Six plant species originating from different ecosystems were chosen for this study: Helianthemum sessiliflorum, the desert shrub host of the desert truffle Terfezia boudieri; Arabidopsis thaliana Col0, representing non-mycorrhizal plants; Vigna radiata (mung beans) and Lens culinaris (lentil), legumes that form a chemo-communication system with nitrogen-fixing bacteria (Rhizobia) and endomycorrhizal fungi; Zea mays (maize), representing monocotyledons; and Persea americana (avocado), representing a tree originating in tropical forests. Seeds of V. radiata, L. culinaris, maize, and avocado were purchased at a local market, and seeds of H. sessiliflorum were collected in April 2018 from the experimental plot at the Bergman campus, Ben-Gurion University of the Negev, Beer Sheva.

Plants were cultured under sterile, controlled conditions for analysis of polar and semi-polar compounds in root exudates as follows: seeds were surface-disinfected by immersion in 70% ethanol, rinsed three times with double distilled water (DDW), immersed in 1% NaOCl (containing a drop of Tween 20), and rinsed again with DDW. For H. sessiliflorum and P. americana, the seed coat was removed by gentle scraping with sandpaper prior to surface disinfection. H. sessiliflorum seeds were soaked in sterile DDW at 6°C for at least 24 h before sowing. Seeds were sown manually in Magenta GA-7 Plant Culture Boxes (bioWORLD) with sterile Perlite (particle size, 4; 78% under 2.4–4.8 mm/22% under 2.4 mm, Agrical) except for A. thaliana seeds that were mixed with 0.4% agar in the same medium. The growth medium was Murashige and Skoog (MS)-zero (Murashige and Skoog, 1962), named N5 containing (in mg/L): 0.025, CoCl₂; 0.025, CuSO₄; 36.7, Fe EDTA; 6.2, H₃BO₃; 0.83, KI; 16.9, MuSO₄; 0.25, Na₂MoO₄; 8.6, ZnSO₄; 332.02, CaCl₂; 170, KH₂PO₄; 380, KNO₃; 180.54, MgSO₄; 330, NH₄NO₃; 0.5, pyridoxine; 2.0, glycine; 100, myo-inositol; 0.1, thiamine HCl; and 0.5, nicotinic acid (pH 5.8; adjusted with 10% KOH). Avocado was grown with sterile DDW. The boxes were placed for 44 days in a controlled growth room under 24 ± 2°C, long daylight (16 h), and fluorescent light (TL-D 36W/54-765, cool daylight). Control magenta boxes contained the matrix without seeds.

Root exudates were collected by sequential immersion of the perlite in DDW. The perlite was removed by vacuum filtration on the filter paper (Whatman®, No. 1). The combined filtered solution was lyophilized, and the dry powder was dissolved in DDW to a concentration of 100 mg/mL and filtered through 0.22-μm filter prior to chemical and physical analysis.

The root exudate was analyzed in the presence of nucleosides and deoxynucleosides using ultra-performance liquid chromatography–mass spectrometry/mass spectrometry (UPLC-MS/MS). A mixture of nine synthetic nucleosides served as the standard.

The system included an ACQUITY UPLC® I Class Waters system with a Q Exactive, Thermo Scientific MS detector (Electrospray Ionization) equipped with ACQUITY UPLC®HSS T3 1.8-μm (2.1 mm × 100 mm) column. The mobile phase A was 0.1% formic acid (FA) in DDW (ULC/MS), and mobile phase B was 0.1% FA in acetonitrile (ULC/MS). Gradient sequence was as follows: 0 min to 1 min, 99%–95% A and 1%–5% B; 1 min to 8 min, 95%–1% A and 5%–99% B; 8 min to 10.9 min, isocratic 1% A and 99% B; 10.9 min to 11 min, 1%–5% A and 99%–95% B; 11 min to 13 min, 5%–99% A and 95%–1% B; and 13 min to 15 min, isocratic 99% A and 1% B, at a flow rate of 0.4 mL/min. The injection volume was 5 μL, and the column and sample tray temperatures were 40°C and 10°C, respectively. MS scan range was 100–1,500 m/z in both positive and negative modes, and data were analyzed using Thermo Xcalibur™ 4.0 software (Thermo Fisher Scientific). Identification of the synthetic standards’ mass to charge ratio (m/z) was done by calculating their theoretical m/z value, identifying the specific m/z’s retention time (RT), and comparing the m/z readings of both individual standard solutions and mixed solution ( ± 0.001) and comparing to Thermo Fisher Scientific mzCloud Mass Spectral Library™. In addition, extracted ion chromatograms were generated for each compound in both the standard and sample solutions to compare the peak profile as another layer of confidence.

Bacteria including B. subtilis [strain 168, American Type Culture Collection (ATCC)], a motile model PGPB (Wadhams and Armitage, 2004; Pérez-García et al., 2011) B. pumilus b-13, Serratia marcescens, and Pseudomonas turukhanskensis were isolated from the rhizosphere of H. sessiliflorum. Xanthomonas campestris served as a model plant pathogen. Soil bacteria were stored in 25% glycerol (v/v) stock at −75°C and grown in Luria broth (LB) at 30°C. Escherichia coli K-12 MG1655 (that approximates wild-type E. coli) was employed in chemotaxis assay as a common/general strain. The E. coli was grown in Tryptone broth (TB) at 34°C for chemotaxis in-plug assay. To determine E. coli and B. pumilus b-13 growth when the carbon source was nucleoside, bacteria were grown in M9 minimal media supplemented with 1 mM 2-deoxy adenosine (2DA) substituting the glucose.

Bacterial colony was grown overnight in LB, at 30°C in a rotary shaker (250 rpm). The cell suspension was diluted 1:100 to LB medium and was grown; for Bacilli species, until it reached O.D. (O.D.)_600nm of 0.5–0.6; and for X. campestris, until optical density_600nm of 0.9–1.2. Bacteria motility was verified by a phase microscope (BH-2, OLYMPUS). These O.D. points were determined due to the highest observed motility of each strain. A solution of 5% glycerol and 0.5 M Na + lactate was added to a final volume of 1%, and the culture was incubated for 15 min with an occasional mixing. Cells were washed twice (centrifuged 3,200xg, 3 min, 25°C) and replaced with chemotaxis buffer (CB) containing 10 mM K₃PO₄ (pH 6.7), 0.14 mM CaCl₂, 5 mM Na + lactate, 0.3 mM (NH₄)₂SO₄, 0.05% (v/v) glycerol, and 0.1 mM EDTA (pH 7). For X. campestris, 1 mM MgCl₂ was added to the CB. The bacterial cultures were resuspended, centrifuged, and washed twice. The bacterial suspension was diluted in CB buffer to an O.D._600nm of 0.005. Chemotaxis was determined by a glass capillary assay (Adler, 1973). The assay included solutions of nucleosides in CB, positive control of amino acids (1 mM asparagine for B. subtilis 168 and B. pumilus b-13, and 5 mM methionine for X. campestris), and negative control (CB), in which methionine served as a positive control based on a previous study with Xanthomonas oryzae (Kumar Verma et al., 2018). A 60-μL solution was loaded into three one-end sealed capillary tube (100 mm × 1.2 mm, Marienfeld Laboratory Glassware, Lauda-Königshofen, Germany) under vacuum. Capillaries were positioned vertically in a 96-well plate, each well contained 270 μL of bacterial suspension and incubated for 30 min at 30°C. Then, each capillary was removed, externally washed with DDW, and broken into an Eppendorf tube to discharge the solution. A sample of 40 μL was serially diluted and from each dilution, and 100 μL was mixed with 3.5 mL of Top agar (0.8% agar, 0.8% NaCl, and 1% tryptone) and plated. The plates were incubated overnight at 30°C, and colonies were counted. The results are presented as colony-forming units (CFU) per capillary.

E. coli, S. marcescens, and P. turukhanskensis colonies were grown overnight in TB (1% Tryptone and 0.5% NaCl) at 34°C. Then, the cell suspension was diluted with TB and re-grown to O.D._600nm of 1 to obtain vigorous motility. Cells were precipitated (4,000 rpm, 6 min, 25°C) and washed with a solution of 10 mM K₃PO₄ (pH 7.2), 0.1 mM EDTA, and 0.001 mM methionine. The cell pellet was resuspended with 10 mL of the same buffer and mixed 1:1 with a buffer containing 0.8% agar and poured into plates containing three plugs of 2% agar, with or without 0.1 mM or 1 mM 2DG. The plates were allowed to solidify and then incubated at 30°C for 2 h to 4 h. Following incubation, the plates were inspected for turbid growth around the agar plugs and a circling clear zone, indicating bacterial chemotaxis toward the plug containing nucleosides.

Chemoattraction of B. pumilus b-13 to 2-deoxy-guanosine (2DG) in soil was tested in a plate assay [adapted from Paul et al. (2006), with major adjustments]. B. pumilus b-13 was grown overnight in LB, and the bacterial culture was diluted and re-grown from an O.D._600nm of 0.05 to ~0.8 (when bacterial motility was vigorous). The plates were prepared as follows: A solidifying medium with 6.6 mM MgSO₄, 6.05 mM (NH₄)₂SO₄, yeast extract (0.02 mg/mL), and 0.4% agarose in 110 mL of phosphate buffered saline (PBS), with or without 0.1 mM 2DG, was poured into 145-mm-diameter Petri dish. An inner circle of the agarose medium was cut out to leave an outer ring of 1-cm width. The plate was filled with sterilized dune sand (taken from the rhizosphere from which b-13 was isolated) moistened to 20% w/w with PBS. A bacterial sample of 10⁹ cells/0.1 mL was dripped onto a shallow pit in the center of the plate that was next incubated at 30°C, overnight. The following morning, sand samples of ~130 mg were taken from four spots at the perimeter of the sand layer (a radius of ~6 cm) of each plate and combined. Each sand sample was suspended in 1 mL of PBS and serially diluted, and an aliquot of 0.1 mL was plated for CFU counting. The results were normalized to soil weight.

All experiments were repeated twice or more, and chemotaxis experiments were carried with triplicate capillaries per treatment. Significant differences were calculated by one-way ANOVA, followed by Tukey’s range test (α ≤ 0.05).

Root-secreted nucleosides and deoxynucleosides were detected in all tested plants by targeted UPLC-MS/MS ( Table 1 ). Uridine, adenosine, and 2-deoxy-adenosine (2DA) were ubiquitous in all plants, whereas 2DG and thymidine were identified by MS2 in all samples except those from P. americana ( Table 1 ). These results correspond with previous reports for A. thaliana (Strehmel et al., 2014), monocotyledons (Triticum aestivum L.) (Warren, 2015), and the legume Phaseolus vulgaris L (Tawaraya et al., 2014). However, guanosine was not detected in A. thaliana as previously shown (Strehmel et al., 2014). In some samples, several nucleoside fragments were not detected by MS2, possibly due to low amounts that were below the detection threshold. Cytidine fragmentation occurred only in legume extracts, indicating its high secretion rate in these species ( Table 1 ). The relative abundance of 2-deoxy-cytidine was very low, as it was detected only in the MS1 (m/z, 228.0973 ± 0.001) of all samples except P. americana. These results suggest that nucleoside secretion is indeed a common phenomenon among plants.

To test whether nucleosides induce chemotaxis in soil-borne bacteria, a capillary assay was performed. The model soil bacteria were Bacillus subtilis 168 and Bacillus pumilus, which are PGPB, and Xanthomonas campestris, which is a generalist plant pathogen. The B. pumilus strain, termed b-13, was isolated from the rhizosphere of mycorrhized H. sessiliflorum plants. Both PGPB B. subtilis 168 and B. pumilus b-13 exhibited a positive chemotaxis toward nucleosides yet with different sensitivities and response trends. A positive correlation was observed between the amount of B. subtilis 168 cells accumulated within capillaries (expressed by CFU) and 2DG concentrations, up to the highest level of 15 mM ( Figure 1A ). The CFU values were significantly different between capillaries loaded with 5 mM, 10 mM, and 15 mM of 2DG with 1,160 CFU, 1,430 CFU, and 1,950 CFU per capillary, respectively. A non-significant difference between the CFU values of the control, with CB alone, and 1 mM 2DG implies that the threshold level for attraction ranges between 1 mM and 5 mM 2DG. Control capillaries showed the lowest number of cells with 363 CFU per capillary, whereas the CFU value of the positive control, 1 mM asparagine, was the highest (3,440 CFU per capillary).

To further verify that secreted nucleosides are indeed potent chemoattractants of soil bacteria, their effect on B. pumilus b-13 was tested. A bell-shaped chemotaxis response was observed for all nucleosides, yet its magnitude varied between compounds ( Figures 1B–E ). For 2DG, chemotaxis peaked at 1 mM with nine-fold higher colony counts than the control with 30,190 CFU and 3,933 CFU per capillary, respectively ( Figure 1A ). The higher and lower concentrations of 2DG (15 mM, 10 mM, 0.01 mM, and 0.001 mM) also induced chemotaxis, but to a lesser extent (11,680 CFU, 10,530 CFU, 12,180 CFU, and 2,800 CFU per capillary, respectively). Thus, the response of B. pumilus toward 2DG saturated between 0.1 mM and 5 mM. This bell-shaped response of B. pumilus b-13 differs from the concentration-depended, almost linear, response of B. subtilis 168. Additionally, B. pumilus b-13 CFU values were nine-fold higher compared to those of B. subtilis 168, indicating that the former has superior motility.

Similar to 2DG, thymidine induced a bell-shaped chemotactic response in B. pumilus ( Figure 1B ), peaking at 1 mM with cell counts 10-fold higher than the control (23,370 CFU and 2,313 CFU per capillary for both, respectively). Lower or higher thymidine concentrations of 10 mM and 0.1 mM yielded 10,190 CFU and 6,608 CFU per capillary, whereas the lowest levels of 0.01 mM and 0.001 mM were similar to controls (data not shown). In assays employing 2DA and guanosine ( Figures 1D, E ), the chemoattraction was observed at 0.1 mM with nine-fold higher colony counts compared to the control for both nucleosides (33,400 and 3,943 for 2DA, and 36,400 and 3,824 CFU per capillary for guanosine). Insignificant reduction in colony counts was observed for 1 mM 2DA (27,350 CFU per capillary) and 0.01 mM guanosine (27,060 CFU per capillary), indicating that maximal chemoattraction occurs between 0.1 mM and 1 mM for 2DA and between 0.01 mM and 0.1 mM for guanosine. A comparison between B. pumilus’ response to various nucleosides showed that, while 2DG and thymidine evoked maximal chemotaxis at 1 mM, its sensitivity to 2DA and guanosine was higher, with a response already peaking at 0.1 mM.

To test whether phytopathogens can exploit nucleoside signaling as well, Xanthomonas campestris, a Gram-negative, motile generalist pathogen was chosen (Ryan et al., 2011). A positive correlation was discovered between nucleoside concentrations and the number of attracted X. campestris cells ( Figures 2A–C ). The data in Figure 2A demonstrate a positive chemotactic response of X. campestris toward 2DG although to a lesser extent than that of B. pumilus b-13 ( Figure 1B ). The most robust chemotactic response was observed at 25 mM 2DG, with an average of 2,986 CFU per capillary ( Figure 2A ). Significant chemotaxis occurred at the lower 2DG concentrations of 0.1 mM, 1 mM, and 10 mM, with 1,095 CFU, 1,494 CFU, and 1,630 CFU per capillary, compared to the control (507 CFU per capillary). Methionine that served as a positive control based on a previous study with Xanthomonas oryzae (Kumar Verma et al., 2018) showed slightly higher (136%) insignificant chemotaxis compared to the control (796 CFU per capillary).

A similar trend of positive correlation between chemotactic response and nucleoside concentration was observed with 2DA and thymidine ( Figures 2B, C ). For both nucleosides, chemotaxis peaked at 25 mM with 3,585 CFU and 2,706 CFU per capillary, whereas the control CFU counts were lower by 88%. Lower nucleoside concentrations and the methionine positive control showed 42%–62% more CFUs compared to CB controls. The presented data demonstrate that nucleosides induce positive chemotaxis in both pathogenic and PGPB, although the sensitivity of X. campestris was significantly lower compared to B. pumilus b-13.

To determine whether nucleosides can induce chemotaxis in rhizobacteria as well as in a warm-blooded originating bacterium, we tested the chemoattraction response of Serratia marcescens, Pseudomonas turukhanskensis spp. and E. coli in a plate assay ( Figure 3 ). The clearing zones around the agar discs containing 0.1 mM 2DA indicate that all three bacteria are positively chemoattracted to 2DA. It is interesting to note that the clearing zones at a concentration of 1 mM 2DA were smaller than that in agar discs containing 0.1 mM 2DA (data not shown), corresponding to the bell-shaped chemotactic response observed in the capillary assay.

To test whether nucleosides can diffuse to substantial distances and evoke chemotaxis under conditions as close as possible to natural environments, a soil plate assay was established. The plate contained an outer agar ring with or without 0.1 mM 2DG, and the rest was filled with wet dune sand. B. pumilus b-13 was placed at the center of the plate, and, following an overnight incubation, sand samples were taken from the edge of the plate ~ 6-cm radius for CFU counts. The assay shows that B. pumilus can traverse 6 cm of wet soil according to 2DG chemical gradient ( Figure 4 ). The number of CFU/g sand in 2DG containing plates was significantly higher (5.1 × 10⁵) than that in the control (3.8 × 10⁵). The results show that root secreted nucleosides at biological concentrations presented here can direct bacterial movement toward plants under soil conditions.

Nucleosides are considered poor nutritional source for some Bacilli (Schuch et al., 1999), implying that, in the rhizosphere, they may function as signaling chemoattractant molecules rather than a carbon source. Bacilli comprises a substantial portion of the rhizosphere bacterial community (Satish et al., 2022); therefore, it was important to determine whether 2DA serves as nutrient or signaling molecule. Growing B. pumilus and E. coli in M9 medium with glucose or 2DA as the carbon source revealed that the growth rate of both bacteria was inhibited ( Figure 5 ). E. coli uptake nucleosides by several transporter proteins (Patching et al., 2005); however, the average inhibition of growth rate in the presence of 2DA was 30% compared to glucose. B. pumilus growth rates were similar on both media up to 4 h, and, after that, the growth rate on 2DA declined to 40% compared to glucose at 10 h. The data indicate that, for B. pumilus, 2DA is a signaling chemoattractant and not efficient carbon source to support bacterial growth.