We selected the structure of the classical LuxR-family protein Agrobacterium tumefaciens TraR (PDB ID: 1L3L) as the template for MD simulations to obtain a stable structure of V. fischeri (also called Aliivibrio fischeri) LuxR. During the modeling and kinetic process, the LuxR protein was surrounded by TIP3P-type water molecules, and sodium and chloride ions were added to the system at an ion concentration of 150 mM to ensure that the molecules were in a normal physiological environment and the system maintained electrical neutrality. Visual Merchandising Design (VMD) software with MODELLER was used for modeling and visualization, and dynamic optimization was performed using the open-source kinetic calculation software NAnoscale Molecular Dynamic (NAMD) (Humphrey et al., 1996; Phillips et al., 2005). The Chemistry at Harvard Macromolecular Mechanics (CHARMM) force field was used for the kinetic process, the temperature of the system was 310 K, and the pressure was 1 bar (MacKerell et al., 1998 ; Teng et al., 2017). The kinetic process used a truncation function calculated from 13 to 15 Å. The MD simulation time was 100 ns.

AutoDock software version 4.2 (http://autodock.scripps.edu) was used for molecular docking as previously described (Morris et al., 2009; Teng et al., 2019). For docking to apo- (i.e., ligand-free) LuxR, polar hydrogen atoms were added to LuxR and its nonpolar hydrogen atoms were merged using AutoDock Tools (http://mgltools.scripps.edu). The structure of apo-LuxR served as the receptor for docking with 3OC6-HSL. Binding energy was calculated as the sum of the intermolecular energy and the torsional free energy. AutoDockTools 1.5.6 software was used to prepare the semi-flexible docking procedures. The default 0.375 Å spacing was adopted for the grid box, and the volume was set as 32 × 32 × 32 Å. The key residues for the binding of LuxR and 3OC6-HSL were selected based on an extensive alignment of LuxR-family proteins from both a sequence and a structural perspective, then selected as the docking pocket. The number of runs was set to 100.

Binding free energy between LuxR and 3OC6-HSL was calculated using the Molecular Mechanics Generalized Born Surface Area (MM-GB/SA) method (Genheden and Ryde, 2015; Teng et al., 2019). For each system, 1,000 snapshots without water and ions were extracted from the last 20 ns of the equilibration trajectory.

Binding free energy (∆G _bind ) was calculated as: Δ G b i n d = G c o m p l e x − ( G r e c e p t o r + G l i g a n d ) (1) where G _complex , G _receptor , and G _ligand are the binding free energies of the complex, the receptor, and the ligand, respectively.

Free energy (∆G _TOT ) was calculated using the equations: Δ G T O T = E M M + G s o l − T D S (2) Δ E M M = Δ E e l e + Δ E v d W (3) Δ G s o l = Δ G p o l a r + Δ G n o n p o l a r (4) G n o n p o l a r = g × S A S A + b (5) where ∆EMM, ∆G _sol , and TDS refer to molecular mechanics free energy, solvation free energy, and entropy in the gas phase, respectively. ∆EMM can be divided into electrostatic interactions (∆E _ele ) and van der Waals interactions (∆E _vd W). ∆G _sol can be divided into polar contributions (∆G _polar ) and nonpolar contributions (∆G _nonpolar ). ∆G _polar is calculated using the Generalized Born method. ∆G _nonpolar is considered to be proportional to the molecular solvent-accessible surface area (SASA) buried upon binding, in which g and b were set to the default values of 0.0072 kcal (mol·Å²) and 0.00, respectively.

The pMAT9S-based plasmids for expressing LuxR-WT–MBP fusions and the corresponding mutants were respectively transformed into E. coli BL21 (DE3) competent cells. Cultures were grown in LB medium containing ampicillin at 310 K until the optical density at 600 nm reached 0.8. Isopropyl-β-D-thiogalactopyranoside was then added to a final concentration of 0.4 mM and culture was continued at 289 K for 16 h. Thereafter, cells were harvested by centrifugation and the bacterial pellets were resuspended in phosphate-buffered saline (140 mM NaCl, 10 mM Na2HPO4, 2.7 mM KCl, 1.8 mM KH2PO4, pH 7.3) supplemented with 15% glycerol. After high-pressure crushing at 277 K, bacterial lysates were centrifuged at 18,000 × g for 40 min at 277 K and the pellet was discarded. The supernatant was loaded onto a disposable column containing amylose resin (New England Biolabs, cat. no. E8021S) to purify MBP-tagged WT or mutant LuxR. The fusion proteins were eluted using 20 mM maltose dissolved in resuspension buffer and further purified by anion-exchange chromatography using a HiTrap Q column (GE Healthcare) with a linear gradient from 25 to 250 mM NaCl in 20 mM Tris-HCl (pH 8.0) and 15% glycerol, followed by gel filtration chromatography using a Superdex-200 column (GE Healthcare) in resuspension buffer.

SPR experiments were performed on a Biacore T200 SPR system (GE Healthcare) at 25°C in PBS-P (20 mM phosphate buffer with 2.7 mM KCl, 0.137 mM NaCl, and 0.05% surfactant P20). MBP-stabilized fusions of LuxR or its variants in PBS-P were respectively immobilized on CM5 chips by amine coupling. After twofold serial dilution of 3OC6-HSL in 5% (v/v) dimethylsulfoxide (DMSO), it was injected into the flow cell for 60 s at a flow-rate of 30 μl/min for association, followed by a 60-s dissociation phase. A high DMSO concentration was found to be necessary for solubility of 3OC6-HSL, helping to minimize nonspecific interactions and aggregation on the chip. The noise effect resulting from differences in DMSO concentration was corrected over a limited range (4.5–5.8%) of standards. Background binding to blank immobilized flow cells was subtracted, and equilibrium dissociation constant (Kd) values were calculated using the steady-state affinity model in BIAcore T200 Evaluation Software v.3.2.

Plasmid constructs were assembled to comply with the BioBrick RCF standard using a variety of molecular cloning techniques. Biological parts were acquired from the Registry of Standard Biological parts or produced by chemical synthesis. The constitutive promoter P_Con (BBa_J23104) with a ribosome-binding site (RBS; BBa_B0034) was inserted upstream of the synthetic V. fischeri luxR gene as the LuxR sender module. RBS (BBa_B0034)–GFP (BBa_E0040)–terminator (BBa_B0015) was inserted downstream of P_lux [a LuxR-HSL-regulated promoter (BBa_C0062)] as the LuxR receiver module. The modules were assembled into vector pSB1A3.

Overnight Escherichia coli cultures (see Supplementary Methods) were diluted 1000-fold into fresh LB medium containing 100 μg/ml ampicillin and transferred to 1.5-ml centrifuge tubes containing different concentrations of 3OC6-HSL (0, 10⁻⁶, 10⁻⁷, 10⁻⁸, 10⁻⁹, 10⁻¹⁰, 10⁻¹¹, 10⁻¹² M). Samples were incubated at 37°C for 8 h, and then transferred to a 96-well microplate and fluorescence was measured using a Synergy H1 hybrid Reader (excitation, 488 nm; emission, 518 nm). Cell densities were measured optically at 600 nm. Mean fluorescence was calculated from at least three biological replicates. In addition, the above 1:1,000 bacterial dilution and 3OC6-HSL (10⁻⁶, 10⁻⁷ M) were added to Corning Black Costar transparent bottom 96-well plates. The samples were incubated in the Synergy H1 hybrid Reader continuously for 8 h (807 rpm, orbital shaking), and the fluorescence intensity and OD₆₀₀ were analyzed at 30-min intervals.

The GFP production of the whole-cell biosensor incubated with culture supernatants of bacterial pathogens was measured as noted below. Overnight E. coli cultures (see Supplementary Methods) were diluted 1:100 and further incubated to logarithmic growth phase in the same conditions. Biosensor cells were collected by centrifugation at 3,000 × g for 4 min and resuspended in an appropriate volume of LB medium to make the OD₆₀₀ = 1. For the detection of native 3OC6-HSL, sterile culture supernatants of Yersinia pestis (Y. pestis), Y. pestis ΔypeRI ΔyspRI, E. coli strain MG1655, and Klebsiella pneumoniae (K. pneumoniae) were collected at 8,000 × g for 5 min and co-cultured with biosensor cells (volume ratio 15: 1) for 11 h at 37°C. We also determined the sensitivity by incubating biosensor cells with sterile culture supernatants of Y. pestis taken from cultures with optical densities 0.2, 0.4, 0.6, 1.0, and 1.5. All GFP fluorescence intensity and OD₆₀₀ measurements were made at 30-min intervals using a microplate reader.

Structural modeling of V. fischeri LuxR (DDBJ accession no. of the coding sequence AF170104) was performed by MD simulation with the LuxR-family protein A. tumefaciens TraR (PDB ID: 1L3L) as the template (Figure 1). The crystal structure of TraR has been obtained by X-Ray diffraction at 1.66 Å resolution. The resulting LuxR structure is depicted in Figure 1B, and the quality of the structure was confirmed by Ramachandran plot analysis (Supplementary Figure S1). Using this structure of LuxR generated as the acceptor molecule and 3OC6-HSL from the TraR crystal structure as the ligand, the semi-flexible docking method in Autodock software was applied to perform molecular docking calculations (Figures 1C,D). PROCHECK software was used to validate the quality of the final LuxR–3OC6-HSL structure; a Ramachandran plot revealed that all residues of the LuxR–3OC6-HSL complex were in favored or allowed regions (91.5 and 8.5% respectively), indicating that the model of the LuxR–3OC6-HSL complex was of good quality (Supplementary Figure S2).

We selected a series of structural states, from 50, 75, and 100 ns (Supplementary Figure S3) in the MD process, to validate the stability of the final LuxR–3OC6-HSL complex structure. As shown in Supplementary Figure S3B, subtle fine-tuning of the position of 3OC6-HSL occurred as the molecular simulation progressed. A final and stable location for the 3OC6-HSL molecule was achieved, in which 3OC6-HSL exactly fitted the binding cavity of LuxR. Based on analysis of the interactions between LuxR and 3OC6-HSL, residues including L42, I58, Y62, W66, D70, and D79 were found to constitute the binding pocket of LuxR. Y62 and D79 form hydrogen bonds with 3OC6-HSL, while W66, I58 and the other residues help to stabilize 3OC6-HSL mainly via hydrophobic interactions.

Based on the above analysis of the interaction profile between LuxR and 3OC6-HSL, we employed the MM-GB/SA method to calculate energies for the key binding sites on LuxR of 3OC6-HSL (Supplementary Figure S4A). Interestingly, I46, a residue >8 Å from the binding site (Figure 2) that is not conserved in A. tumefaciens TraR (Figure 1), had a marked effect on the binding potential between LuxR and 3OC6-HSL. In the structural model, we mutated residues that affected 3OC6-HSL binding into the other naturally occurring proteinaceous amino acids and docked the 3OC6-HSL molecule into each mutant using semi-flexible docking, MD, and kinetic calculations. The change in docking energy relative to the wild-type (WT) protein served to indicate the importance of each site (Supplementary Figure S4B). When I46 was mutated, the heatmap indicated that LuxR-I46F would display the strongest binding energy to 3OC6-HSL, and I46R the weakest interaction (Supplementary Figure S4B), i.e., that mutation LuxR-I46F would improve the response of LuxR to 3OC6-HSL compared with the WT protein, and mutation LuxR-I46R would weaken it. LuxR was thus rationally engineered at position 46.

Because of the tendency of LuxR-family transcription factors to aggregate in the absence of HSLs (Qin et al., 2000), no direct biochemical characterization of the binding affinity between LuxR and 3OC6-HSL has been reported, except for a roughly calculated dissociation constant retrieved from a model fitted to luminescence data (Colton et al., 2015). Here, we applied the rigid MBP technique, which has enabled the crystallization and structural determination of numerous proteins, to production of stable apo-LuxR. In this approach, a rigid helix at the C-terminus of MBP constrains the flexibility of attached recombinant proteins and thereby improves their solubility (Waugh, 2016).

Applying this method, we produced a stable, soluble, apo-form of V. fischeri LuxR (residues 1–179, the N-terminal ligand-binding domain) as a fusion with MBP. We determined the dissociation constants for 3OC6-HSL binding to the MBP-fused LuxR and the LuxR-I46F and LuxR-I46R mutants by surface plasmon resonance (Figure 3; one-phase dissociation fitting). The dissociation constant values were 258.5 μM (the MBP-fused LuxR and 3OC6-HSL), 52.8 μM (the MBP-fused LuxR-I46F and 3OC6-HSL), and 365.0 μM (the MBP-fused LuxR-I46R and 3OC6-HSL), which indicated that—in accordance with the docking energy calculations—I46F mutation of LuxR increased the sensitivity of LuxR for 3OC6-HSL, and I46R decreased it.

We constructed a novel whole-cell biosensor using the various forms of LuxR (WT, I46F and I46R mutants) with different sensitivities to 3OC6-HSL in a one-plasmid system (Figures 4A,B). One module, P_Con–luxR, expressed the transcription factor LuxR under the control of a constitutive promoter as an intermediate for signal transduction. The following reporting module, P_lux–GFP, expressed GFP under the control of the P_lux promoter. The amount of GFP production from P_lux is dependent on the degree of transcriptional activation that occurs as a result of functional LuxR binding to a specific acyl-HSL signaling molecule.

Bioassay using these gene circuit constructs allowed us to quantitatively characterize the different forms of LuxR by GFP production rates in response to varying concentrations of 3OC6-HSL. LuxR-I46F displayed a sensitive and stable response to 3OC6-HSL: this system emitted significant fluorescence at 3OC6-HSL concentrations ∼1 × 10^–10 M, while the construct with LuxR-WT did not respond until the concentration was 1 × 10⁻⁹ M, and that with LuxR-I46R did not respond until it was 1 × 10^–7 M (Figures 4C,D). In terms of the response time, the gene circuit based on LuxR-I46F also displayed sensitivity and stability: The LuxR-I46F system showed a significant fluorescence difference for the first time at 1.5 h with ∼1 × 10⁻⁶ M 3OC6-HSL, and at 2.5 h with ∼1 × 10⁻⁷ M 3OC6-HSL (Figures 4E,F).

To determine the ability of whole-cell biosensors based on LuxR-WT and LuxR-I46F to detect 3OC6-HSL produced by bacterial pathogens, the biosensor cells were incubated in the culture supernatant of Yersinia pestis. Y. pestis is the causative agent of plague, a highly virulent bacterium, and a serious threat to public health. The type I quorum sensing system of Y. pestis includes two sets of LuxI/LuxR QS proteins, namely YpeI/YpeR and YspI/YspR. Signal molecules including 3OC6-HSL are synthesized by YpeI/YspI. The whole-cell biosensors based on LuxR-WT and LuxR-I46F were incubated with culture supernatants of different bacterial strains in logarithmic growth phase (Y. pestis, Y. pestis ΔypeRI ΔyspRI, K. pneumoniae and E. coli MG1655). Only the culture supernatant of Y. pestis contained the target signal molecules; the latter three bacterial culture supernatants served as negative controls. Knockout of YpeI/YpeR and YspI/YspR does not affect the growth of Y. pestis (see Supplementary Figure S6 for growth curves of the various bacteria).

The whole-cell biosensors based on LuxR-I46F displayed a sensitive and stable response to the culture supernatant of Y. pestis: Compared with the negative controls, both LuxR-WT and LuxR-I46F biosensors can detect 3OC6-HSL in the supernatant of Y. pestis containing LuxI/LuxR QS systems. After incubating for 2.5 h, the relative fluorescence intensity of the LuxR-I46F biosensor was significantly higher than that of the LuxR-WT biosensor (Figure 5). The difference in the relative fluorescence intensity of the LuxR-I46F and LuxR-WT biosensors increased significantly with incubation time (Supplementary Figure S5). In addition, the biosensor based on LuxR-I46F can respond accurately and quickly to the culture supernatant of Y. pestis in the initial growth stage [OD₆₀₀ = 0.2 (representing incubation for 0.5 h)] (Figure 5B). In all, the results provide solid evidence that the whole-cell biosensor based on LuxR-I46F can sensitively and rapidly detect native 3OC6-HSL produced by a bacterial pathogen.