All bacterial strains and plasmids used in this study are listed Supplementary Table S1. Wild-type A. caulinodans ORS571 and its mutant derivatives were grown in L3 minimal medium (Jiang et al., 2016a), TY medium or on agar plates at 37°C containing 25 μg ml^-1 nalidixic acid and 100 μg ml^-1 ampicillin. E. coli was grown in LB broth or on LB agar plates at 37°C with the following concentrations of antibiotics as required: 50 μg ml^-1 kanamycin, 100 μg ml^-1 ampicillin, 50 μg ml^-1 gentamycin, 10 μg ml^-1 tetracycline.

A tlpA1-deficient strain was constructed as follows: A 693-bp fragment spanning a region immediately upstream of tlpA1 and the 5′ end of the gene (Supplementary Figure S1A) was amplified by PCR from the genomic DNA of A. caulinodans ORS571 using primer pair TlpA1UF and TlpA1UR (Supplementary Table S2). A 653-bp fragment encompassing 501-bp at 3′ end of tlpA1, the stop codon and a 152-bp region immediately downstream of the gene (Supplementary Figure S1A) was amplified from the genomic DNA of the wild-type strain using the primer pair TlpA1DF and TlpA1DR (Supplementary Table S2). The 693-bp fragment was digested with KpnI and NdeI and ligated into KpnI and NdeI sites of the vector pCM351 (Marx and Lidstrom, 2002). One of the resulting recombinant plasmid was designated as pCM351::UF. The 653-bp fragment was digested with ApaI-SacI enzymes and cloned into ApaI-SacI sites of the recombinant plasmid pCM351::UF. The ligation mixtures were transformed into E. coli DH5α to recover individual clones. One of the resulting recombinant plasmids, designated as pCM351::UF::DF, was verified by sequencing. The recombinant plasmid pCM351::UF::DF was transferred into wild-type A. caulinodans ORS571 by tri-parental conjugation with a helper plasmid pRK2013 (Figurski and Helinski, 1979). Colonies were screened for loss of the recombinant plasmid and for double homologous recombination. Potential mutants were obtained from on TY plates by selecting for gentamicin-resistance (from pCM351 vector) and tetracycline-sensitivity (Marx and Lidstrom, 2002). The gentamicin gene was then deleted by introduction of plasmid pCM157 (Marx and Lidstrom, 2002). Mutations in the strains were verified by PCR with primer pair TlpA1F and TlpA1R. A confirmed mutation was named ΔtlpA1.

For complementation, the tlpA1 ORF with 737-bp upstream non-coding sequence was amplified by PCR using primer pair TlpA1comF and TlpA1comR (Supplementary Table S2). The amplified 2250-bp fragment was inserted into HindIII and BamHI sites of the broad-host-range cloning vector pBBR1MCS-2 (Kovach et al., 1995). The ligation mixture was transformed into E. coli DH5α. DNA sequences from individual clones were sequenced and one verified clone was designated as pBBR1MCS-2-tlpA1-com. This recombinant plasmid was then introduced into the tlpA1 mutant via tri-parental mating and the transformants were recovered by selection for kanamycin resistance. One of the resulting strains was designated as tlpA1-com.

To construct a cheA deletion mutant of A. caulinodans ORS571, a 696-bp upstream fragment (UF) was amplified by PCR using primer pair CheAUF and CheAUR (Supplementary Table S2), and a 675-bp downstream fragment (DF) was amplified by PCR using primer pair CheADF and CheADR (Supplementary Table S2). The upstream PCR product was digested with NsiI-NdeI and cloned into NsiI-NdeI sites of vector pCM351. The resulting plasmid was designated as pCM351::UF. The downstream PCR product was digested with AgeI-SacI and cloned into AgeI-SacI sites of pCM351::UF. The final plasmid pCM351::UF::DF was recovered from the ligation mixture by transforming into E. coli DH5α and verified by sequencing. The plasmid was transferred into A. caulinodans ORS571 by a tri-parental mating with the aid of a helper plasmid pRK2013. The gentamicin gene replaces the cheA gene by double homologous recombination in the potential mutant, which was then removed by introducing the vector pCM157 (Marx and Lidstrom, 2002). The cheA-deletion mutant was thus generated by deletion of an internal 2340-bp fragment (Supplementary Figure S1B) and was verified by PCR with primer pair CheAF and CheAR. The confirmed mutant was named ΔcheA.

Strains were grown overnight in TY medium containing 25 μg ml^-1 nalidixic acid and 100 μg ml^-1 ampicillin. Cultures were diluted either into L3 medium containing 10 mM succinate and 10 mM NH₄Cl or into TY medium to an initial OD₆₀₀ of 0.02 and grown at 200 rpm and 37°C. All data were depicted as means and standard deviations from three individual repetitions.

For the soft agar assay, overnight cultures were washed and resuspended in chemotaxis buffer (Jiang et al., 2016a). The cultures were then adjusted to an optical density at 600 nm (OD₆₀₀) of 1.0. To compare chemotaxis response to different carbon sources, 5 μl aliquots of cell suspensions of the wild-type, mutant and complemented strains were inoculated on L3 soft agar plates (0.3% agar) containing 10 mM various carbon sources (succinate, citrate, proline, tartrate, malate, and glycerol) with or without 10 mM NH₄Cl. The inoculated plates were incubated for three days at 37°C.

A quantitative capillary chemotaxis assay was carried out as described in Reyes-Darias et al. (2016) with a few modifications. The wild-type and ΔtlpA1 strains were grown overnight in TY medium at 200 rpm and 37°C. The cells were then washed with chemotaxis buffer and the cell pellets were resuspended to a final density of OD₆₀₀ = 0.05. 200 μl of bacterial suspension mixtures (containing 100 μl wild-type and 100 μl mutant) were added into each well of a 96-well plate. Open ends of the capillaries containing chemotaxis buffer or 10 mM succinate as carbon source were then inserted into the wells. After incubation for an hour, the portions of the capillaries that were in contact with bacteria were rinsed with sterile water. The sealed ends were broken and the contents were transferred into microfuge tubes containing 1 ml sterile water. From each of these tubes, five individual 20 μl aliquots were placed onto TY plates and incubated at 37°C for 48 h prior to cell counting. Cells on TY plates were tested by PCR using primer pair TlpA1F and TlpA1R.

Total RNA was isolated from cultured free-living cells. cDNA was generated using GoScript^TM Reverse Transcription Kit (Promega) and diluted 10-fold (analysis of 16S rRNA expression) and 500-fold (analysis of the cheA gene expression) for subsequent PCR amplification. Quantitative PCR was performed using GoTaq^® qPCR Master Mix kit (Promega) with gene-specific primer pairs (TlpA1QF and TlpA1QR for tlpA1, CheAQF and CheAQR for cheA, 16SQF and 16SQR for 16S rRNA). Genomic DNA of ORS571 strain was used as template to construct a standard curve for determination of gene copy numbers. Transcript levels of the cheA gene were normalized against 16S rRNA levels. Real-time PCR was carried out on a Bio-Rad CFX96^TM real-time PCR system. The program for amplification included an initial denaturation step at 95°C for 2 min, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Specificity of the reaction was monitored by melting curve analysis.

Single colonies were inoculated into L3 liquid medium and incubated at 37°C and 200 rpm. The cell culture (OD₆₀₀ around 2.5) was washed once with L3 medium and resuspended in an equal volume of L3 medium. For biofilm formation assay, 1.5 ml L3 medium was added to each glass tube, into which 150 μl cell cultures was added and incubated at 37°C for 3–5 days. For Crystal Violet (CV) staining, cultures were removed carefully with a pipette, and the tubes were washed 3–5 times with ddH₂O and 1.6 ml of 0.5% CV was added into each tube. After incubation for 20 min, CV was removed and the tubes were washed five times with ddH₂O. Finally, 1.5 ml of 30% acetic acid was added to each tube to wash off the “CV ring”. The OD value of each tube was determined at 570 nm after CV staining.

Sesbania rostrata seeds were treated as described previously (Suzuki et al., 2007) with a few modifications. Seeds were immersed in concentrated sulfuric acid for 30 min to induce uniform germination, rinsed five times with sterile water, and soaked in sterile water for two days at 37°C in dark. Cells were grown in TY medium and harvested by centrifugation at mid-log phase (OD₆₀₀ of 0.5–0.6). The cells were washed thrice and resuspended in sterile chemotaxis buffer (Greer-Phillips et al., 2004) to an OD₆₀₀ of 0.5. The concentration of the cell suspensions (OD₆₀₀ = 0.5) was further verified by serial dilution and plating onto TY plates. 50 ml cultures of each strain were used to soak (50 rpm, 28°C) two day-old seedlings of S. rostrata for 4 and 24 h. The seedlings were then washed four times with sterile water at 100 rpm for 1 min each in a rotary bath. Equal-sized roots from five seedlings were then dispersed in sterile water by using a homogenizer. Serial dilutions from this suspension were spread onto TY agar plates and incubated for two days at 37°C. The adsorption pattern of root surface was examined by analysis of the wild-type and ΔtlpA1 colonies on TY plates. In competition adsorption experiments, surface-sterilized seedlings were co-inoculated with wild-type ORS571 and ΔtlpA1 strains at approximately 1:1 and 1:10 ratios for 4 and 24 h. Adsorbed bacteria were detected by PCR using the primer pair TlpA1F and TlpA1R. A non-inoculated flask was used as a control for all experiments.

Nodulation and competitive nodulation assays of S. rostrata were performed as previously described in Yost et al. (1998) with a few modifications. For nodulation assays on roots, surface-sterilized seedlings were co-inoculated with wild-type ORS571 or ΔtlpA1 strains at an OD₆₀₀ of approximately 0.5. For nodulation assays on stem, S. rostrata seeds were germinated and the seedlings were transferred to vermiculite in moistened pots containing a low-N nutrient (Callow and Vincent, 1971). After 3–4 weeks of plant growth, S. rostrata stems were inoculated with either the wild-type or ΔtlpA1 strains at an OD₆₀₀ of approximately 0.5. For nodulation competition assays on roots and stems, surface-sterilized seedlings and stems were co-inoculated with the wild-type ORS571 and ΔtlpA1 mixtures at approximately 1:1 and 1:10 ratios. All S. rostrata plants were grown in the greenhouse at 27°C. Nodules were harvested 3 weeks post inoculation and bacteria were re-isolated from root and stem nodules. Competitive abilities of the wild-type and ΔtlpA1 strains were analyzed by PCR using the primer pair TlpA1F and TlpA1R. Non-inoculated plants served as controls.

To measure symbiotic ARA, 50 root or stem nodules were weighed and immediately placed into individual 3 ml sealed tubes. 300 μl acetylene (C₂H₂) gas was added to each tube up to 10% (vol/vol) and incubated at 37°C for 4 h. 100 μl gas was taken out from each tube and analyzed using a gas chromatograph (Agilent Technologies 7890A). Nitrogen-fixing ability was expressed as μmol C₂H₄ produced h^-1 g ^-1 of fresh nodules. Each measurement was repeated at least three times.

We used MiST2 and SMART databases and identified a chemotaxis gene cluster in the genome of A. caulinodans ORS571. This cluster encodes homologs of known chemotaxis proteins, which we termed as che gene cluster (Jiang et al., 2016b). This unique che gene cluster contains cheA, cheW, cheY, cheB, and cheR genes with a chemotaxis receptor, which we named tlpA1 (transducer-like protein in A. caulinodans), located upstream of cheA and predicted to be transcriptionally co-oriented (Figure 1A). TlpA1 is predicted to be a ∼60 kDa protein. N-terminal domain of TlpA1 consists of two transmembrane helices (TM) and a periplasmic ligand-binding domain (LBD), while the C-terminal region consists of a HAMP and a highly conserved MA domain (Figure 1B).

Similarity searches using BLASTP revealed that TlpA1 is homologous to chemoreceptors from closely related alpha-proteobacteria such as Azorhizobium doebereinerae, Afifella, Ancylobacter, and Chelatococcu (Supplementary Figure S2A). Sequence similarity searches with the N-terminal periplasmic ligand-binding region of TlpA1 followed by multiple sequences alignment of related proteins showed that this region is found exclusively in the N-terminal variable extracellular regions of chemoreceptors from various distantly related bacterial species (Supplementary Figure S2B). The presence of a conserved LBD domain in various classes of chemoreceptors suggests a possible fitness advantage to bacteria in different environments. Interestingly, 30 other chemoreceptors of A. caulinodans ORS571 also contain two transmembrane regions in the N-terminal region, and HAMP and chemoreceptor signaling domains in their C-terminal regions as predicted by MiST2 and SMART. However, sequence similarity searches using BLASTP indicated that neither the entire TlpA1 protein sequence nor the N-terminal periplasmic region is homologous to the other 30 transmembrane chemoreceptors (Supplementary Figure S2).

To determine the role of the chemotaxis receptor TlpA1 in A. caulinodans ORS571, we generated strain containing a tlpA1 deletion mutant (named ΔtlpA1). Soft agar plates were used to assess chemotaxis towards carbon sources such as succinate, citrate, tartrate, malate, proline, and glycerol, which are strong attractants of A. caulinodans (Jiang et al., 2016a). The ΔtlpA1 mutant was significantly impaired (∼25 to 40% of wild-type) in chemotaxis towards all carbon sources tested, regardless of presence or absence of combined nitrogen (Figures 2A,B). We complemented the ΔtlpA1 strain with a plasmid carrying the wild-type tlpA1 gene. The chemotaxis defect of the ΔtlpA1 mutant was partially rescued by expression of the wild-type tlpA1 gene driven by its native promoter (Figures 2A,B).

Since the tlpA1 gene is located directly upstream of cheA and transcribed in same direction (Figure 1A), we determined whether deletion of tlpA1 affected the cheA promoter region and thus cheA gene expression. We thus compared cheA gene expression in the wild-type and ΔtlpA1 strains and found that cheA was expressed in both strains (Figures 3A,B). Furthermore, there was no difference in growth rates of the mutant and wild-type cells in TY medium (Supplementary Figure S3A) or in L3 minimal medium (Supplementary Figure S3B). These data suggested that the diminished diameters observed on the soft agar plate are related to defects in chemotaxis.

To further assess chemotaxis and motility of the ΔtlpA1 mutant, competitive quantitative capillary assays of the wild-type and ΔtlpA1 strains were performed. In this assay, capillaries filled with chemoeffector are immersed into a bacterial suspension, which causes cells to preferentially swim into the capillary filled with chemoattractant. Buffer without chemoattractant was used as a control (Figure 4A). After this assay, the number of colony forming units in the capillaries was determined. For competitive quantitative capillary chemotaxis assay, a bacterial suspension containing a mixture of the wild-type and ΔtlpA1 strains at an approximate 1:1 ratio was used. Figure 4A shows competitive quantitative capillary chemotaxis assay with buffer and succinate as the attractants for the wild-type and ΔtlpA1 mutant. The cell count of the ΔtlpA1 from capillaries filled with buffer was similar to that of the wild-type (about 1:1) (Figure 4B). This observation indicates that swim motility behavior of the free-swimming cells is not affected by the ΔtlpA1 mutant. However, cell count of the ΔcheA mutant from capillaries filled with buffer was significantly less than that of the wild-type cells (Supplementary Figure S4). With succinate as the attractant, the wild-type cell count was almost twice that of the ΔtlpA1 mutant (Figure 4B). This data shows that the ΔtlpA1 mutant is impaired in chemotaxis and that TlpA1 is essential for chemotaxis but not for cell motility in A. caulinodans ORS571.

Chemosensory systems have been previously implicated in regulation of biofilm formation (Hickman et al., 2005). To test whether deletion of tlpA1 affected biofilm formation, we compared biofilm formation ability of the ΔtlpA1 with the wild-type. Quantitative data reveal that biofilm formation in the ΔtlpA1 mutant is less than that in the wild-type (Figures 5A,B). These data indicate that TlpA1 may directly modulate biofilm formation under certain growth conditions.

To investigate the role of TlpA1 in symbiosis, the colonization ability on surface of roots was assessed by a quantitative assay. When inoculated individually, cell number of the ΔtlpA1 strain re-isolated from S. rostrata root surfaces was not significantly different from that of the wild-type (Supplementary Figure S5). This result suggested that the ΔtlpA1 is able to colonize on root surface. Surface-sterilized seedlings were co-inoculated with the wild-type ORS571 and ΔtlpA1 at approximately 1:1 and 1:10 ratios and cell numbers of the wild-type and ΔtlpA1 strains on the root surface of S. rostrata seedlings were determined after inoculation. The results indicated that colonization efficiency of the ΔtlpA1 mutant was significantly reduced when compared to the wild-type strain (Figure 6). These data therefore suggest that the ΔtlpA1 mutant affects competitive colonization of host root surfaces.

Root colonization is an important step in the establishment of symbiosis between rhizobia and host plants (Benizri et al., 2001). To further test whether the ΔtlpA1 mutant is defective for symbiosis and nodulation inducing ability in host, the wild-type and ΔtlpA1 strains were inoculated individually or in combination onto roots and stems of S. rostrata. The nodule occupancy of each strain was then determined. As shown in Supplementary Figures S6A–C, the ΔtlpA1 mutant induced functional nodules on roots and stems when inoculated alone, with no differences in nodule number and morphology as compared to the wild-type. Leghemoglobin of the stem nodules induced by the ΔtlpA1 mutant was orange–brown 20 days post inoculation (Supplementary Figure S6B). This result indicates that TlpA1 does not affect leghemoglobin production of the host plant in mature nodules. Likewise, nitrogenase activities of the root and stem nodules formed by the ΔtlpA1 strain were similar to the wild-type (Supplementary Figure S7). When we tested the ability of the mutant to induce nodule formation on roots in competition with the wild-type, we found that it was impaired in its ability to form nodules (Figure 7). The ΔtlpA1 mutant could not successfully compete for nodule formation with the wild-type even at 10-fold excess. We also tested the ability of the mutant to form stem nodules in competition with the wild-type. The results were similar to those obtained for root nodules (Figure 7). Taken together, these results indicate that the TlpA1 chemoreceptor is implicated in the processes related to competitive nodulation.