High-Affinity Chemotaxis to Histamine Mediated by the TlpQ Chemoreceptor of the Human Pathogen Pseudomonas aeruginosa

Genome analyses indicate that many bacteria possess an elevated number of chemoreceptors, suggesting that these species are able to perform chemotaxis to a wide variety of compounds. The scientific community is now only beginning to explore this diversity and to elucidate the corresponding physiological relevance. The discovery of histamine chemotaxis in the human pathogen Pseudomonas aeruginosa provides insight into tactic movements that occur within the host. Since histamine is released in response to bacterial pathogens, histamine chemotaxis may permit bacterial migration and accumulation at infection sites, potentially modulating, in turn, quorum-sensing-mediated processes and the expression of virulence genes. As a consequence, the modulation of histamine chemotaxis by signal analogues may result in alterations of the bacterial virulence. As the first report of bacterial histamine chemotaxis, this study lays the foundation for the exploration of the physiological relevance of histamine chemotaxis and its role in pathogenicity.

not alter histamine production in mast cells, which are the classical histamine reservoirs (36). Furthermore, it has been shown that histamine might play divergent roles in the immune response: it has been implicated in mediating the defense against infection (37) as well as increasing the susceptibility to infection (38). While there has been preliminary evidence that histamine is a signal molecule for bacteria, the underlying mechanisms remain largely unknown (39). The present study provides important insight into the molecular mechanisms that permit bacteria to sense and respond to histamine.

RESULTS
Identification of histamine and additional polyamines as novel ligands for the P. putida KT2440 McpU chemoreceptor. By screening 190 compounds for binding to the purified McpU-LBD, we previously found that McpU binds to and mediates chemotaxis to putrescine, cadaverine, and spermidine (20). In the present study, we extended this screening to include 285 additional compounds. These compounds were mostly bacterial nitrogen, phosphorous, and sulfur sources (see Materials and Methods). We used a thermal shift assay to monitor changes in the midpoint of protein unfolding (T m ) caused by ligand binding (40). In the absence of ligand, McpU-LBD had a T m of 46.5°C. Of the 95 nitrogen sources screened (Biolog plate PM3B), three additional compounds-agmatine, ethylenediamine, and histamine-caused T m increases greater than 2°C (Fig. 1).
Using isothermal titration calorimetry (ITC), we found that all three compounds bind to McpU-LBD (see Fig. S1A in the supplemental material). Very tight binding was observed for agmatine with a K D (equilibrium dissociation constant) in the nanomolar range, whereas histamine and ethylenediamine bound with much lower affinities (Table 1). It should be noted that of these three new McpU ligands and the previously identified ligands (i.e., putrescine, cadaverine, and spermidine), all except for histamine are polyamines (Fig. S1B).
Identification of TlpQ as a histamine receptor in Pseudomonas aeruginosa. Because histamine plays an important role in the immune response, we aimed to identify McpU homologues in P. aeruginosa that may also sense and mediate chemotaxis to histamine. To this end, we carried out a sequence clustering analysis of all dCACHE-containing chemoreceptors in PAO1 and KT2440 (see Fig. S2A). This analysis revealed that the LBD of the TlpQ receptor shares 62% sequence identity with the McpU-LBD homologue (Fig. S2B). To verify TlpQ function, we purified TlpQ-LBD for ITC binding studies. The results showed that five McpU-LBD ligands bind to TlpQ-LBD with nanomolar affinities, whereas the binding of ethylenediamine was slightly weaker ( Fig. 2A, Table 1). Spermidine had a K D of 56 nM, which is the highest ligand affinity ever observed for a chemoreceptor. Histamine had a K D of 639 nM, which is an affinity 41 times higher than its affinity for McpU-LBD (Table 1, Fig. 2B). Thus, the affinities of the ligands to TlpQ-LBD were 3 to 147 times higher than their affinities to the McpU-LBD (Table 1). Previous studies showed that TlpQ mediates chemotaxis to ethylene (31), but the titration of TlpQ-LBD with a saturated ethylene solution did not show any binding (data not shown).
Characterization of histamine chemotaxis. KT2440 and PAO1 both contain chemoreceptors that bind histamine. In initial experiments, we identified the optimal culture conditions for motility of both strains (see Fig. S3). Using these conditions, we carried out capillary chemotaxis assays of PAO1 toward the six TlpQ ligands (Fig. 2C). All ligands caused chemotaxis, with significant responses observed for some ligands at concentrations as low as 500 nM, whereas optimal responses occurred at 5 mM. In subsequent experiments, we compared the histamine dose response for KT2440 with that of PAO1 (Fig. 3A). KT2440 showed only moderate chemotaxis over the entire concentration range tested, whereas PAO1 responses were much stronger. In accordance with the different binding affinities observed by ITC, the response onsets between strains also differed. Thus, PAO1 required 500 nM histamine, while KT2440 required 5 M.   Upper graphs show raw titration data, while lower graphs show integrated corrected peak areas of the titration data fit using the "one binding site model." The derived thermodynamic parameters are provided in Table 1. (C) Quantitative capillary chemotaxis assays of P. aeruginosa PAO1 toward TlpQ ligands. Shown are the ratios of cells after 2 min of exposure to the chemoattractant relative to the number of cells at the beginning of the experiment. The horizontal line marks the ratio of 1, which is indicative of no chemotaxis; n ϭ 3.
To assess the metabolic value of these ligands, we conducted growth experiments with PAO1 and KT2440 in minimal medium containing each of the ligands as the sole carbon or nitrogen source. We found that most of the ligands supported growth either as the carbon or nitrogen source (see Fig. S4). The exceptions were spermidine and ethylenediamine that were either not or were poor growth substrates for PAO1 and KT2440 (Fig. S4). Histamine permitted the growth of both strains as the sole C and N source.
Additional experiments were conducted to assess histamine chemotaxis in other bacteria. First, we assessed the motility of P. aeruginosa strains 227, 233, 287, 401, and 428, which were isolated from patients with urinary tract infections (41). Strains 233 and 401 exhibited motilities comparable to that of PAO1 and were therefore selected for further studies. P. aeruginosa PA14 as well as the plant pathogen Ralstonia pseudosolanacearum Ps29 were also included in these experiments. We found that all analyzed P. aeruginosa strains showed significant chemotaxis to 5 mM histamine, and their chemotactic phenotype was significantly higher than that of KT2440. On the other hand, the strain Ps29 was not attracted to histamine (Fig. 3B). Growth experiments with Ps29 in minimal medium containing histamine as the sole carbon and nitrogen source revealed no significant growth (Fig. S4B), suggesting a link between chemotaxis and the capacity to use histamine for growth.
Three-dimensional structure of TlpQ-LBD in complex with histamine. To determine the molecular determinants for histamine recognition by TlpQ, we solved the Bacterial Histamine Chemotaxis ® high-resolution structure of TlpQ-LBD in complex with histamine. There are four monomers in the asymmetric unit,and the superimposition of their C ␣ atoms resulted in root mean square deviation (RMSD) values of 0.4 to 0.8 Å, indicative of high similarity. An inspection of the structure revealed that it is a dCACHE domain (12) (Fig. 4). A long N-terminal helix is followed by two globular ␣/␤ modules, termed membrane-proximal and membrane-distal modules. The membrane-distal module contained bound histamine in all four monomers of the asymmetric unit (Fig. 4).
Structural alignments of TlpQ-LBD with entries in the protein data bank identified structural homologues (see Table S1). Most of the homologues are categorized as dCACHE_1 Pfam domains (12). This domain is found in histidine kinases and chemoreceptors, as well as in a novel cytosolic receptor protein (PDB identifier [ID] 5ere), and are found in different bacteria as well as in Arabidopsis. The average size of the domains, while taking into account the segment in between both transmembrane regions, is 268 Ϯ 17 amino acids (Table S1). Of all the homologous domains that we identified, TlpQ-LBD was the largest, at 334 amino acids, namely due to particularly long inserts between ␤-strands 1 and 2, extended helices ␣1 and ␣2, and an extended loop between helices 3 and ␣3 (see Fig. S5).
A well-defined electron density for histamine was observed in all four monomers, enabling the ligand placement to be determined (Fig. 5A). TlpQ ligands are present as protonated polycations at neutral pH, which explains why the ligand binding pocket is highly negatively charged (Fig. 5B). All three histamine nitrogen atoms establish hydrogen bonds (Fig. 5C). TlpQ ligands contain at least one primary amino group, and the primary amino group of histamine plays a central role in binding because it forms hydrogen bonds with the side chains of Tyr208, Asp210, and Asp239. In addition, this histamine amino group interacts with a water molecule coordinated by the main chain oxygen of Lys211 and the hydroxyl group of Tyr158. Each of the histamine imidazole nitrogen atoms forms hydrogen bonds with Asp210 and Glu170. The LBDs of TlpQ and McpU of P. putida KT2440 share approximately 50% sequence identity (Fig. S2B). When their structures containing either histamine or putrescine were superimposed (Fig. 5D), it became apparent that the primary amino groups of both ligands are coordinated in a similar manner via hydrogen bonds with Y208/D210/D239 of TlpQ-LBD or their equivalents in McpU-LBD.
Histamine chemotaxis is mediated by multiple chemoreceptors in PAO1. To assess the role of TlpQ in histamine chemotaxis, we generated a tlpQ mutant. Control experiments showed that its response to Casamino Acids was comparable to that of the wild type. However, the response of this mutant to 5 mM histamine was also similar to that of the wild type ( Fig. 6A), indicating that additional chemoreceptors may be involved.
To identify these additional chemoreceptors, we screened a number of mutants, in which 3 to 7 chemoreceptor genes had been deleted. Our results showed that the deletion of the pctA, pctB, and pctC chemoreceptor genes (strain PCT2) abolished chemotaxis to 5 mM histamine (Fig. 6A). PctA, PctB, and PctC are chemoreceptors for L-amino acids (23,24), while PctC also mediates chemotaxis toward GABA (21).
To clarify the roles of PctA, PctB, PctC, and TlpQ in histamine chemotaxis, we conducted dose-response experiments using wild-type PCT2 as well as a mutant in which the pctABC as well as the tlpQ gene had been deleted, named PCT2Q (Fig. 6B). The latter mutant was devoid of histamine chemotaxis over the entire concentration range (50 nM to 50 mM), whereas significant chemotaxis was observed for the PCT2 mutant at a concentration range between 500 nM and 500 M. This indicates that TlpQ mediates chemotaxis to low histamine concentrations, which is in agreement with the very high affinity observed in vitro. In contrast, one or more of the PctA, PctB, and PctC receptors mediate chemotaxis to elevated histamine concentrations.
To assess the role of the individual chemoreceptors, the PCT2Q mutant devoid of histamine chemotaxis was complemented with plasmids containing one of the four chemoreceptors-an approach that has previously proven effective to study complex chemotactic processes (42). To confirm the phenotypes of these strains, chemotaxis was measured toward previously identified ligands, namely L-Ile (PctA), L-Arg (PctB), and GABA (PctC) (23,24), and the three complemented strains responded to these ligands. Histamine chemotaxis measurements revealed that the pctC and tlpQ genes in trans recovered histamine chemotaxis using an exposure time of 1 min. At 2.5 min, complementation with pctA, pctC, and tlpQ resulted in significant chemotaxis (Fig. 6C). Thus, these data reveal that histamine chemotaxis is mediated by the concerted action of PctA, PctC, and TlpQ. To assess the role of these receptors in another strain, we generated a triple mutant in the homologous receptors of P. aeruginosa PA14. As shown in Fig. S6, the deletion of these receptors also abolished histamine chemotaxis.
TlpQ, PctA, and PctC employ different mechanisms to mediate histamine chemotaxis. To determine the mechanism by which PctA and PctC respond to histamine, microcalorimetric binding studies with purified PctA-LBD and PctC-LBD were conducted. Whereas the proteins bound L-Ala and L-Gln (24), respectively, histamine did not bind. Direct microcalorimetric titrations can only provide information on highaffinity binding events because of the limitations presented by ligand dilution heats. To assess the possibility of low-affinity histamine binding, we conducted a competition assay. PctA-LBD was titrated with L-Ala in the presence and absence of 20 mM histamine. However, the resulting titration curves were almost identical (see Fig. S7), confirming that histamine does not bind directly to PctA-LBD.
To assess whether PctA and PctC may be activated by histamine-containing periplasmic binding proteins, pulldown experiments with immobilized PctA-LBD and PctC-LBD as well as PAO1 protein extracts were conducted using previously verified protocols (25). However, our results provided no evidence for binding partners to either domain.

DISCUSSION
The elevated numbers of chemoreceptors in many bacteria suggest that this abundance confers chemotactic capabilities to many different stimuli, and the scientific community is only beginning to explore the diversity of these responses. In general, chemoeffectors can be classified into three groups according to their physiological role. First, the majority of chemoattractants are important nutritional sources, as evidenced by numerous receptors that respond to different organic or amino acids (11). Second, chemoattraction has been observed for signal molecules such as plant hormones (31,43), neurotransmitters (44), and quorum sensing signals (45), which inform bacteria about their environment. Lastly, chemoreceptors can signal the presence of compounds, such as histamine, that are involved in multiple functions.
Thus, what is the physiological relevance of chemotaxis toward histamine? One possibility is certainly that, like most of the other McpU/TlpQ ligands, histamine supports growth as the sole C and N source. However, chemotaxis to host signals has been shown for many different pathogens to be essential for efficient infection and virulence (46). Importantly, P. aeruginosa PAO1 was shown to greatly increase neutrophil histamine content and secretion in mouse models (36), and chemotaxis to this host-derived signal will result in an accumulation of bacterial cells at the infection site. This increase in bacterial cell density likely alters the expression of quorum-sensingcontrolled genes, including those responsible for the production of virulence determinants and biofilm formation in P. aeruginosa (47). Nonetheless, the precise assessment of the role of histamine chemotaxis in the virulence of P. aeruginosa is technically a difficult undertaking, since it is unfeasible to generate a mutant that is deficient in histamine chemotaxis without impairing taxis to the remaining identified ligands for PctA (17 amino acids), PctC (GABA and 2 amino acids), and TlpQ (5 polyamines) (24).
The interference with motility and chemotaxis is an alternative strategy to block bacterial pathogens (48). Previous work has shown that some chemoreceptors recognize chemoattractants and antagonists (27,49,50), and the identification of antagonists that specifically interfere with histamine chemotaxis may thus be an alternative approach to modulate the virulence properties of P. aeruginosa. Remarkably, the identification of these antagonists may be facilitated by the resolution of the threedimensional structure of TlpQ-LBD in complex with histamine ( Fig. 4 and 5).
High sensitivity histamine responses are mediated by the TlpQ chemoreceptor, which binds histamine directly. TlpQ is in many aspects an atypical chemoreceptor.
First, its LBD, which is 334 amino acids, is larger than any other known chemoreceptor LBD (11). Second, it has the highest affinity ever observed for the binding of a chemoattractant to the recombinant LBD of a chemoreceptor. Histamine binding occurred with an affinity of 639 nM, which is among the highest affinities observed for chemoattractants. This unusually high affinity permits responses to very low histamine concentrations, and the onset of chemotactic response occurred at the unusually low concentration of 500 nM (Fig. 2C).
Frequently, the deletion of a chemoreceptor abolishes taxis to a given compound, indicating that there is a single receptor for a given chemoattractant (18,29). However, histamine chemotaxis is mediated by the concerted action of three receptors. Thus, what is the advantage of having multiple receptors for the same chemoattractant? In this context, close similarities exist between histamine chemotaxis and the mechanisms by which P. aeruginosa is attracted to inorganic phosphate (P i ). P i is a key signaling molecule that controls the expression of many virulence genes (51,52). Chemotaxis to a low P i concentration is mediated by the CtpL receptor, whereas CtpH is responsible for responses to high concentrations (25,26). Whereas CtpH recognizes P i directly at its LBD, CtpL is stimulated by the P i -loaded periplasmic binding protein PstS (25). A chemoreceptor, either stimulated by direct or indirect signal recognition, is characterized by a response range (53). The combined action of multiple chemoreceptors with different sensing abilities permits the microorganism to expand its response range to a given chemoattractant. The presence of multiple receptors for a given chemoeffector may suggest that a compound is particularly physiologically relevant.
Four human histamine receptors have been described, termed H1, H2, H3, and H4 (54). Histamine was shown to mediate the chemotaxis of mast cells via the H4 receptor, and this mechanism might be responsible for mast cell accumulation in allergic tissues (55). However, the topology of eukaryotic histamine receptors differs entirely from that of their bacterial counterparts. All four receptor types form a barrel composed of seven transmembrane helices (54). The three-dimensional structure of the human H1 receptor has been solved (56), which revealed that ligands bind within this transmembrane barrel. Therefore, the evolutionary strategies to sense histamine greatly differ between bacteria and humans.
Here, we provide the first report of bacterial chemotaxis toward histamine. Histamine is vital to cellular processes in mammals, and the initial evidence suggests that histamine also functions as a bacterial signal molecule. This study expands the range of known bacterial chemoeffectors and lays the foundation for deciphering the molecular mechanisms underlying histamine chemotaxis and its role in bacterial virulence.

MATERIALS AND METHODS
Bacterial strains and plasmids. The bacterial strains and plasmids used are listed in Table 2, and oligonucleotides are listed in Table S2 in the supplemental material.
Construction of bacterial mutant strains and plasmids. The pctA and tlpQ genes were deleted in different mutant strains by unmarked gene deletion. Plasmids pK18mobsacB-pctA and pK18mobsacB-tlpQ were generated by amplifying 0.6-to 1.2-kb regions up-and downstream of the target gene. PCR products were digested with the restriction enzymes listed in Table S2 and cloned into pK18mobsacB. The resulting plasmids were introduced into E. coli S17-1 pir by electroporation. Plasmids were transferred to PAO1 by conjugation, and cells were selected in Simmons citrate (BBL; Becton, Dickinson) agar plates supplemented with kanamycin. For plasmid excision, LB medium was inoculated with a kanamycin-resistant colony and grown for 12 h, and then spread on LB plates containing 20% (wt/vol) sucrose. To construct the triple deletion mutant in pctA, pctC, and tlpQ in P. aeruginosa PA14, pK18mobsacB-pctA and pK18mobsacB-tlpQ were consecutively conjugated into PA14 to generate the double mutant. Subsequently, the plasmid pK18mobsacB-pctC, generated by amplifying regions up-and downstream of pctC, was conjugated into PA14 ΔpctA ΔtlpQ. Kanamycin-resistant colonies were grown on LB agar plates supplemented with 20% (wt/vol) sucrose for selection. For complementation purposes, the pctC gene was amplified by PCR and cloned into plasmid pUCP18 using the restriction enzymes listed in Table S2. The ligation mixture was electroporated into E. coli JM109, and transformants were selected on carbenicillin-containing LB plates. The resulting plasmid pPctC was transferred to PCTC1 and PCT2Q by electroporation.
Construction of the TlpQ-LBD expression plasmid. The DNA fragment encoding the LBD of TlpQ was amplified, digested with NdeI and BamHI, and cloned into pET28b(ϩ) linearized with the same enzymes.  The pa4308 gene (orf-1), which forms part of the pctABC operon, encodes a hypothetical protein that is not involved in chemotaxis (23). medium, and washed twice with M9 medium. The pellet was resuspended in 1 ml M9 medium. Ten-microliter aliquots of 5 mM chemoattractant solutions were placed onto plates containing M9 medium, 2.5 mM glucose, and 0.25% (wt/vol) agar. Two-microliter aliquots of bacterial suspensions were placed horizontally to each of the chemoattractant spots. The plates were incubated at 30°C for 16 to 20 h.
(ii) Quantitative capillary chemotaxis assays. Two protocols were used that differed in the ways the cells were counted. The first protocol was used to generate the data shown Fig. S6, whereas the second protocol was used for the remaining chemotaxis experiments. In the first protocol, overnight cultures of strains were diluted to an OD 660 of 0.05 in MS medium (58) supplemented with 6 mg·liter Ϫ1 Fe citrate, trace elements, and 15 mM glucose and grown at 37°C. At an OD 660 of 0.4, the cultures were centrifuged at 1,700 ϫ g for 5 min and the pellet was washed twice with chemotaxis buffer (50 mM potassium phosphate, 20 mM EDTA, 0.05% [vol/vol] glycerol, pH 7.0). The cells were resuspended in this buffer and adjusted to an OD 660 of 0.1, and 230-l aliquots were placed into 96-well plates. Capillary tubes (P1424, Microcaps; Drummond Scientific) were heat sealed at one end and filled with chemotaxis buffer or chemotaxis chemoattractant solution. The capillaries were then immersed in bacterial suspensions at their open ends. After 30 min at room temperature, the capillaries were removed and rinsed with sterile water, and the content was expelled into 1 ml of M9 medium. Serial dilutions were plated on LB medium, and the CFU were determined. In all cases, the data were corrected to the number of cells that swam into the buffer-containing capillaries. In the second protocol, we used computer-assisted image analysis as reported previously (59). Briefly, capillaries were filled with chemoattractant solutions in 10 mM HEPES buffer (pH 7.0) containing 1% (wt/vol) agarose and heat sealed on one side. The cells were grown in 2ϫ yeast extract-tryptone (YT), medium and 10-l aliquots of the cell suspension were placed onto a microscope slide within the U-shaped spacer, which was then covered by a coverslip. The chemoattractant-filled capillaries were introduced into the chemotaxis chamber, and cell movement was videotaped, with images taken at the beginning and at different time intervals. If not otherwise stated, the contact time between the cell and the chemoattractant was 2 min. The Bioinformatics Assistant Icy Sport detector software (60) was used to determine the number of cells per image. The magnitude of chemotaxis was expressed as the number of cells after a given time over the number of cells at the beginning of the experiment. The data shown are the means and standard deviations from three experiments conducted in triplicates.
Growth experiments. PAO1 was grown overnight in MS minimal medium (29) containing 20 mM D-glucose. Cultures were diluted to an OD 600 of 0.02 in MS medium supplemented with 5 mM carbon or nitrogen source. The assays were performed in 100-well polystyrene plates and incubated at 30°C (KT2440) or 37°C (PAO1) in a Bioscreen microbiological growth analyzer. The data represent the means and standard deviations from three biological replicates conducted in triplicates.
Crystallization and structure resolution. Crystallization trials were carried out with TlpQ-LBD in the absence and presence of histamine. TlpQ-LBD in polybuffer was incubated with a 2-fold molar excess of histamine on ice for 30 min. Unbound ligand was removed by buffer exchange using 10-kDa-cutoff filters (Amicon) and polybuffer. The apo-(6 mg/ml) and ligand-bound (26 mg/ml) proteins were loaded into 0.3-mm-diameter capillaries for counter diffusion crystallization using screen kits from Triana S & T (Granada, Spain). Only the LBD-TlpQ-histamine complex produced crystals in 1.5 M ammonium phosphate and 0.1 M sodium citrate pH 5.6. The same protocol was used to crystallize the Se-Met TlpQ-LBD. The capillaries were emptied into mother solution containing 10% to 25% (vol/vol) glycerol as the cryoprotectant. Crystals were diffracted at the European Synchrotron Radiation Facility and the Spanish Synchrotron ALBA. The data were indexed and integrated with XDS (61) and scaled with SCALA (62). All attempts to obtain a molecular replacement solution failed. Phases were obtained from the Semethionine derivative by combining SAD data, at the selenium peak, with an initial model generated by Phyre2 (63), as input files for Auto-Rickshaw (64). All 26 expected heavy atom positions were identified by SHELXD (65) using the data to 3.5 Å. The model generated was refined with phenix.refine (66). Further refinement was performed against the best data set (2.45 Å) with phenix.refine (66) using Coot (67). Model quality was checked using MolProbity (68). The refinement statistics and quality indicators of the final model are summarized in Table S3. The structure was deposited at the protein data bank with identifier (ID) 6fu4.

ACKNOWLEDGMENTS
We acknowledge the European Synchrotron Radiation Facility (beamlines ID23-1 & 2, ID30B, and ID30A-3) and the Spanish Synchrotron ALBA (beamline XALOC) and thank the staff for their invaluable support.
This work was supported by FEDER funds and Fondo Social Europeo through grants held by T. Krell from the Spanish Ministry for Economy and Competitiveness (grants BIO2013-42297 and BIO2016-76779-P) and by grant BIO2016-74875-P to J. A. Gavira.