The strains and plasmids used in this study are listed in Supplementary Table S1. C. jejuni strains were grown in Mueller–Hinton (MH) medium (Duan et al., 2023) and Brucella broth supplemented with 10% filter-sterilized fetal bovine serum (ExCell Bio, China) (Duan et al., 2023) for the growth and chemotaxis experiments, respectively, under microaerobic conditions (85% N₂, 10% CO₂, and 5% O₂), at 37°C. E. coli strains were grown in Luria-Bertani medium (Oxoid, United States), at 37°C, for routine culture; in Tryptone broth (1% tryptone and 0.5% NaCl), at 34°C, for the chemotaxis experiments; and in Medium A (Yuan et al., 2017), pH 7.2, at 37°C, for measuring the responses of hybrid kinases in E. coli MG1655 strains, under aerobic conditions. The selection medium contained antibiotics, including 10 μg mL⁻¹ chloramphenicol and kanamycin for both C. jejuni and E. coli, 50 μg mL⁻¹ ampicillin for E. coli, 3 μg mL⁻¹ tetracycline for C. jejuni, and appropriate concentrations of inducers (Supplementary Table S1).

Plasmids pZQ1-5 and pYFL1-8 were constructed to express the hybrid chemoreceptors and hybrid kinases, respectively. The DNA fragment encoding the LBD of each chimera was amplified from the genomic DNA of C. jejuni ATCC 33560. A fragment of the Tar or PhoQ cytoplasmic region was amplified from the genomic DNA of E. coli MG1655. Overlap PCR was performed to connect the fragments encoding Tlp11-Tar or Tlp11-PhoQ. The pKG116 plasmid was digested using NdeI and BamHI enzymes and ligated to the amplified fragments Tlp11-Tar or Tlp11-PhoQ by means of Red/ET recombination of E. coli GB05-dir (Li et al., 2023). To generate the plasmid pET28b-Tlp11-LBD, the codons of Tlp11-LBD from C. jejuni ATCC 33560 were optimized to those preferred by E. coli (Ruibiotech, China), and the optimized fragment was ligated to the NdeI- and XhoI-digested pET28b, by means of Red/ET recombination. Point mutations were introduced into the sequence using specific primers and cloned into the corresponding vectors. Plasmids expressing the LBDs from other bacterial species were generated using similar protocols. All plasmids were verified by means of sequencing.

E. coli BL21 (DE3) competent cells were transformed with plasmids encoding C. jejuni Tlp11-LBD, its mutants, or homologous LBD proteins from other bacterial species. An overnight culture of E. coli BL21 (DE3) containing the plasmid was inoculated into 100 mL LB medium containing 10 μg mL⁻¹ kanamycin and grown at 37°C, with shaking at 200 rpm. When the OD₆₀₀ reached 0.6 to 0.8, 500 μM isopropyl ß-D-1-thiogalactopyranoside was added into the culture, to induce expression, following which the cells were continually cultured overnight, at 18°C, with shaking at 110 rpm.

For protein purification, E. coli cells expressing recombinant proteins were harvested by means of centrifugation at 7,000 rpm, 4°C, for 10 min, to obtain cell pellets, which were then resuspended into Buffer A (25 mM Na₂HPO₄, 25 mM NaH₂PO₄, and 500 mM NaCl, pH 7.0) and lysed using an ultrahigh-pressure homogenizer (JN-2.5, JNBIO, China). The crushed cells in solution were centrifuged at 20,000 rpm and 4°C for 1.5 h, to remove the insoluble fraction. The soluble supernatant was applied to a 5 mL HisTrap^™ column (GE Healthcare, United States) equilibrated with Buffer A. The column was washed with different concentrations of Buffer B (25 mM Na₂HPO₄, 25 mM NaH₂PO₄, 500 mM NaCl, and 500 mM imidazole, pH 7.0) and the eluted proteins were collected using fast protein liquid chromatography (AKTA^™ Go System, Cytiva, United States). The eluted proteins were validated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The target proteins were concentrated using a 10-kDa centrifugal filter (Merck Millipore, United States), and the residual imidazole in the protein solution was removed using a desalting column (GE Healthcare) with Buffer C (25 mM Na₂HPO₄, 25 mM NaH₂PO₄, and 150 mM NaCl, pH 7.0).

The protein sequences from Campylobacter coli (NCBI accession number: WP_201459806.1, residues 32-332), Helicobacter equorum (WP_115570384.1, residues 34-332), Helicobacter himalayensis (WP_066386874.1, residues 33-320), Helicobacter mesocricetorum (WP_199770133.1, residues 31-330), Helicobacter ganmani (WP_115552006.1, residues 31-329), and Campylobacter upsaliensis (WP_257425542.1, residues 31-331) were purified to obtain Tlp11-LBD homologue proteins.

TSA measurements were performed using a quantitative Real-Time PCR System (LightCycler^® 480; Roche, United States), to monitor the T_m of the LBD proteins. The compounds listed in Supplementary Table S2 were used as ligand candidates for the high-throughput screening. Each 25 μL mixture for the standard assay contained 20 μM protein, 1–2 mM compound in Buffer C, and SYPRO^™ Orange dye (Life Technologies, United States), at 5× concentration. Samples were heated from 26°C to 85°C, with a ramp rate of 1.2°C min⁻¹, to denature the protein. Protein unfolding curves were recorded by detecting the changes in fluorescence. The T_m values were calculated using first-derivative values (dF/dT) from the raw data.

The His-tag dye (MO-L018 RED-tris-NTA, NanoTemper, Germany) was adjusted to a final concentration of 25 nM and used to label the purified protein with a His-tag. The compound was diluted to a series of concentrations in phosphate-buffered saline (Duan et al., 2023) with 0.05% Tween 20 (PBS-T) buffer and mixed in a ratio of 1:1 with protein. The mixtures were loaded into capillaries (Monolith^™ Series capillaries, NanoTemper) by means of capillary action. Thermophoresis was measured using a Monolith^™ NT.115 (NanoTemper) instrument, with 40% excitation and 40% MST power configuration. The time-traces of changes in fluorescence, which reflected the thermophoretic movement of labeled protein affected by different concentrations of compound were recorded, and the dose–response curves were fitted with the ‘temperature jump and thermophoresis’ mode. Data analysis was performed using MO. Affinity Analysis version 2.3 (NanoTemper) and Prism version 8.0.2 (GraphPad Software, United States).

The C. jejuni WT NCTC 11168 was used to generate the 11168ΩTlp11 strain by means of double-crossover homologous recombination, using a suicide plasmid containing homologous arms that flanked the target gene (Duan et al., 2023). The insertion site for tlp11 was selected between 16 s RNA and 23 s RNA in the genome of C. jejuni NCTC 11168. The linear fragment 16sRNAup-tlp11-km-23sRNAdown, which contained 1,044 bp of the upstream 16sRNA, full-length tlp11 with its promoter, a kanamycin resistance gene, and 1,026 bp of the downstream 23sRNA, was constructed using overlap PCR. The fragment 16sRNAup-tlp11-km-23sRNAdown was transferred into E. coli GB08-red containing pBJ114, to generate the plasmid pBJ110-Ωtlp11. The inserted sequences in the transformants were verified by means of enzyme digestion and sequencing. pBJ110-Ωtlp11 was electroporated into C. jejuni NCTC 11168 using a modified protocol adapted from a previous report (Duan et al., 2023), and screened using kanamycin, to obtain C. jejuni NCTC 11168ΩTlp11. To construct the 11168ΩTlp11/ΔCheY strain, the cheYup-cm-cheYdown fragment containing 1 kb homologous arms of upstream and downstream cheY and the chloramphenicol resistance gene, was ligated into pBJ114 by means of Red/ET recombination, to generate the plasmid pBJ110-ΔcheY. Next, pBJ110-ΔcheY was electroporated into C. jejuni NCTC 11168ΩTlp11 competent cells and screened with chloramphenicol, to generate C. jejuni NCTC 11168ΩTlp11/ΔCheY. The C. jejuni mutants were verified by means of PCR and DNA sequencing.

The C. jejuni ATCC 33560, C. jejuni WT NCTC 11168 and 11168ΩTlp11 cells were grown on a MH agar plate overnight, at 37°C, under microaerobic conditions. The colonies were suspended in PBS (Duan et al., 2023), pH 7.2, inoculated into Minimal Essential Medium broth (Life Technologies) containing 10% fetal bovine serum to an initial OD₆₀₀ of 0.01, and grown at 37°C, 100 rpm, under microaerobic conditions. When required, Minimal Essential Medium was supplemented with different concentrations of methyl pyruvate or pyruvate (both from Macklin, China).

The C. jejuni WT NCTC 11168, 11168ΩTlp11, and 11168ΔCheY cells were cultured on MH agar plates, at 37°C, for 18–24 h, under microaerobic conditions. The colonies were then harvested and resuspended in MH broth, until an OD₆₀₀ of 0.05 was attained. Semi-solid MH agar (0.4%) was used to observe the spreading of C. jejuni. Briefly, 1 μL of the cell solutions were added to the semi-solid MH agar plates and incubated at 37°C, for 48 h, under microaerobic conditions. The diameter of the colony ring on the semi-solid MH agar plate was measured.

The E. coli cells expressing hybrid chemoreceptor and GFP were grown in tryptone broth, up to an OD₆₀₀ of 0.55, at 34°C, with shaking at 250 rpm. The cells were harvested by means of centrifugation at 5,000 rpm, washed twice with Tethering Buffer (10 mM KH₂PO₄, 10 mM K₂HPO₄, and 0.1 mM EDTA, pH 7.0), and resuspended in Tethering Buffer such that an OD₆₀₀ of 5.5 was attained. The chemotactic responses of E. coli cells to the compounds were measured using a microfluidic device (Si et al., 2012). The collected E. coli cells were added to the sink pores of the device and allowed to freely diffuse into the observation channel for 40 min. Subsequently, the compound solution was added to the source pores, to establish a concentration gradient gradually in the observation channel of the device. Fluorescence intensity in the observation channel was detected using an LSM 800 laser scanning confocal microscope (Zeiss, Germany). Data were analyzed using ImageJ software.

C. jejuni chemotaxis assays were performed using the microfluidic device described above. C. jejuni was grown in Brucella broth containing 10% fetal bovine serum, at 37°C, with shaking at 100 rpm, under microaerobic conditions, until an OD₆₀₀ of 0.2 was attained. The cells were collected by means of centrifugation at 3,000 rpm for 5 min, and resuspended in PBS, such that an OD₆₀₀ of 2.0 was attained. The collected C. jejuni cells were loaded into sink pores and allowed to diffuse into the observation channel. After 20 min, compound solutions were added to the source pores, to form a concentration gradient. Chemotactic responses were observed in the phase-contrast mode, using an inverted fluorescence microscope (TI-E, Nikon, Japan). The results were analyzed by counting the numbers of bacteria, about 200–600 cells, in the analysis region (100 × 100 μm) of the observation channel of microfluidic device. Data analysis was performed using Prism version 8.0.2.

The promoters of mgtLA and ompC were amplified from the E. coli MG1655 genome and attached to the gfp gene by means of overlap PCR, to obtain P_mgtLA-gfp and P_ompC-gfp. The linear fragment, P_mgtLA-gfp or P_ompC-gfp, was cloned into the vector pUA66 at the XhoI and BamHI restriction sites, to generate pUA66-P_mgtLA-GFP and pUA66-P_ompC-GFP, respectively, as plasmids encoding the reporter. These reporter plasmids were then transformed into E. coli MG1655, and the fluorescence intensity triggered by 400 mM NaCl (mediated via the PhoQ-PhoP-GFP system) or 20% sucrose (mediated via the EnvZ-OmpR-GFP system) was detected using an Imaging Flow Cytometer (ImageStream^X Mark II, Merck, United States).

The reporter plasmid pUA66-P_mgtLA-GFP and the plasmid encoding the hybrid kinase were together transformed into E. coli MG1655/ΔPhoQ strain. E. coli cells containing the reporter and Tlp11-PhoQ chimera were incubated in Medium A, at 37°C, with shaking (200 rpm), until an OD₆₀₀ of 0.4 was attained, following which different concentrations of methyl pyruvate were added to the medium and the cells were incubated for 40 min. The fluorescence intensity of the cells was monitored using a plate reader (Varioskan^™ LUX, Life Technologies) equipped with a 488 nm laser and a 528/12 bandpass filter. The acquired data were analyzed using Prism 8.0.2.

The linear fragment phoQup-tet-phoQdown, which contained 500 bp of the upstream phoQ, a tetracycline resistance gene, and 500 bp of the downstream phoQ, was generated by means of overlap PCR, and then ligated into pRE112 by means of Red/ET recombination, to construct the plasmid pRE112-ΔphoQ. The helper plasmid pTKRED (Lange et al., 2019) was transformed into E. coli MG1655, rendering the cells resistant to spectinomycin. Transformed cells were cultured in LB with 60 μg mL⁻¹ spectinomycin, culture at 30°C, 200 rpm, to an OD₆₀₀ of 0.2, then add 0.5 mM IPTG to an OD₆₀₀ of 0.5. The cells were harvested to produce competent cells, and the linear fragment phoQup-tet-phoQdown amplified from pRE112-ΔphoQ was transformed by electroporation into the E. coli MG1655-pTKRED competent cells and screened on agar plates containing tetracycline and spectinomycin. The pTKRED plasmid was removed after 3 generations at 42°C on agar plates containing tetracycline. The mutant cells obtained were validated using PCR and sequencing molecular dynamics.

The three-dimensional structure of Tlp11-LBD was predicted using AlphaFold 2 (Jumper et al., 2021). MD simulations were conducted for conformation optimization of Tlp11-LBD, using GROMACS 2021.5 package (Sarkar et al., 2022). Tlp11-LBD was parameterized using the Amberff14sb force field (Maier et al., 2015). A cubic box was established by extending at least 1.4 nm outward along the protein (10 × 10 × 10 nm³), following which the system was solvated in TIP3P water, and 0.15 M NaCl (99 Na⁺ and 90 Cl⁻) was added to maintain electrical neutrality. Energy minimization was performed using the steepest descent algorithm, with a force tolerance of 500 kJ mol⁻¹ nm⁻¹. Periodic boundary conditions were imposed in all the three directions. The system was relaxed for 1 ns under the NPT ensemble and position restraints with a constant of 1,000 kJ mol⁻¹ nm⁻² in three directions were applied to the heavy atoms of the protein.

After completing the above steps, 100 ns NPT MD simulations were performed. The pressure was maintained at 1 bar using a Parrinello-Rahman barostat (Braga and Travis, 2006), in an isotropic manner, while the temperature was maintained at 310 K using a V-rescal thermostat (Bussi et al., 2007). The LINCS algorithm was used to constrain the bond lengths of hydrogen atoms (Hess, 2008). Lennard-Jones interactions were calculated within a cutoff of 1.2 nm, and electrostatic interactions beyond 1.2 nm were treated using the particle-mesh Ewald method, with a grid spacing of 0.16 nm. PyMOL (Seeliger and De Groot, 2010) was used to visualize the results.

Molecular docking was performed using AutoDockTools-1.5.6 (Sousa et al., 2006), to predict the interaction between Tlp11-LBD and compounds. A docking box was constructed around the membrane-distal and -proximal pockets of Tlp11-LBD, as docking sites for compounds. The optimal configuration was determined by calculating the binding free energies of each conformation. The docking results were displayed using PyMOL.

Tlp11-LBD and its mutants were dissolved in Buffer D (25 mM Na₂HPO₄, 25 mM NaH₂PO₄, pH 7.0). Far-ultraviolet CD spectra were recorded with the protein concentration of 0.06 mg mL⁻¹, at 25°C, with a scan rate of 100 nm min⁻¹, in the wavelength range of 190–240 nm, using a JASCO J-1500 CD spectrometer (JASCO, Japan). The spectra were corrected using a solvent. The curves were smoothed using Prism version 8.0.2.

The Tlp11-LBD homologues were searched against the NCBI RefSeq Database (Tatusova et al., 2014). The Tlp11-LBD (NCBI accession number: AZU51669.1) sequence from C. jejuni ATCC 33560 was used as a query in a BLAST search against the NCBI RefSeq Database, and the maximum target sequence was set to 5,000, using default parameters. Tlp11-LBD homologue sequences were aligned using ClustalOmega.¹ A phylogenetic tree was constructed using the neighbor-joining method in MEGA 7.0.26 (Filipski et al., 2014), with 1,000 bootstraps, and the results were displayed in iTOL version 6 (Letunic and Bork, 2021). Conservation pattern analysis of Tlp11-LBD homologues was conducted using WebLogo 3 server.² Amino acids were numbered based on the C. jejuni ATCC 33560 Tlp11 sequence. The results of the multiple sequence alignments have been displayed using GeneDoc (Lanave et al., 2002).

Fluorescence-based TSA, which has been widely used to discover direct-binding molecules of proteins with high efficiency (Fernández et al., 2018), was used to identify the ligand specificity of the C. jejuni chemoreceptor Tlp11. The Tlp11-LBD protein (residues 32–332 of Tlp11) from the strain C. jejuni ATCC 33560 was expressed and purified from E. coli. Hydrophobic reactive dyes were added to the purified protein sample, and the protein was heated to record the change in fluorescence intensity as the temperature increased during the denaturation process of the protein (see Materials and Methods). The melting temperature (T_m) of the target protein was obtained and represented as the midpoint of the protein unfolding transition. Most ligands stabilize the protein during binding, leading to an increase in T_m. Using TSA, we screened a compound library composed of 131 molecules (Supplementary Table S2) that commonly act as microbial carbon and nitrogen sources or metabolites in the human GI tract. Among the compounds in the library, methyl pyruvate elicited the most significant increase in T_m. In the absence of the ligand, the T_m of Tlp11-LBD was 38.3°C, while in the presence of 1 mM methyl pyruvate, the T_m increased by 1.1°C compared to buffer (Figure 1A). The elevated T_m of Tlp11-LBD was concentration-dependent, with an increased concentration of methyl pyruvate triggering a larger ΔT_m (ΔT_m of 1.3°C and 2.8°C at 3 mM and 10 mM methyl pyruvate, respectively; Figures 1A,B), indicating that methyl pyruvate might bind to Tlp11-LBD as a ligand.

We measured the in vitro binding affinity of methyl pyruvate for C. jejuni ATCC 33560 Tlp11-LBD protein using MST. MST quantifies interaction affinities by detecting the direct movement of fluorescent molecules along temperature gradients in capillaries. Ligand binding changes the thermophoretic movement of proteins, which can be used to derive Kds, by sequentially scanning capillaries with different ligand concentrations. MST has been applied to the study of weak binding systems and validated to provide reliable results (Linke et al., 2016; Gao et al., 2021). We observed that compared to the negative control in the buffer, supplementation with different concentrations of methyl pyruvate significantly affected the thermophoresis of Tlp11-LBD in the MST experiments (Figure 1C). The derived Kd of Tlp11-LBD binding to methyl pyruvate was 688 ± 140 μM at pH 7.0, indicating that methyl pyruvate is a direct-binding ligand for Tlp11-LBD. The Kd of methyl pyruvate measured by MST was consistent with the rapid increase in ΔT_m of Tlp11-LBD elicited by 100 μM to 1 mM methyl pyruvate observed in the TSA experiments (Figure 1A). We also tested the binding abilities of methyl pyruvate analogues to Tlp11-LBD using MST (Supplementary Table S3). Except for ethyl pyruvate that could bind to Tlp11-LBD with a Kd of 2.36 ± 0.17 mM, the other analogues, including pyruvate, did not bind to Tlp11-LBD (Supplementary Figures S1A–G), suggesting that Tlp11-LBD has a high specificity for binding to methyl pyruvate.

To understand the correlation between the physiological function and specific interaction of Tlp11 with methyl pyruvate, we detected the chemotactic responses of C. jejuni towards methyl pyruvate. Although C. jejuni ATCC 33560 encodes Tlp11 and has motile and invasive abilities (Konkel and Joens, 1990; Hazeleger et al., 1998), difficulties in the genetic operation of this strain make it difficult to achieve functional verification of Tlp11. Moreover, ATCC 33560 is more sensitive to oxygen, which makes it difficult to perform chemotaxis measurements using this strain in our microfluidic assays. Therefore, we expressed Tlp11 of ATCC 33560 in C. jejuni NCTC 11168, which does not encode Tlp11, but has been widely used to study C. jejuni chemotaxis in recent studies (Hartley-Tassell et al., 2010; Duan et al., 2023), to obtain the Tlp11-expressing strain 11168ΩTlp11. Although the spreading ring of 11168ΩTlp11 cells on the semi-solid MH agar plate was smaller than that of WT NCTC 11168 (Supplementary Figure S2), 11168ΩTlp11 and 11,168 WT had a similar chemotactic strength to formate (Figures 2A,C), indicating that the overexpression of Tlp11 in NCTC 11168 might induce a higher frequency of tumbling. 11,168 ΔCheY strain, which was reported with low motile ability (Chandrashekhar et al., 2015), had a much smaller spreading ring than 11168ΩTlp11 (Supplementary Figure S2).

We measured the responses of 11168ΩTlp11 cells towards methyl pyruvate using a microfluidic device reported previously (Supplementary Figure S3) (Duan et al., 2023). C. jejuni cells were loaded into the sink pores of the device and allowed to swim into the observation channel. The methyl pyruvate solution was loaded into the source pores and allowed to diffuse through the agarose gel into the observation channel, to form a concentration gradient. If methyl pyruvate is the attractant, C. jejuni cells sense the attractant gradient and move from the sink pore into the observation channel, thereby accumulating towards the source. However, if it is the repellent, cells move out of the observation channel towards the sink pore, thereby decreasing the cell intensity in the observation channel. The cell number was counted (~200–600 cells) in each of the analysis region (100 × 100 μm) of the observation channel in microfluidic device, in response to the ligand or without ligand (buffer), for 30 min. The number of cells in response to ligand gradient was normalized to that before adding the ligand, to obtain the chemotactic index (CI). The CI of cells in response to ligand gradient was then normalized to that of cells in the buffer, to get relative CI. A relative CI >1 indicates an attractant response, whereas a relative CI <1 indicates a repellent response.

We used formate, a reported attractant for the C. jejuni chemoreceptor Tlp1, as the positive control (Vegge et al., 2009; Duan et al., 2023), and a blank buffer as the negative control, for measuring the chemotactic response. The 11168ΩTlp11 cells swam up the formate gradient in the observation channel, and the number of accumulated cells increased over time, indicating an attractant response to formate (Figure 2A). In contrast, cell density remained almost unchanged in the blank buffer. Similar to the response towards formate, 11168ΩTlp11 cells exhibited a robust attractant response towards methyl pyruvate, with cells moving up the gradient and accumulating in the observation channel (Figures 2A,B). This attractant response was concentration-dependent, indicating that methyl pyruvate is a novel attractant for ΩTlp11. In contrast, although the WT NCTC 11168 cells also exhibited an attractive response to formate, they failed to show any chemotaxis towards methyl pyruvate (Figures 2B,C), suggesting that the WT NCTC 11168 cells could not respond to methyl pyruvate, and the attractant response of ΩTlp11 towards methyl pyruvate is mediated specifically via Tlp11.

As a control, the non-chemotactic mutant of 11168ΩTlp11 with CheY deletion (11168ΩTlp11/ΔCheY) showed no chemotaxis towards formate or methyl pyruvate (Figure 2D), indicating that the chemotaxis system mediates the response towards methyl pyruvate in 11168ΩTlp11. Therefore, our results demonstrated that the chemoreceptor Tlp11 of C. jejuni ATCC 33560 mediates the attractant response towards methyl pyruvate through a chemotaxis system.

To understand the physiological significance of the attractant response to methyl pyruvate, we explored the effects of methyl pyruvate on the growth of C. jejuni. A previous study suggested that methyl pyruvate improves C. jejuni growth (Wagley et al., 2014). We measured the growth curves of the C. jejuni ATCC 33560 strain in Minimal Essential Medium supplemented with different concentrations of methyl pyruvate or pyruvate under the shaking condition. The results showed that at concentrations above 1 mM, methyl pyruvate significantly promoted the growth of C. jejuni ATCC 33560 (Figure 2E). This is possibly because methyl pyruvate can be converted to pyruvate, the product of glycolysis, and metabolized by the tricarboxylic acid cycle, to serve as a carbon source for C. jejuni growth, as suggested by a previous study (Wagley et al., 2014). Similar to that observed for methyl pyruvate, pyruvate-mediated promotion of the growth was also observed (Figure 2F). Therefore, methyl pyruvate has metabolic value, and its chemotaxis may be beneficial for the growth of C. jejuni. In addition, the WT NCTC 11168 and 11168ΩTlp11 showed similar growth in addition of methyl pyruvate (Supplementary Figure S4), indicating that the presence of Tlp11 does not affect the C. jejuni methyl pyruvate growth response under the shaking condition.

To determine whether methyl pyruvate stimulates transmembrane signaling through Tlp11-LBD, we fused C. jejuni ATCC 33560 Tlp11-LBD with the cytoplasmic region of the E. coli chemoreceptor Tar to construct Tlp11-Tar hybrid chemoreceptors (Figure 3A). Five hybrid receptors with different fusion positions in the second transmembrane helix (TM2) were obtained: Tlp11[1-342]-Tar[200-553] (Tlp342Tar200), Tlp11[1-343]-Tar[200-553] (Tlp343Tar200), Tlp11[1-346]-Tar[204-553] (Tlp346Tar204), Tlp11[1-347]-Tar[204-553] (Tlp347Tar204), and Tlp11[1-348]-Tar[204-553] (Tlp348Tar204) (Supplementary Figure S5A). To verify the activities of these hybrid chemoreceptors, we expressed each protein as a chimera in green fluorescent protein (GFP)-labeled E. coli VS188 without any other chemoreceptors, such that each hybrid receptor served as the only chemoreceptor in the E. coli. Glucose was used as the effector to screen for hybrid receptor activity, as it is a substrate of the phosphotransferase system that stimulates functional chemoreceptors in the receptor-CheA-CheW ternary complex and triggers an attractant response (Somavanshi et al., 2016; Bi and Sourjik, 2018). These chemotactic responses triggered by glucose through the phosphotransferase system are independent of the chemoreceptor LBD (Neumann et al., 2012). Using the microfluidic device described above, we observed that of the five hybrid receptors, GFP-labeled E. coli expressing Tlp342Tar200 showed the strongest chemotaxis in the glucose gradient (Figure 3B), suggesting that Tlp342Tar200 had the best communication with the E. coli chemosensory pathway. Thus, we used this hybrid receptor for subsequent chemotaxis measurements.

Observation of the response of E. coli expressing Tlp342Tar200 as the sole receptor for methyl pyruvate revealed that Tlp342Tar200 mediated a strong attractant response to methyl pyruvate, with the cells moving up the methyl pyruvate gradient and accumulating in the observation channel (Figure 3C), in a concentration-dependent manner (Figure 3D). As a control, cells expressing the pin-head Tar (Tar without LBD) (Bi et al., 2016) or WT Tar as the sole chemoreceptor showed no response or a much weaker response to methyl pyruvate (Supplementary Figures S5B,C), thus supporting that methyl pyruvate is a signaling molecule specifically sensed by Tlp11-LBD, and its binding can stimulate transmembrane signaling in Tlp11.

To further explore the signal transduction ability of Tlp11-LBD in sensing methyl pyruvate, we enrolled C. jejuni ATCC 33560 Tlp11-LBD in engineered TCSs, to evaluate its response to methyl pyruvate, using the fluorescence intensity of GFP as a reporter. E. coli PhoQ/PhoP and EnvZ/OmpR were chosen as the target TCSs for the first design. To construct the signal output module, gfp was placed under the control of the promoter region of mgtLA, which is regulated by PhoP, or the promoter of ompC, which is regulated by OmpR (Supplementary Figure S6). E. coli MG1655 expressing the PhoQ-PhoP-GFP system showed a clear response to osmotic pressure triggered by 400 mM NaCl (Supplementary Figure S7A), the stimulus reported for PhoQ (Yuan et al., 2017). However, the EnvZ-OmpR-GFP system showed fluorescence leakage and could not respond properly to the osmotic pressure triggered by 20% sucrose (Wang et al., 2012) (Supplementary Figure S7B); therefore, we utilized the PhoQ-PhoP-GFP system for the subsequent design.

To construct the signal input module, we fused the Tlp11-LBD with the PhoQ cytoplasmic region, to design the Tlp11-PhoQ hybrid kinases (Supplementary Figure S6), Tlp11[1-337]-PhoQ[200-486] (Tlp337PhoQ200), Tlp11[1-339]-PhoQ[200-486] (Tlp339PhoQ200), Tlp11[1-339]-PhoQ[201-486] (Tlp339PhoQ201), Tlp11[1-340]-PhoQ[202-486] (Tlp340PhoQ202), Tlp11[1-340]-PhoQ[200-486] (Tlp340PhoQ200), Tlp11[1-341]-PhoQ[202-486] (Tlp341PhoQ202), Tlp11[1-342]-PhoQ[202-86] (Tlp342PhoQ202), Tlp11[1-343]-PhoQ[202-486] (Tlp343PhoQ202), which connected Tlp11 and PhoQ in TM2 (Supplementary Figure S7C). The activity of each chimera in sensing methyl pyruvate was determined by measuring the fluorescence intensity of GFP in the E. coli MG1655/ΔPhoQ strain. Cells expressing Tlp343PhoQ202 showed the largest increase in fluorescence intensity upon addition of methyl pyruvate, with the highest fold-change representing the ratio of fluorescence intensity in the presence of methyl pyruvate to that without methyl pyruvate, normalized to cell density (Figure 3E). Moreover, the response of Tlp343PhoQ202 towards methyl pyruvate exhibited significant concentration-dependence (Figure 3F), suggesting that Tlp343PhoQ202 senses methyl pyruvate and triggers downstream signals. However, control cells expressing full-length PhoQ or an empty vector showed almost no response towards methyl pyruvate (Figures 3E,F). These results indicated that the signal transduction elicited by methyl pyruvate occurs through Tlp11-LBD.

To understand how methyl pyruvate binds to C. jejuni ATCC 33560 Tlp11-LBD, the structure of Tlp11-LBD was modeled using Alphafold 2 (Jumper et al., 2021) and molecular dynamics (MD) simulations were conducted to optimize the conformation obtained (Supplementary Figure S8). The binding free energy for each predicted binding mode of methyl pyruvate to Tlp11-LBD was calculated by means of molecular docking. We tried to dock methyl pyruvate into the pockets of both membrane-proximal and -distal subdomains of Tlp11-LBD, and found that it could only be docked into the membrane-proximal pocket, and the lowest binding free energy was −3.04 kcal mol⁻¹. In addition, the predicted lowest binding free energy of pyruvate interacting with Tlp11-LBD membrane-proximal pocket was −1.39 kcal mol⁻¹, consistent with the MST result that the binding of pyruvate to Tlp11-LBD was not detectable (Supplementary Figure S1C). The docking results showed that methyl pyruvate may interact with the residues L264, N268, I276, Y291, V318, and T320 in the membrane-proximal pocket, at a distance of <3.5 Å. The ester and keto groups of methyl pyruvate formed hydrogen bonds with the residues N268, Y291, and T320. The residues L264, I276, and V318 were also spatially adjacent, and may have interacted with methyl pyruvate through hydrophobic interactions (Figure 4A).

In order to confirm the contribution of L264, N268, I276, Y291, V318, and T320 to the binding to methyl pyruvate, we generated Tlp11-LBD proteins with an alanine point mutation at individual residues, to obtain the mutant proteins L264A, N268A, I276A, Y291A, V318A, and T320A, and then analyzed the binding affinities of methyl pyruvate to these mutants using MST. These mutant proteins significantly impaired the binding of Tlp11-LBD to methyl pyruvate, and the MST measurement-derived binding curves of these mutants to methyl pyruvate were unable to obtain fitted Kd values (Figures 4B–D; Supplementary Figures S9A–C). Circular dichroism spectroscopy showed that a single mutation did not affect the secondary structures of L264A, N268A, I276A, Y291A, V318A, or T320A (Supplementary Figure S10), indicating that these residues are crucial for binding to methyl pyruvate. Compared to pyruvate, the additional alkyl group of methyl pyruvate or ethyl pyruvate might form hydrophobic interactions with L264 and V318 (Figure 4A; Supplementary Figure S11), which could stabilize the binding inside the pocket.

Mutations on hydrophobic residues mentioned above greatly affected the binding of Tlp11-LBD to methyl pyruvate, indicating the importance of hydrophobic interactions for ligand binding. In addition, there are aromatic residues F243 and Y291 present in the membrane-proximal pocket of ATCC 33560 Tlp11-LBD (Supplementary Figure S12), which might form π–π interactions with aromatic molecules (Oshita and Shimazaki, 2022). We thus selected some aromatic compounds and detected their binding abilities to Tlp11-LBD using MST (Supplementary Table S4). Among these compounds, toluene and quinoline could bind directly to Tlp11-LBD, with Kd of 844 ± 172 μM and 905 ± 180 μM, respectively (Figures 5A,B). Molecular docking showed that both compounds might bind to the same membrane-proximal pocket as methyl pyruvate (Supplementary Figure S12). Toluene and quinoline were predicted to orientate similarly, and they might interact with additional hydrophobic residues compared to methyl pyruvate, including F243, I245, I251, and I267, while lose the interaction with T320 (Figures 5C,D; Supplementary Figure S12).

To explore the signal transduction induced by the binding of Tlp11 to toluene and quinoline, we first observed the chemotaxis of E. coli expressing the hybrid chemoreceptor Tlp342Tar200 towards these compounds. However, both toluene and quinoline failed to elicit a chemotactic response in a concentration-dependent manner (Supplementary Figures S13A,B). We then observed the chemotactic behavior of C. jejuni towards toluene and quinoline, but neither WT NCTC 11168 nor 11168ΩTlp11 cells exhibited chemotaxis towards them (Supplementary Figure S13C), indicating that toluene and quinoline act as the antagonists of Tlp11 and could not trigger the signal transmission.

To further ascertain the antagonistic function of toluene and quinoline, we performed competitive binding experiments with methyl pyruvate using MST. When Tlp11-LBD was saturated with 5 mM toluene or quinoline, the Kd of Tlp11-LBD interacting with methyl pyruvate increased to 18 ± 3 and 3.9 ± 1.1 mM, respectively (Supplementary Figures S14A,B), indicating that toluene and quinoline compete with the methyl pyruvate-binding pocket. We also measured the effects of toluene and quinoline on the chemotaxis of E. coli expressing Tlp342Tar200 to methyl pyruvate. When E. coli cells with Tlp342Tar200 were incubated in 10 mM toluene or quinoline, they lost the chemotaxis towards methyl pyruvate, while their responses to glucose were not affected (Figure 5E). All these results indicated that toluene and quinoline function as the antagonists by competitively binding to the methyl pyruvate-binding site in Tlp11-LBD.

Analysis of 100 genomes from different C. jejuni strains in the NCBI Genome Database showed that Tlp11 is encoded by ~15% of the C. jejuni genomes (Supplementary Table S5), which is consistent with a previous report (Day et al., 2016). Next, we analyzed the distribution of Tlp11-LBD homologous proteins that may also bind to methyl pyruvate in other bacterial species. The Tlp11-LBD sequence (residues 32-332) from C. jejuni ATCC 33560 (NCBI accession number: AZU51669.1) was used as a query in a BLAST search against the NCBI RefSeq Database (Tatusova et al., 2014), and 44 sequences (including ATCC 33560 Tlp11) from different species with e-values less than e⁻⁵ were obtained (Supplementary Table S6). The identity of these sequences with those of C. jejuni Tlp11-LBD was >22% (Supplementary Table S6). We found that these potential Tlp11-LBD homologues mainly exist in the order Campylobacterales, including species from the host-associated genera Campylobacter, Helicobacter, Wolinella, and the family Arcobacteraceae (Figure 6A). The distribution of key residues for methyl pyruvate-binding in the active pocket of Tlp11 was analyzed in these 44 protein sequences by means of alignment, which revealed the presence of five crucial residues L264, N268, I276, Y291, and T320 in Tlp11 at the highest frequencies for binding to methyl pyruvate (Figure 6B). Two protein sequences contained all six key residues, while totally 21 sequences contained the conserved hydrophilic residues N268, Y291, and T320, which are distributed in the genera Campylobacter and Helicobacter (Supplementary Figure S15). All 44 proteins were chemoreceptors.

To evaluate whether these potential Tlp11-LBD homologues could bind to methyl pyruvate, we conducted a structural analysis of the LBDs of six chemoreceptors from C. coli, H. equorum, H. himalayensis, H. mesocricetorum, H. ganmani, and C. upsaliensis using Alphafold 2. All of these LBDs are dCache structures, and the spatial orientations of N, Y, and T corresponding to Tlp11-N268Y291T320, which is involved in the binding of methyl pyruvate in the membrane-proximal pocket, are similar to those of Tlp11-LBD (Figure 6C; Supplementary Figures S16A–F). Assessment of the binding abilities of these homologues to methyl pyruvate using MST showed that the chemoreceptors of all these 6 LBDs from C. coli, H. equorum, H. himalayensis, H. mesocricetorum, H. ganmani, and C. upsaliensis had the ability to bind to methyl pyruvate, with Kd values of 1.4 ± 0.5 mM, 340 ± 43 μM, 1.1 ± 0.3 mM, 1.2 ± 0.4 mM, 1.6 ± 0.2 mM, and 1.9 ± 0.5 mM, respectively (Figures 6D–F; Supplementary Figures S17A–C). These results demonstrated that chemoreceptors with a dCache domain that bind to methyl pyruvate are conserved in the order Campylobacterales.