes such as CD and fluorescence.

Stabilizing the sensor domain by the binding of divalent cations leads to a state that is inactive in signaling. Similarly, stabilizing the mutant domain by neutralizing negative charges that are close in space also reduces signaling. As the acidic-cluster residues seem to form part of the binding site for divalent cations, the simplest model is that the mutant adopts a conformation similar to that of the Mg²⁺-bound conformation of the wild-type sensor domain. The ``active'' and ``inactive'' conformations are unlikely to differ significantly in secondary structure, since the CD spectra of the wild-type and mutant domains are nearly identical with or without Mg²⁺. This may be reminiscent of ligand binding by Tar, where only subtle structural changes occur upon binding of aspartate (6). Although the nature of the PhoQ conformational change is difficult to guess at present, we have grown crystals of the wild-type sensor domain that diffract to 2.7 Å. Solving the crystal structure of this protein should allow more detailed models to be proposed and tested.