Curcumin Reduces the Motility of Salmonella enterica Serovar Typhimurium by Binding to the Flagella, Thereby Leading to Flagellar Fragility and Shedding

S. Typhimurium strain 12023 with pFPV25.1 and strain 14028 (a kind gift from M. Hensel, Germany) were grown at 37°C in lysogeny broth (LB; HiMedia). The medium was supplemented with 50 μg/ml ampicillin for growing S. Typhimurium 12023 with pFPV25.1. lacZ transcriptional fusions to flhD, fliL, and fliC or fljB in S. Typhimurium, a kind gift from Kelly Hughes, University of Utah, were grown in LB containing 50 μg/ml kanamycin. Escherichia coli BL21 was grown in LB at 37°C. Solvent (DMSO) for dissolving all the phenolic compounds or curcumin (catalog no. C7727; Sigma) was added to LB wherever mentioned. The phenolic compounds ferulic acid, piperine, and vanillin (Sigma-Aldrich) were added to the medium whenever required.

The DNA sequence that codes for flagellin protein (fliC gene) from S. Typhimurium 14028 was amplified and cloned into pTrc99A. The cloned fliC gene was sequenced to confirm complete identity with the published sequence in the database. The clone was transformed into the S. Typhimurium ΔfliC mutant to obtain S. Typhimurium ΔfliC::pTrcfliC. S. Typhimurium ΔfliC::pTrcfliC was induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at 30°C, and the whole-cell lysate was probed with anti-FliC antibody (Difco) to confirm the expression of the flagellin protein (see Fig. S1 in the supplemental material). All the mutations were introduced into pTrc99A fliC using standard PCR mutagenesis techniques, and mutations were confirmed by sequencing. The mutants generated were the following: (i) pTrc99A fliC N120A D123A, (ii) pTrc99A fliC N163A S164A, (iii) pTrc99A fliC N173A Q175A, and (iv) pTrc99A fliC triple mutant (all three mutations were incorporated). For generation of the triple mutant, individual mutations were introduced sequentially.

A swimming and swarming assay was performed as described by Kim and Surette (17). Briefly, 2 μl of an overnight grown bacterial culture was spot inoculated at the center of the swim or swarm plates. The plates were dried for 50 min and incubated at 37°C for 6 to 8 h. The plates were imaged, and the diameter of the bacterial growth front was measured using the ImageJ software. The rate of bacterial swimming or swarming was calculated as the growth of a colony per unit of time.

The media used for swimming and swarming assay were (i) swimming medium (LB plus 0.25% agar [bacteriological grade; HiMedia] plus 0.5% glucose), and (ii) swarming medium (LB plus 0.5% agar [bacteriological grade; HiMedia] plus 0.5% glucose).

Curcumin (20 μM), DMSO (volume equivalent to that used for 20 μM curcumin), or the indicated phenolic compounds at different concentrations were added along with the medium (after cooling to 60°C), wherever mentioned.

The bacteria grown in the presence of curcumin (20 μM) or DMSO were fixed with 2.5% glutaraldehyde and mounted on silane-coated coverslips. The fixed cells were incubated with antiflagellin antibody (1:100 in PBS; Difco) for 1 h. The coverslips were washed twice with PBS, and the cells were incubated with Cy2-conjugated anti-rabbit antibody (1:100 in 2% bovine serum albumin [BSA]; Dianova). The cells were observed under a Carl Zeiss meta 710 microscope, and the images were captured.

S. Typhimurium cells grown in LB, LB plus 20 μM curcumin, or LB plus DMSO were fixed in 2.5% glutaraldehyde at room temperature for 30 min. One hundred microliters of fixed samples was put onto Parafilm. Formvar-coated nickel grids (3 mm, 200 mesh) were placed on top of the sample drop and left for 1 to 2 min. The grid surface was washed twice with Tris-EDTA (TE) buffer (pH 6.9), blotted dry, and rinsed in distilled water. The bacteria were negatively stained with 2% uranyl acetate solution. The grid surface was washed with TE buffer, air dried, and imaged using a transmission electron microscope (JEM-100 CX II; JEOL).

The flagellar density of S. Typhimurium was determined as described previously (18).

S. Typhimurium grown in the presence or absence of curcumin (20 μM) was pelleted at 10,000 × g for 5 min and incubated with rabbit anti-S. Typhimurium flagellin antibody (1:100 in 2% BSA; Difco) for 1 h. Bacterial cells were washed with PBS containing 1% BSA by centrifugation at maximum speed, followed by resuspension of the bacterial pellet in PBS. The samples were then incubated with horseradish peroxidase-conjugated anti-rabbit antibody (1:2,000; Bangalore Genie) for 1 h. The samples were again washed with PBS, and the peroxidase activity was assessed using 3,3′,5,5′-tetramethylbenzidine as a substrate. The absorbance of the solution was recorded at 450 nm in an enzyme-linked immunosorbent assay (ELISA) microplate reader (ELx 800MS).

Flagellin protein was isolated using a modified protocol from that of Ibrahim et al. (19). Briefly, bacteria were grown in 5 ml of LB at 37°C for 12 h. The bacteria were subcultured at a 1:100 ratio in LB plus 0.5% glucose and incubated at 37°C and 80 rpm for 16 h. The bacteria were harvested by pelleting at 6,000 × g for 15 min at 4°C. The pellet obtained from a 1.5-liter culture was resuspended in 5 ml of PBS. The pH of the solution was adjusted to 2 with 1 N HCl. The sample was maintained under shaking conditions for 30 min and then centrifuged at 10,000 × g for 15 min. The supernatant thus obtained was spun at 100,000 × g for 1 h to pellet any insoluble debris. The pH of the recovered supernatant was adjusted to 7 with 1 N NaOH, and 2.6 M ammonium sulfate was added with vigorous shaking. After overnight incubation at 4°C, the mixture was spun at 14,000 × g for 20 min at 4°C. The pellet thus obtained was dissolved in distilled water and the mixture dialyzed against distilled water in the presence of activated charcoal. The water was changed every 1 h for the next 8 h. The solution was stored at −20°C until use. The purity of the sample was assessed by SDS-PAGE, followed by Coomassie brilliant blue staining and mass spectrometry.

In order to obtain purified mutant flagellin protein, S. Typhimurium ΔfliC was transformed with either pTrc99A fliC (wild type [WT]) or the mutant gene. S. Typhimurium ΔfliC::pTrcfliC was induced with 1 mM IPTG at 30°C. Cells were pelleted 6 h postinduction, and flagellin protein was purified as described above.

S. Typhimurium was grown in the presence or absence of curcumin for 6 h. The culture was vortexed at 5,000 rpm for 5 min. The cells were pelleted to obtain the culture supernatant. The total protein from the culture supernatant was precipitated using 2.6 M ammonium sulfate. The protein thus obtained was dialyzed using a 10-kDa-cutoff dialysis membrane. The protein solution was heated at 90°C for 10 to 15 min to monomerize the flagellin polymer. The monomerized flagellin was used to treat HEK293 cells.

A total of 5 × 10⁴ HEK293 cells were seeded into a 24-well plate. The medium in each well was replaced with 700 μl of transfection medium containing Opti-MEM (Gibco), polyethylenimine, and 250 ng of each plasmid, pNF-κB-TA-Luc and pCMVβ-gal (both plasmids were a kind gift from Takashi Yokochi, Aichi Medical University, Japan). The final concentration of each plasmid was 50 ng/l0⁴ cells. The cells were incubated at 37°C in humidified air (with 5% CO₂) for 6 h. The transfection medium was replaced with complete medium (RPMI with 10% serum), and the cells were further incubated for 36 h. The transfected cells were induced with a concentration of 250 ng/ml of total protein supernatant (heated at 90°C for 10 min) obtained from S. Typhimurium grown in the presence of either curcumin (20 μM) or DMSO for 6 h. The luciferase assay was performed 12 h postinduction, in accordance with the manufacturer's instruction (Promega). A β-galactosidase assay was performed by measuring the hydrolysis of the chromogenic substrate O-nitrophenyl-β-d-galactoside to normalize for the transfection efficiency.

S. Typhimurium culture was grown in the presence of curcumin, phenolic compounds, or DMSO for 6 h. Bacterial cultures were either first vortexed for 3 min and then pelleted or directly pelleted at 10,000 × g for 5 min at 4°C. The supernatants from curcumin-treated and untreated samples were processed in one of the following ways and processed for immunoblot analysis: (i) concentrated under a vacuum; (ii) filtered using a 0.2-μm-pore-size filter and concentrated under a vacuum; (iii) acidified to pH 2 and concentrated under a vacuum; or (iv) acidified to pH 2, filtered using a 0.2-μm-pore-size filter, and concentrated under a vacuum.

The whole-bacterium culture was lysed using a lysis solution (0.01 M Tris-HCl, 1 mM EDTA, 8 M urea, 0.05 M dithiothreitol [DTT], 10% [vol/vol] glycerol, 5% [vol/vol] NP-40). Concentrated culture supernatant (400 μg) and the lysed whole-bacterium culture (50 μg) were probed with antiflagellin antibody (1:1,000; Difco). A portion of the gel not used for transfer was stained using the silver staining protocol and used as the loading control.

S. Typhimurium harboring lacZ transcriptional fusions to flhD, fliL, and fliC or fljB were used for the assay. The transcriptional activity of each gene was determined by performing a β-galactosidase assay, as described previously (20).

The change in the intrinsic tyrosine fluorescence was used to estimate the binding affinity of curcumin to flagella. Curcumin (1.5 to 8 μM) or an equivalent volume of DMSO was titrated against a fixed concentration of flagellin protein (4 μM). The solution was excited at 270 nm, and the emission spectrum from 280 to 340 nm was recorded at each curcumin concentration.

Various concentrations of curcumin (0.125 to 8 μM) were titrated against a fixed concentration of flagellin (10 μM) in Tris buffer (25 mM [pH 7]) and the curcumin fluorescence estimated at 495 nm by exciting the solution at 425 nm.

The inner filter effect at a high curcumin concentration was minimized using the following formula (21): F_corrected = F_observed × antilog [(A_ex + A_em)/2], where F_corrected is the corrected fluorescence intensity, F_observed is the observed fluorescence intensity, and A_ex and A_em are the absorbance of curcumin at excitation and emission wavelengths, respectively.

A single isolated bacterial colony (S. Typhimurium or Escherichia coli) was inoculated in LB and grown at 37°C and 160 rpm for 6 to 7 h. The bacterial cultures were pelleted and resuspended in Tris-HCl buffer (25 mM [pH 7]). These cultures were then incubated in the presence of curcumin (20 μM) or DMSO and antiflagellin antibody (1:200; Difco) for 3 to 5 min. Aliquots of 20 to 50 μl were placed on a slide and observed for agglutination/formation of clumps due to cross-linking of bacteria.

The crystal structure of the flagellin monomer, F41 fragment (22), a bacterial flagellar protofilament (Protein Data Bank identifier 1IO1) was obtained at a resolution of 2 Å from RCSB (23). The obtained structure was then prepared for docking by the removal of water molecules and addition of Gasteiger partial charges for adding hydrogens to polar atoms, availed for docking in a .PDBQT format using python script available in the AutoDock Tools package (24, 25). The structure files of curcumin, ferulic acid, vanillin, piperine, and EF-24 were obtained from PubChem in SDF format and translated to the required formats using Open Babel 2.3.0 (26).

The possible ligand binding sites were predicted using PocketDepth (27), an in-house geometry-based approach that uses depth-based clustering for the identification of binding sites in proteins. Further support for the predicted sites was obtained by using alternative binding site prediction methods, Q-SiteFinder (28) and SiteHound (29). These two methods are energy-based approaches that use different types of probes to calculate the interaction energy, with SiteHound also including Lennard-Jones and electrostatics energy terms, and LIGSITEcsc (30), which predicts ligand binding sites by using the Connolly surface and degree of conservation of the involved surface residues. The degree of conservation as defined by LIGSITEcsc is the average conservation of all residues within a sphere of a certain radius (8 Å, in this study) of the center of mass of the cluster.

The protein sequences of flagellin from E. coli (GenBank accession no. P04949) and S. Typhimurium (GenBank accession no. P06179) were retrieved from the UniProtKB (31) database and subjected to a BLASTP search (32) with default BLOSUM62 matrix for pairwise alignment.

Autodock4 (AD4) (24) and Vina (33) were employed for docking (22), and the individual parameter files were prepared as required by the program. A 70- by 70- by 70-point grid, centered at the binding site with 0.375-Å and point grid spacing, was used to calculate an atomic affinity map. The conformational space around the binding site was assigned as a cubic grid of 26.6 Å, and the coordinates assigned to the grid center were the same as those used in AD4.

AD4 and Vina generated a docking log file that gives complete information about different binding modes and their corresponding binding free energies of the docked ligand. AD4 gives an additional feature of clustering, which was studied for similarity in binding modes and binding energies. A standalone version of Ligand Protein Contacts (LPC [34]) was used to analyze the interaction profile explaining different kinds of interactions (hydrogen bonds and electrostatic, van der Waals, and hydrophobic interactions) across docked complexes. PyMOL was used for visualization and image construction of the docked complexes (35).

A database search for sequence homologues of S. Typhimurium FliC was carried out in the bacterial genomes of the NR sequence database using the BLASTP server and the BLOSUM62 matrix. An E value of 0.0001 was imposed as the search criterion. BLAST hits having the highest sequence identity, query coverage area, and BLAST score and lowest E value in each species were selected for further analysis. A multiple-sequence alignment was carried out for the hits obtained, using Clustal Omega (36) with default parameters, the results of which were visualized using BoxShade version 3.31, written by K. Hofmann and M. Baron (37).

Student's t test was used for testing the significance of the differences observed, if any. All analyses were done with the GraphPad Prism (version 5) software.

Curcumin (20 μM) decreased the motility of S. Typhimurium as visualized under the light microscope. This reduction in motility was confirmed by a swim and swarm assay. There was a significant decrease in bacterial motility, as revealed by the reduction (65 to 80%) in the diameter of the swim and swarm colony of S. Typhimurium in the presence of different curcumin concentrations (Fig. 1). Curcumin (20 μM) also reduced the motility of S. Typhi strain Ty2 (see Fig. S2A in the supplemental material).

When assessed for cytotoxicity, curcumin (10 to 120 μM) showed no toxic effect on the growth of bacteria (see Fig. S2B in the supplemental material). Thus, curcumin exhibited an apparent effect on the motility of S. Typhimurium and not on its growth.

To address the defect in motility, we investigated curcumin's effect on flagellar machinery. We analyzed the density of flagellar filaments possessed by the bacterium using transmission electron microscopy (TEM) and fluorescence imaging techniques. TEM and fluorescence imaging gave a visual appreciation of the reduction in the filament number around the curcumin-treated (20 μM) bacteria (Fig. 2A). Almost 41% of the curcumin-treated bacteria were aflagellated. The remaining ∼59% (flagellated) possessed, on average, five filaments, contrasting with the 8 or 9 filaments possessed by untreated bacteria (Fig. 2A). A modified form of ELISA (see Materials and Methods) that estimates the flagellar density on the bacteria (18) was performed to validate our results. Similar results were obtained (Fig. 2B), confirming the reduction in the number of flagellar filaments upon curcumin treatment.

Careful observation of TEM and fluorescence microscopic images revealed detached flagellar filaments in the background of curcumin-treated bacteria (Fig. 2C), indicating that curcumin increased shedding of the flagellar filaments from the bacterial surface. In curcumin-treated samples, the length of the flagellar filaments either attached to or detached from the bacteria (5.37 ± 0.12 μm [mean ± standard error]) was significantly shorter than that for the untreated samples (8.01 ± 0.17 μm) (Fig. 2C). Almost 80% of the filaments in curcumin-treated samples had a length of <6 μm, whereas in curcumin-untreated samples, 80% of them had a length of >6 μm (Fig. 2C), implying that curcumin fragments the flagellar filaments.

Next, we assessed the expression of flagellar genes (temporally divided into three classes, I, II, and III [3]) on curcumin treatment using a reverse transcription-PCR (RT-PCR) and reporter (β-galactosidase) assay. The RT-PCR and reporter assay indicated that there was no effect of curcumin on the expression of the flagellar genes flhD (class I gene) and fliC and fljB (class III genes) (Fig. 3). However, there was an approximately 2-fold increase in the promoter activity of fliL, a class II gene (Fig. 3B). fliL is the first gene of the operon fliLMOPQR, which suggests that the expression of other genes, viz., fliM, fliO, etc., is also affected by curcumin.

Curcumin reduced the number of flagella per bacterium by fragmenting the flagellar filaments without altering the expression of flagellin genes fliC and fljB (Fig. 2 and 3). To explain these results, we speculated that curcumin might induce some physical changes in the flagella and weaken it, thus leading to its fragmentation. To test this, we applied an external mechanical force in the form of vigorous vortexing and checked for shedding/fragmentation of flagellar filaments. The bacterial cultures grown in the presence and absence of curcumin (20 μM) were vortexed at 5,000 rpm for 5 min to enhance the shedding of flagellar filaments. The amount of flagellin protein in the culture supernatants, bacterial pellet, and whole-cell suspension was estimated by immunoblot analysis. The culture supernatants of curcumin-treated samples had a greater amount of flagellin protein than that of the untreated samples (Fig. 4A). However, the amount of flagellin protein in the whole-cell suspension was unaltered upon curcumin treatment. These results suggest that application of external mechanical force in the presence of curcumin increases the shedding of flagellar filaments (Fig. 4A). Another strategy used was to examine the presence of flagellar fragments in culture supernatant of a curcumin-treated sample. The bacterial cultures were vortexed and the culture supernatants filtered through a 0.2-μm-pore-size filter, with which the monomers or fragments <0.2 μm in length would easily pass through the filter pores. This step would prevent the passage of fragments of >0.2 μm through the filter, unless its orientation is such that it passes through the filter surface lengthwise (filament diameter, ∼20 nm, smaller than the filter pore size). Immunoblot analysis revealed that the filtrate obtained from the curcumin-treated samples had a smaller amount of flagellin protein than that of the DMSO-treated samples (Fig. 4B), indicating the presence of flagellar fragments >0.2 μm in length. This also indicates that the flagellin monomers were found in smaller amounts in curcumin-treated samples. This might be due to the spontaneous polymerization of the monomers in the presence of flagellar fragments present in the curcumin-treated cultures (38, 39). The flagellar fragments act as a seed for polymerization, and the monomers become incorporated into a polymer (39). To further confirm the presence of flagellar fragments in curcumin-treated cultures, the culture supernatants were acidified (pH 2) to depolymerize the fragments (40) and then passed through the 0.2-μm-pore-size filter. The flagellin monomers thus formed could now pass through the filter. When the filtrates were probed with antiflagellin antibody, the samples of curcumin-treated culture showed higher flagellin content than that in the DMSO-treated control (Fig. 4B).This demonstrates that curcumin-treated bacterial cultures have more flagellar fragments that cannot pass through the filter until they are depolymerized before filtration. Thus, it is clear that curcumin increased the fragmentation and shedding of the flagella from the bacterial surface.

To account for the presence of fragmented flagellar filaments, the culture supernatant from the vortexed sample was spun at 100,000 × g for 1 h to pellet the flagellar fragments but not monomers (39) and then probed with antiflagellin antibody. Curcumin-treated samples had a greater amount of flagellin than the DMSO-treated control (Fig. 4C), reaffirming that curcumin treatment leads to the fragmentation of flagellar filaments.

If we assume that curcumin (20 μM) can mechanically modulate the fragmentation of flagellar filaments, transient treatment of bacteria with curcumin might also lead to increased shedding of the flagella. To test this hypothesis, we incubated the bacteria at log phase with curcumin for 10 to 12 min, followed by vigorous vortexing for 5 min, and the amount of flagellar fragments released was analyzed by immunoblotting. The release of fragmented filaments was evident from the greater amount of flagellin protein in the culture supernatant of curcumin-treated samples than that in the DMSO-treated control (Fig. 5A).

Flagellin molecules bind to the TLR5 receptor on the host cell surface, triggering a signal transduction cascade via activation of NF-κB. A functional assay was performed to validate the shedding of flagellar filaments from the bacteria, in which HEK293 cells (transfected with pNF-κB-TA-Luc) were treated with the protein isolated from the culture supernatants of curcumin-treated and untreated cultures. The assay was performed with HEK293 cells, as these cells do not express any TLRs except TLR5 (41, 42). HEK293 cells transfected with pNF-κB-TA-Luc were induced with 250 ng/ml total protein (heated at 90°C for 10 min) from the culture supernatants of curcumin-treated (20 μM) or untreated S. Typhimurium cells and assayed for the activation of NF-κB. The NF-κB activity (Fig. 5B) was significantly higher in the cells induced with the protein from the culture supernatant of curcumin-treated S. Typhimurium than in the DMSO-treated control. This corroborated our result that curcumin treatment increased the shedding of flagella from the bacteria.

It is believed that the FliL protein of Salmonella is required for protection of the flagellar rod against the increased torsional stress experienced by the flagella on the swarm agar (43). In our study, we observed increased fliL expression (Fig. 3B), along with the increased shedding of flagella in the presence of curcumin (Fig. 2). To test whether the increased expression of fliL could be linked to curcumin-induced flagellar fragmentation, we assessed the expression of fliL in aflagellated bacteria (Salmonella Typhimurium ΔfliC) by a reporter (β-galactosidase) assay. We observed no change in the fliL promoter activity (β-galactosidase) of the aflagellated strain upon exposure to curcumin (Fig. 6), indicating that curcumin-induced fragmentation could trigger the expression of fliL.

All the above-mentioned results hint at a physical interaction between curcumin and flagella. To assess the binding of curcumin to flagella, a fluorescence binding assay was performed that measures the change in protein fluorescence on conformational change due to ligand binding. Flagellin protein lacks tryptophan residues in its sequence; hence, the change in the intrinsic fluorescence of tyrosine was monitored at various curcumin concentrations (0.125 to 16 μM). The intrinsic tyrosine fluorescence of flagellin reduced significantly with increasing concentrations of curcumin (Fig. 7A), indicating the binding of curcumin to flagellin molecules. The dissociation constant for specific binding, K_d, which was obtained using a one-site binding hyperbola nonlinear regression analysis (Fig. 7A), was 5.51 ± 0.78 µM (mean ± standard error). We further investigated the binding of curcumin to flagellin by a fluorimetric titration assay, in which the change in the fluorescence of curcumin, if any, in the presence of flagellin protein was monitored. At neutral pH, curcumin shows very weak fluorescence. The addition of flagellin protein to the curcumin (8 μM) solution significantly increased the fluorescence of the solution (Fig. 7B). The dissociation constant obtained using one-site specific hyperbola was 4.84 ± 0.58 µM (mean ± standard error), which is similar to that obtained from the tyrosine fluorescence data. The flagellin protein isolated using the ammonium sulfate precipitation method stays in a polymeric form (19, 44). To investigate whether curcumin interacts with flagellin monomers, a fluorescence binding assay was performed with increasing concentrations of curcumin. The intrinsic fluorescence of flagellin monomers (produced by heat denaturation of flagella at 65°C) was not affected upon the addition of increasing concentrations of curcumin (see Fig. S3 in the supplemental material). These results indicate that curcumin interacts preferentially with flagellar filaments.

Curcumin has a bivalent structure. Hence, it could simultaneously bind two flagellin monomers. If these monomers happen to be on flagellar filaments of two adjacent bacterial cells, it could cross-link the filaments and cause agglutination (clumping). An agglutination assay revealed that curcumin did not cause any agglutination compared to that observed with antiflagellin antibody (Fig. 8). This indicates that curcumin can either bind directly to residues at the interface of filaments or destabilize filament formation through steric hindrance.

The binding site of curcumin on flagellin was predicted using computational analysis, as explained in Materials and Methods. The top-ranked pocket predicted by PocketDepth was located at the interface of domains D1 and D2 (see Fig. S4A in the supplemental material). A pocket at the same location was predicted to be the top pocket by all other methods as well, with residues GLN112, THR116, LEU119, ASN120, ASP123, ASN163, SER164, GLN165, LEU169, ASP170, THR171, LEU172, ASN173, GLN175, GLY365, ALA366, THR370, ALA381, LYS384, LEU396, ALA397, GLU398, ALA399, and ALA400 as consensus binding site residues. The grid center of the binding site predicted by PocketDepth, which showed a good match with results from other methods as well, was provided for docking using softwares AD4 and Vina (24, 33).

The possible binding poses of curcumin on the flagellin molecule were determined using AD4 and Vina. The top two conformers of curcumin were selected from AD4 and Vina. The binding energies of the best conformers generated by Vina and AD4 were −7.6 kcal/mol and −9.65 kcal/mol, respectively. Since the searching and scoring processes were implemented differently in the two docking tools using a different scoring function, both have been used for evaluating the binding potential of curcumin to flagellin. With differences in scoring functions, ranking of the poses, and variations in the finer conformations picked, it is useful to explore both. The two programs docked curcumin at similar locations (see Fig. S4A in the supplemental material), mainly in two conformations (a bent conformation and an elongated conformation of curcumin), albeit with slightly different relative ranks. The two poses shared similarity in their binding sites (see Fig. S4B and C in the supplemental material).

To confirm the binding of curcumin to flagellin at the residues identified using Vina and AD4, the residues predicted to form H bonds with curcumin were mutated, except A397 (see Fig. S4D in the supplemental material), and curcumin-flagellin binding was monitored using the fluorescence binding assay. Mutant proteins were generated with the mutations M1 (N120A and D123A), M2 (N163A and S164A), and M3 (N173A and Q175A) and all of the mutations (M1, M2, and M3) combined together. Mutated protein M3 showed a significant reduction in binding compared to mutant proteins M1 and M2 (Fig. 9). However, when all the residues mentioned above were mutated, there was no change in fluorescence with the addition of increasing concentrations of curcumin (0.125 to 8 μM), indicating abolition of binding of curcumin to a flagellin molecule (Fig. 9). These results suggest that curcumin binds to a flagellin molecule at positions N120, D123, N163, S164, N173, and Q175.

To understand whether the effect shown by curcumin is specific to its backbone structure and/or functional groups, we used computational analysis to test the binding of compounds, such as piperine and EF-24 (structural analogue of curcumin), and the degraded products of curcumin, namely, ferulic acid and vanillin, with the flagellin molecule. All the compounds showed an interaction with the flagellin molecule. The binding mode and the interaction profile of all compounds with flagellin are shown in Fig. S5 in the supplemental material.

The effect of these compounds on the swarm motility of S. Typhimurium was tested. We found that these compounds inhibited the swarm motility of S. Typhimurium (Fig. 10A) without any apparent effect on the growth of the bacteria (see Fig. S6A in the supplemental material). We further evaluated the alteration in the percentage of flagellated bacteria on treatment with these compounds. There was a 10 to 25% reduction in the number of flagellated bacteria on treatment of S. Typhimurium with these phenolic compounds (Fig. 10B). Immunoblot analysis of the protein from the culture supernatant and whole-cell lysate suggested that these compounds increased the shedding of flagellar filaments from S. Typhimurium and had no effect on the production of flagellin protein (see Fig. S6B in the supplemental material).

To determine whether the effect of curcumin is specific to S. Typhimurium or is a general phenomenon against flagellated bacteria, we assayed the effect of curcumin on the motility of E. coli. Curcumin (20 μM) retarded the spread of the E. coli colony by ∼50% on swarm agar (see Fig. S7A in the supplemental material). Curcumin did not show any cytotoxic effect on E. coli grown in LB (see Fig. S7B in the supplemental material). This indicates that the reduction in the spread of the swarm colony of E. coli was due to an effect of curcumin on the motility of E. coli.

E. coli and S. Typhimurium flagellins showed 55% sequence identity and 67% similarity, as judged by a sequence alignment using the BLOSUM62 substitution matrix. Residues predicted to interact with curcumin, namely, ASP123, ASP170, and THR370, were conserved, whereas others at the putative binding site, including ASN120, ASN163, SER164, ASN173, GLN175, and SER383, had conservative substitutions. Overall, conservation of the binding site between the two species lends further support to the result of the computational analysis.

In order to test whether curcumin could have a generalized effect on the flagella of other gammaproteobacteria, the conservation of the consensus binding site residues across a few species, belonging to both pathogenic and nonpathogenic categories, was assessed using multiple-sequence alignment (see Fig. S8 in the supplemental material). Some of the residues in the expected curcumin-binding site in S. Typhimurium FliC are conserved in a flagellin protein of many other bacteria as well, implying that it may bind curcumin in a similar fashion in these species. In particular, GLN112, LEU119, ASP123, SER164, and LEU169 appear to be well conserved.