The strains and plasmids used in this study are listed in Table 1. Different primers for molecular biology manipulations are provided in Supplementary Table S1.

The DNA fragment of mcpK encoding amino acids Arg³⁸–Ser²⁹³ was amplified using primers McpK-LBD_fw and McpK-LBD_rv and genomic DNA of P. aeruginosa PAO1. The resulting PCR product was cloned into pGEM-T and digested with NdeI and BamHI and then subcloned into the expression plasmid pET28b(+) linearized with the same enzymes. The resulting plasmid, termed pET28-McpK-LBD, was verified by DNA sequencing of the insert and flanking regions.

Escherichia coli BL21 (DE3) containing pET28-McpK-LBD was grown in 2 L Erlenmeyer flasks containing 500 ml LB medium supplemented with 50 μg ml^-1 kanamycin at 30°C until an OD₆₆₀ of 0.6, at which point protein production was induced by adding 0.1 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG). Growth was continued at 18°C overnight before cell harvest by centrifugation at 10,000 g for 30 min. All subsequent manipulations were carried out at 4°C. Cell pellets were resuspended in buffer A [20 mM Tris/HCl, 200 mM NaCl, 10 mM imidazole, 5% (vol/vol) glycerol, pH 8.0] and broken by French press treatment at 1000 psi. After centrifugation at 20,000 g for 1 h, the supernatant was loaded onto a 5 ml HisTrap column (Amersham Bioscience), washed with five column volumes of buffer A and eluted with a 30–300 mM imidazole gradient in buffer A. Protein-containing fractions were pooled.

Thermal shift assays were performed using a BioRad MyIQ2 Real-Time PCR instrument. Ligands were prepared by dissolving Biolog Phenotype Microarray compounds in 50 μl of MilliQ water to obtain a final concentration of around 10–20 mM (as indicated by the manufacturer). Screening was performed with plates PM1, PM2A, PM3B, PM4A, and PM5. Each plate contains 95 compounds and a control (Supplementary Figure S1). Each 25 μl assay mixture contained 40 μM protein in 5 mM Tris, 5 mM Pipes, 5 mM Mes, pH8.0 and SYPRO orange (Life Technologies) at 5x concentration. Aliquots of 2.5 μl of the resuspended Biolog compounds were added to each well. Samples were heated from 23°C to 85°C at a scan rate of 1°C/min. The protein unfolding curves were monitored by detecting changes in SYPRO Orange fluorescence. Melting temperatures were determined using the first derivative values from the raw fluorescence data.

Experiments were conducted on a VP-microcalorimeter (Microcal, Amherst, MA, USA) at 20°C or 10°C. McpK-LBD was dialyzed overnight against 5 mM Tris, 5 mM PIPES, 5 mM MES, pH 8.0 and placed into the sample cell. Typically, protein at 20–100 μM was introduced into the sample cell and titrated with 1–3 mM ligand solutions that were prepared in dialysis buffer immediately before use. For fitting, data were integrated using NITPIC (Keller et al., 2012) before global fitting to a two symmetric-site binding model in SEDPHAT (Houtman et al., 2007). The binding constants expressed are corrected as K_d1 = 2^∗K_d1′ and K_d2 = 0.5^∗K_d2′ in order to express an estimation of the microscopic constants for the two binding sites model, where Kx’ are the macroscopic constants measured. The cooperativity factor α is expressed as α = K_d1/K_d2. Statistical uncertainties for best-fit estimates of K_d and ΔH were calculated using standard error surface projection methods built into SEDPHAT.

Experiments were performed in a Beckman Coulter Optima XL-I analytical ultracentrifuge (Beckman-Coulter, Palo Alto, CA, USA) equipped with UV-visible absorbance as well as interference optics detection systems, using an An50Ti 8-hole rotor and 12 mm path-length charcoal-filled epon double-sector centerpieces. The experiments were carried out at 7°C with at least 1 h stabilizing after reaching 7°C in the rotor chamber, using 5 mM Tris, 5 mM PIPES, 5 mM MES, pH 8.0. Samples of 10–50 μM for McpK-LBD were analyzed in the presence and absence of 1 mM αKG.

Sedimentation velocity (SV) runs were carried out at a rotor speed of 128,793 × g using 400 μL samples with McpK-LBD dialysis buffer as reference. A series of 180 scans without time intervals between successive scans were acquired for each sample. Laser at a wavelength of 236 nm was used in the absorbance optics mode. A least squares boundary modeling of the SV data was used to calculate sedimentation coefficient distributions with the size-distribution c(s) method and the non-interacting discrete species model (Schuck, 2000) implemented in the SEDFIT v14.1 software. The molecular weight was extracted from the sedimentation profiles, via the Lamm equation solution included in the c(s) model of SEDFIT (Schuck, 2000). The best fit values obtained for the sedimentation coefficient (s, in S or Svedbergs) and diffusion were used to estimate the molar mass of the molecule using the Svedberg equation. Buffer density (ρ = 1.00036 g/ml) and viscosity (η = 0.01433 Poise) at 7°C were estimated by SEDNTERP software (Laue et al., 1992) from the buffer components. The partial specific volume used was 0.71274 ml/g as calculated from the amino acid sequence also using SEDNTERP software.

Sedimentation equilibrium (SE) experiments were performed at 7°C, measuring absorbance at 280 nm as a function of radius. Three different concentrations of McpK-LBD (40, 50, and 60 μM) both in the absence or in the presence of 1 mM αKG were loaded with the dialysis buffer as reference and a multi-speed (9,740, 20,606, and 185,462 × g) run was used. The data were analyzed globally by the SEDPHAT (Vistica et al., 2004) “species analysis with mass conservation constraints” model. The goodness of fit was evaluated on the basis of the residuals, expressed as the difference between the experimental data and the theoretical curve.

An unmarked non-polar deletion of pa5072 was created by allelic exchange as described (Schweizer, 1992) with the following modifications: two DNA fragments comprising the 234 bp upstream and 336 bp downstream regions of pp5072 were obtained by PCR amplification of genomic DNA using primers PA5072UpF-HindIII and PA5072UpR-XbaI (upstream region) and primers PA5072DownF-XbaI and PA5072DownR-EcoRI (downstream region). The upstream DNA fragment was digested with HindIII and XbaI and cloned into pUC18NotI digested with the same restriction enzymes, resulting in pUC18NotI-5072Up, whereas the downstream DNA fragment was digested with XbaI and EcoRI and cloned into pUC18NotI-5072Up digested with the same restriction enzymes, resulting in pUC18NotI-5072UpDw. The 570 bp 5072UpDw DNA fragment was digested from pUC18NotI-5072UpDw using NotI and subcloned into the suicide vector pKNG101 a mobilizable suicide vector, hosted routinely in the permissive E. coli strain CC118λpir bearing the positive counter selection marker sacB and conferring resistance to streptomycin. The resulting plasmid, pKNG101-PA5072UpDw was transformed into E. coli CC118λpir. The pKNG101-PA5072UpDw plasmid was mobilized from this strain into P. aeruginosa PAO1 by three-partner conjugation using the E. coli HB101(pRK600) helper strain. Selection for plasmid cointegration in P. aeruginosa PAO1 was accomplished using M9 minimal medium supplemented with 10 mM succinate and 2000 μg/ml streptomycin (Sm). The Sm^r colonies were unable to grow on LB medium containing 10% sucrose (Suc), confirming that the plasmid pKNG101-PA5072UpDw with its sacB gene had integrated into P. aeruginosa PAO1. Transconjugants were plated on LB plates and incubated for 48 h at room temperature. PCR analysis of the Suc^r Sm^s colonies confirmed that gene replacement had occurred.

The DNA fragment corresponding to the mcpK gene and the 607 bp upstream region was amplified by PCR using primers McpK-comp_fw and McpK-comp_rv and genomic DNA of P. aeruginosa PAO1. The resulting fragment was digested with KpnI and XbaI and cloned into pBBR1MCS-2, linearized with the same enzymes. The resulting plasmid pBBR1MCS-2-McpK was verified by DNA sequencing of the insert and flanking regions. The mutant strain was transformed with pBBR1MCS-2-McpK by electroporation (2800 V). Transformed colonies were selected on LB agar plates containing 300 μg/ml kanamycin and verified by colony PCR.

Assays were carried out at a temperature of 25°C. Bacterial cultures were grown to an OD₆₀₀ of 0.35–0.4 in MS medium, washed and resuspended in chemotaxis buffer (30 mM K₂HPO₄, 19 mM KH₂PO₄, 20 μM EDTA and 0.05% (v/v) glycerol, pH 7.0) to an OD₆₀₀ of 0.08-0.1. Polystyrene multi-well plates were filled with 230 μl of the resulting bacterial suspension. For filling with chemoeffector solutions, capillaries (Microcaps, Drummond Scientific, USA) were heat-sealed at one end, warmed over the flame and the open end inserted into the chemoattractant solution. The capillary was immersed into the cell suspension at its open end. After incubation for 30 min, the capillary was removed from the cell suspension, rinsed with water and emptied into an Eppendorf tube containing 1 ml M9 medium. Serial dilutions were made and 20 μl aliquots of the resulting cell suspension were plated onto agar plates containing M9 minimal medium supplemented with 15 mM succinate and incubated at 30°C. Colonies were counted after growth for 24 h. Positive (casamino acids) and negative (buffer only) controls were included in each experiment. Data shown are means from three independent experiments conducted in triplicate.

Two flasks with 20 ml of minimal medium M9 containing 10 mM glucose were inoculated with an overnight culture to an OD₆₀₀ of 0.05. At mid-exponential phase, 1 mM of αKG was added to one of the flasks and 0.5 ml samples were taken after 0, 15, 30, and 45 min. Total RNA was extracted using the High Pure RNA Isolation Kit (Roche Diagnostics) according to the manufacturer’s instructions. RNA was treated with Turbo DNAse (Ambion) and the RNA quality and quantity was analyzed by NanoDrop Spectophotometer and agarose gel electrophoresis. cDNA was synthesized from 500 ng of RNA using the SuperScript II Reverse Transcriptase (Invitrogene) and 200 ng of random primers following manufacturers’ instructions. Quantitative PCR was performed using the iQ SYBR green supermix (BIO-RAD) in a MyiQ^TM2 thermocycler (BIO-RAD). The PCR protocol used was 95°C for 5 min followed by 35 cycles of 95°C (10 s) and 60°C (30 s) and melting curve analysis from 55 to 95°C, with increment of 0.5°C/10 s. Gene expression data were normalized to the expression of the reference gene rpoD and reported as normalized fold expression. The primers for mcpK and rpoD were designed using the Primer3 Plus software.

The kamanycin-resistant strain P. aeruginosa PAO1-Km was constructed by homologous recombination using a derivative plasmid of the suicide vector pKNG101. The initial plasmid pMAMV257 was generated by separately amplifying the 5′ ends of the convergently transcribed genes pa5548 and pa5549 (glmS) of P. aeruginosa PAO1. These PCR products were obtained using primers glmS-EcoRI_fw and gmlS-BamHI_rv (for amplifying 5′ end of glmS) and glmS-BamHI_fw and glmS-HindIII_rv (for amplifying 5′ end of pa5548). Subsequently, the PCR products were digested with EcoRI and BamHI (5′ end of glmS) or BamHI and HindIII (5′ end of pa5548) and ligated in a three-way ligation into pUC18Not. For the generation of the final PAO1-Km strain, the suicide plasmid pMAMV261 was transferred to P. aeruginosa PAO1 by triparental conjugation using E. coli CC118λpir and E. coli HB101 (pRK600) as helper. PAO1 cells, in which pMAMV261 has integrated into the chromosome, were selected on minimal medium containing 400 μg/ml streptomycin and 200 μg/ml kanamycin. To select derivatives that had undergone a second crossover event, sucrose was added to a final concentration of 10% (w/v). The final PAO1-Km strain was confirmed by PCR and sequencing.

Sterilization, germination and inoculation of maize seeds was carried out as described previously, with minor modifications (Matilla et al., 2007). Briefly, sterile seeds were incubated for 1 h at 30°C with a 10⁷ CFU/ml 1:1 mixture of PAO1-Km and ΔmcpK. Thereafter, seeds were rinsed with sterile deionized water and planted in 50 ml Sterilin tubes containing 40 g of sterile washed silica sand and 10% (v/w) plant nutrient solution supplemented with Fe-EDTA and micronutrients. Plants were maintained at 24°C with a daily light period of 16 h. After 7 days, bacterial cells were recovered from the rhizosphere or from 1 mm of the main root apex, as described previously (Matilla et al., 2007). Serial dilutions were plated in LB-agar and LB-agar medium supplemented with 400 μg/ml of kanamycin, to select the wild type strain PAO1-Km.

To identify the putative LBD of chemoreceptor PA5072, its sequence was analyzed using the DAS transmembrane prediction algorithm (Cserzo et al., 1997). Two transmembrane regions were identified that flank the segment of amino acids 38–293, which likely corresponds to the periplasmic LBD (Supplementary Figure S2). The DNA fragment encoding the PA5072-LBD was cloned into an expression vector, the protein overexpressed in E. coli and purified from the soluble fraction of the E. coli lysate.

To identify whether and which ligand bind may bind to this domain, we carried out thermal shift assays in high throughput screening format as reported by McKellar et al. (2015). In this assay, a temperature gradient is applied to a mixture of the purified protein and a fluorescent compound. During protein unfolding buried hydrophobic parts of the protein will be exposed leading to additional dye binding, causing fluorescent changes, which is the signal recorded. Consequently, this assay permits the calculation of the Tm value, which corresponds to the temperature at which half of the protein is in its native form and half is in the unfolded form (Krell, 2015). Ligand binding to a protein causes typically Tm changes, which gives useful initial information as to the identification of potential ligands. This assay has been essential to gain initial insight into the specificity of several chemoreceptors (McKellar et al., 2015; Corral-Lugo et al., 2016; Fernandez et al., 2016).

The thermal shift assays of ligand-free PA5072-LBD resulted in a Tm of 38.7°C. Ligands tested included different bacterial carbon-, nitrogen-, phosphorous-, and sulfur sources as well as to nutrient supplements, which are listed in Supplementary Figure S1. As a representative example, the Tm changes induced by compounds of plate PM1 are shown in Figure 1. A total of 13 compounds were identified that increased or decreased the Tm by at least 2°C (Supplementary Table S2), which is the generally accepted threshold for a relevant ligand-induced change in protein stability. This analysis showed also that α-keto-glutaric acid caused with 5.2°C the most pronounced Tm increase.

Thermal shift assays provide initial information on ligands that bind but represent no evidence of binding. A valid criterion to ascertain binding are isothermal titration calorimetry (ITC) (Krell, 2008) binding experiments. PA5072-LBD was titrated with all compounds that caused Tm changes of at least 2°C (Supplementary Table S2). In these experiments α-ketoglutarate (αKG) was the only compound that showed binding and the corresponding titration data are shown in Figure 2. The titration caused exothermic heat changes (down going peaks), but the biphasic titration curve indicates that the reaction is more complex than the binding of a ligand to a single site at a macromolecule. Data analysis was carried out with several models for the dependent or independent binding of ligands to multiple sites. A very satisfactory curve fit was obtained using a model for the cooperative binding of a molecule to two sites. The initial binding event was characterized by a K_d1 of 301.4 ± 0.2 μM (ΔH1 = -0.16 ± 0.05 kcal/mol), whereas the second event had a K_d2 of 80.90 ± 0.05 μM (ΔH₂ = -2.99 ± 0.15 kcal/mol). The cooperativity factor μ was of 4.41 (with μ of 1 for a non-cooperative process), indicating an approximately fourfold increase in binding affinity of the second binding event as compared to the initial event. Alpha-KG binds thus with positive cooperativity to PA5072-LBD.

The criterion to establish that a given compound does not bind is the absence of binding heats in experiments conducted at two different analysis temperatures to exclude the possibility that exothermic and endothermic contributions to binding cancel out each other at a given analysis temperature. Therefore, binding of compounds that failed to bind at 20°C were also analyzed at 10°C, which in all cases confirmed the absence of binding. As additional control experiment, αKG was titrated into mixtures of protein with ligands that did not cause binding heats to verify protein integrity. In summary, of the compounds that caused significant Tm shifts, only αKG was confirmed as PA5072 ligand.

We then explored by ITC whether other compounds that are structurally similar to αKG may bind to PA5072-LBD and the 15 ligands selected for further ITC studies are provided in Supplementary Table S2. In analogy to the above results, we were unable to detect any binding and concluded that PA5072 binds exclusively αKG. The receptor was therefore named McpK (methyl-accepting chemotaxis protein K).

The unexpected observation of binding with positive cooperativity indicates the presence of higher oligomeric states of the protein analyzed. To assess this issue, we carried out analytical ultracentrifugation (AUC) studies of McpK-LBD in the absence and presence of αKG. Initial SM studies of ligand free protein revealed two species with standard sedimentation coefficients of s_20,w = 2.45 S and s_20,w = 3.18 S (Figure 3; note: these values are standard values normalized for migration in water, whereas Figure 3 shows the experimental data recorded in buffer). The frictional ratio (fr) for the peak at 2.45 S is 1.5, indicative of a rather elongated protein morphology, which agrees with the structure of HBM domains (Pineda-Molina et al., 2012; Ortega and Krell, 2014). Based on this frictional ratio and using the diffusional scaling law of SEDFIT and the Svedberg equation, the average molar mass of the protein was determined 31.1 kDa, which is very close to the sequence-derived mass of the monomer (30.5 kDa). The peak of the fast sedimenting species (s_20,w = 3.18 S) shows a similar fr corresponding to an elongated particle. The molar mass extracted was 45.8 kDa, which may point to a virtual intermediate species resulting from the fast equilibrium between the monomers and dimers. Such virtual species have been observed previously for the analysis of the homologous domain of the P. putida KT2440 McpQ chemoreceptor (Martin-Mora et al., 2016). When the experiment was conducted in the presence of 1 mM αKG a single peak with s_20,w = 3.58 S was observed that translates to a species with a molecular mass of 58.7 kDa, close to the sequence derived mass of the protein dimer (61 kDa).

The self-association behavior observed by SM was further demonstrated by multi-speed SE experiments. The SE of McpK at three different concentrations was analyzed by a global fit that confirmed the presence of both the monomeric and dimeric species at all concentrations (Figure 4). For ligand-free McpK-LBD a dimer self-dissociation constant of 55.0 μM could be determined. When this experiment was repeated in the presence of 1 mM αKG, a tighter association was observed, with an approximately 10-fold lower self-dissociation constant of 5.9 μM, confirming that αKG causes McpK-LBD dimer stabilization.

At the protein concentration used for ITC binding studies, McpK-LBD is thus partially present as dimer and the cooperativity observed is thus likely due to ligand binding to the different monomers of the dimer in a way that the initial binding to one monomer of the dimer enhances the affinity for the second monomer.

Chemoreceptors can either mediate chemotaxis, have alternative cellular functions or are responsible for type IV pili mediated motility (Wuichet and Zhulin, 2010). To determine McpK function, we generated a mcpK deletion mutant and carried out quantitative capillary chemotaxis assays. Initially, control experiments were conducted to assess chemotaxis of the wt and mutant strain toward casamino acids. The choice of this chemoattractant was based on the fact that the chemotaxis toward amino acids is mediated by the three well-characterized chemoreceptors PctA, PctB and PctC (Taguchi et al., 1997; Rico-Jimenez et al., 2013). As shown in Supplementary Figure S3, deletion of mcpK did not alter significantly chemotaxis to casamino acids, indicating that this mutation did not cause any general motility defect.

Subsequently, chemotaxis assays toward different αKG concentrations were performed. Experiments showed that the wt strain has significant taxis to αKG concentrations between 5 μM and 5 mM, with an optimal response at 500 μM (Figure 5). Importantly, the deletion of the mcpK gene resulted in a drop in the chemotactic response to close to baseline levels. Complementation of this mutant by the in trans expression of the mcpK gene restored wild type like chemotaxis at all concentrations (Figure 5). In order to determine whether the observed chemotaxis defect in the mcpK mutant is the result of an altered metabolism, we performed growth curves in M9 minimal medium supplemented with αKG and succinate as carbon sources. As shown in Supplementary Figure S4, there were no differences between the wt and mutant strain. Taken together, these data show that McpK is the primary chemoreceptor for αKG in P. aeruginosa PAO1.

We have recently assessed the effect of chemoeffectors on the gene expression of their cognate chemoreceptors in P. putida KT2440 (López-Farfán et al., 2016). We were able to show that the expression of a significant number of chemoreceptor genes is either up- or downregulated by the cognate chemoeffectors. However, this was not the case for genes encoding chemoreceptors that respond to TCA cycle intermediates, namely mcpS, mcpQ and mcpR that were expressed independently of the presence or absence of their cognate ligands.

To assess mcpK expression, we carried out real-time quantitative PCR measurements. Figure 6A shows mcpK transcript levels of samples taken during mid-exponential phase in comparison to those of the housekeeping genes gyrB and rpoD as well as to genes encoding functionally characterized chemoreceptors such as the amino acid sensors PctA and PctC (Taguchi et al., 1997), the Pi responsive CtpH (Wu et al., 2000) or the TlpQ chemoreceptor that mediates taxis to ethylene (Kim et al., 2007). Data show that mcpK transcript levels are low as compared to the housekeeping genes and in between most and less abundant chemoreceptor transcripts.

In subsequent experiments we assessed the effect of αKG on mcpK expression. However, the addition of αKG to P. aeruginosa cultures grown in M9 minimal medium supplemented with glucose did not alter mcpK expression (Figure 6B). Data thus show that, in analogy to the TCA cycle intermediate responsive chemoreceptors in P. putida KT2440, the cognate chemoeffector αKG does not modulate mcpK expression. Further experiments will show whether the constitutive expression of TCA cycle intermediate responsive chemoreceptors, as observed in P. putida and P. aeruginosa, is a general feature.

Pseudomonas aeruginosa is a universal pathogen that is also able to colonize and infect different plants (Rahme et al., 2000; Cao et al., 2001; Walker et al., 2004; Attila et al., 2008). The web-based resource PIFAR allowed the identification of 175 gene products in P. aeruginosa putatively involved in the interaction with plants (Martinez-Garcia et al., 2016) and the TlpQ chemoreceptor was found to mediate chemotaxis toward the plant hormone ethylene (Kim et al., 2007). αKG is present at significant levels in plant root exudates (Tawaraya et al., 2014; Ganie et al., 2015) and chemotaxis to root exudate components was shown to be essential for efficient root colonization (de Weert et al., 2002; Scharf et al., 2016).

To determine the role of McpK in the colonization of the rhizosphere, we performed competitive colonization assays using maize as model plant. Initial experiments showed that P. aeruginosa colonizes the maize rhizosphere at a density of around 5 × 10⁷ bacteria per gram of root. To distinguish between the wild type and the mutant strain, a kanamycin-resistant P. aeruginosa wild type strain was generated. This strain contains a kanamycin cassette inserted downstream of the glucosamine-6-phosphate synthetase encoding gene, glmS. This region was demonstrated to be neutral in multiple Pseudomonas strains, including P. aeruginosa (Koch et al., 2001; Matilla et al., 2007). The resulting strain, P. aeruginosa PAO1-Km, was shown to mediate chemotaxis to αKG (and other known chemoattractants) at the same levels as PAO1 (Supplementary Figure S5). Competitive colonization assays showed that the fitness of the mutant in mcpK in the rhizosphere was similar to the strain PAO1-Km (Figure 7). Additionally, the strain ΔmcpK also colonized root tips at the wild type levels (Figure 7).