The bacterial strains and plasmids used in this study are listed in Table 1. The primers used are listed in Table 2. Bacteria were grown in Luria–Bertani broth (LB): 1% tryptone (Difco Laboratories Inc., United States), 0.5% yeast extract (Difco Laboratories Inc., United States), and 0.5% sodium chloride (Merck, Germany). LB was also used for bacterial soil inoculation and molecular methods (i.e., bacterial DNA extraction and plasmid purification). For the glass attachment assay and wheat or cucumber root extracts media, 50% LB media without sodium chloride was used. Solid media were prepared by adding 1.5% Bacto^TM agar (Difco Laboratories Inc., United States) to LB media. Where appropriate, antibiotics were added to maintain or select for plasmids as follows: for Escherichia coli, ampicillin (Ap, Calbiochem, United States) at 100 μg/mL and gentamicin (Gm, Sigma, United States) at 15 μg/mL and for all Pseudomonas isolates, Gm at 30 μg/mL.

All basic molecular techniques were executed according to standard protocols: DNA (from roots and isolates) was extracted using soil GeneAll kit (soil DNA production kit, GeneAll, South Korea) according to the manufacturer protocol. Plasmids were purified with the QIAprep Spin Miniprep Kit (Qiagen, Germany). All primers were obtained from Integrated DNA Technologies (IDT, United States), and PCR primers are listed in Table 2. For quantifying plant gene copy number by qPCR, we used the known DNA concentration of a specific plasmid containing the target region coding for translation elongation factor 1 (tef). In order to examine the ability to colonize wheat roots (as described below), the isolates were chromosomally labeled with the gfp or mCherry gene. Construction of chromosomally mCherry-expressing strains: pUCP18-miniTn7T2.1Gm-GW: Gm-mCherry (Zhao et al., 2013) was inserted into Pseudomonas isolates (NT0128, NT0133) together with pTNS2 helper plasmid using electroporation (Choi and Schweizer, 2006; Zhao et al., 2013; Marmont et al., 2017) as described by Tovi et al. (2019). All plasmids are listed in Table 1.

Wheat (Triticum turgidum cv. Negev, Hazera, Israel) or cucumber seeds (Cucumis sativus cv. Kfir, Zeraim Gedera, Israel) were prepared as described previously by Tovi et al. (2019). Briefly, wheat seeds were surface sterilized by soaking in 3% sodium hypochlorite for 1.5 min, followed by 70% ethanol for 1.5 min, then rinsing three times with water. The sterilized seeds were cultivated in sterile mix of sandy loam soil with perlite 9:1 (wt/wt), hydrated with half-strength Hoagland solution (Ofek et al., 2014). For soil inoculation, bacteria were grown in LB overnight, diluted 1:100 in LB, and further grown in LB for 3–4 h and inoculated at 10⁶ bacteria per gram of soil–perlite mixture. Plants were grown for 10–12 days, after which roots were carefully removed and rinsed in sterile saline, and soil adhering to the roots was removed by vortex. Roots were weighted and used for either DNA extraction and qPCR or for imaging bacterial colonization by CLSM.

Roots were stained with Hoechst 3334 (NucBlue, Thermo, United States) according to the manufacturer protocol. Images were acquired using either OLYMPUS IX 81 (Olympus Corporation, Japan) inverted laser scanning confocal microscope (FLUOVIEW 500) equipped with 405-, 488-, 515-, and 543-mm laser lines and a 20 × 0.7 NA UPlanApo objective or a Leica SP8 laser scanning microscope (Leica, Germany) equipped with a solid state laser with 405-, 488-, 514-, and 552-nm light, and HC PL APO CS2 20×/0.75 objective (Leica, Germany) and Leica Application Suite X software (LASX, Leica, Germany).

Wheat roots were prepared as described above. An average of 0.25 g of wheat root was used for DNA extraction. Each sample represents three different individual roots from the same pot, and a minimum of six different replicates (individual pots) of wheat plants were used for quantification. Bacterial abundance on roots was quantified using qPCR by targeting GFP, mCherry (for labeled isolates), and normalizing it to the copy number of the plant tef gene. In all samples, GFP and mCherry target numbers were divided by the tef target number, as described by Ofek et al. (2011). A plasmid standard containing the target region was generated for the tef gene as described previously (Ofek et al., 2011; Tovi et al., 2019). For GFP, mCherry, and tef quantification, the plasmids containing the target region listed in Table 1 were used as standard, and primers used for quantifying each gene are listed in Table 2. Each gene copy number was calculated using StepOne software v2.3 (Applied Biosystems, United States), using the known DNA concentration and the specific plasmid plus the molecular weight of the insert, estimated from their lengths (Table 2). All qPCR assays were conducted in polypropylene 96-well plates and StepOnePlus real-time PCR system (Applied Biosystems, United States). Plasmids DNA concentrations were measured using Qubit fluorometric quantification (Thermo Fisher Scientific, United States) and Qubit dsDNA BR assay kit (Thermo Fisher Scientific, United States). Sevenfold dilution series of the plasmids containing the target genes were conducted within a range of 10⁸ copies per 1 mL to 10² copies per 1 mL. The standards and each sample within each treatment were tested in triplicate. The slope of the standard curve, correlation coefficient, and amplification efficacy were calculated using StepOne software v2.3. Each 20-μL reaction contained 10 μL Absolute Fast SYBR^® Green Master Mix (Thermo, United States), 0.6 μL of each primer (100 μM), 7.8 μL H₂O, and 1 μL template DNA (diluted 1:10). PCR conditions were as follows: 5 min at 95°C, followed by 40 cycles of 95°C for 5 s, 60°C for 30 s. Melting curve analysis of the PCR products was performed following each assay to confirm that the fluorescence signal originated from specific PCR products.

Wheat or cucumber roots were washed of all soil particles. The washed roots were shaken in 0.25% sodium chloride for 2 h. In order to prepare a particle free extract, the roots were centrifuged (4,500 revolutions/min, 10 min, 22°C), and the supernatant was filtered through 0.22-μm Autofill Filtration System PES (Foxx, Life Science, United States) and kept at −80°C. When needed, 50% LB without sodium chloride was added to the filtered supernatant (termed here “root extract”) for the preparation of root extract containing medium that was sterilized by filtration.

gfp- or mCherry-labeled isolates were grown in LB overnight and then diluted 1:100 and left to grow up to approximately 0.4 OD in 3 mL LB. The bacteria cultures were diluted again to 1:100,000 and left to grow overnight in 50% LB or in cucumber or wheat root extract media. The overnight cultures were centrifuged, and the pellets were kept at −20°C until used for genomic DNA extraction. DNA was extracted and diluted to 20 ng/μL for each qPCR reaction. GFP or mCherry copy numbers were calculated as described above. p value was calculated by Mann–Whitney U test by comparing the monocultures (triangle) to co-cultures (circle) strains.

gfp- or mCherry-labeled isolates were grown in LB media overnight, diluted 1:100, and left to grow up to approximately OD₆₀₀ = 0.4 in 3 mL LB. The bacteria cultures were diluted again to a final concentration of 1:100,000 in 1 mL of 50% LB or in cucumber or wheat root extract media in glass-bottom 12-well plates (Cellvis, United States). Image acquisition was performed using an Eclipse Ti-E inverted microscope (Nikon, Japan) equipped with a Plan Apo 40x/0.95 NA air objective. An LED light source (SOLA SE II, United States) was used for fluorescence excitation. GFP fluorescence was excited with a 470/40 filter, and emission was collected with a T495lpxr dichroic mirror and a 525/50 filter. mCherry fluorescence was excited with a 560/40 filter, and emission was collected with a T585lpxr dichroic mirror and a 630/75 filter (all filters from Chroma Technology Corp., Brattleboro, VT, United States). Images were acquired with an SCMOS camera (ZYLA 4.2 PLUS, Andor, United Kingdom) at 1-h intervals for a period of 17 h. NIS software (version 5.02, Nikon Instruments, Inc.) was used for acquisition and basic image processing.

Pair cross-correlation (PCC) function g(r) (Lotwick and Silverman, 1982) describes the spatial organization of two strains on a glass surface at several time points (Figure 4). In general, PCC quantifies the relationship between a pair of populations by examining the distances between their respective members. For a given pair of populations on a two-dimensional (2D) surface,

where p(r) is the probability of finding a pair of members of both populations at distance r, and D_i is the fraction of area covered by population i. The mean and confidence interval of g(r) can be calculated by combining the analysis results of several separate 2D areas (Steinberg et al., 2021). If the confidence interval of g(r) at distance r includes 1, then the two populations do not exhibit significant attraction or repulsion with respect to each other at distance r; if g(r) at distance r is significantly higher than 1, then the populations are clustered at distance r, and conversely, if g(r) at distance r is significantly lower than 1, then the populations are repelled at distance r. PCC analysis was performed by the 2D dipole algorithm, using DAIME software package (Daims et al., 2006). Mean and 95% confidence intervals were calculated by pooling the results of 2 × 2 sections of each image (total image area is 0.85 mm² at 0.16 μm/pixels). The analysis is performed at the pixel level; therefore, the measurements pertain to a continuous fraction of the population by area. The analysis parameters are a sample size of 500,000 dipoles and binning of distances at 0.96-μm intervals.

Statistical analysis was performed using the Mann–Whitney U test, and p < 0.01 was considered statistically significant.

To assess whether it is possible to predict co-cultures outcomes from monocultures behavior, we followed and examined spatial scales of interactions and abundance of three pairs of related Pseudomonas species. These strains have previously been shown to colonize roots differently when inoculated separately: NT0124 (tip colonizer) is specialized in efficient colonization of the root tips; NT0128 colonizes along the root; and NT0133 colonizes distant from the tip (Tovi et al., 2019). In the current study, we followed bacteria inoculation in pairs using two methods: qualitatively on root topography, using CLSM (Figure 1), and quantitatively, using qPCR analysis (Figure 2).

Wheat seedlings were grown in soil inoculated with the various pairs. The pattern of root colonization was followed in 10- to 12-day-old seedlings using CLSM. Remarkably, CLSM images of each pair that were taken from the coinoculated roots showed that the spatial distribution of each strain could be either affected or independent of the companion strain (Figure 1). Co-inoculation of isolates NT0124, the “tip bacterium” (shown in green) and NT0128 (in red) exhibited a root-wide colonization distribution: Both isolates colonized along the root, without apparent preference for a specific area (Figures 1A.1–A.3). When inoculated together with NT0128, the spatial distribution preference of the “tip bacterium” (NT0124) changed, and it was absent from the tip (Figure 1A.1). In contrast, Co-inoculations of NT0124 with NT0133 retained their original spatial preference, with each isolate occupying its preferred specific niche on the root (Figures 1B.1–B.3), as was the case when inoculated as single culture: NT0124 (shown in green) was localized mainly at the root tip (Figure 1B.1), and NT0133 (in red) was localized at the distant part (Figure 1B.2). Similarly, the pair NT0128 and NT0133 did not change its typical spatial distribution on wheat roots (Figures 1C.1–C.3): NT0128 (in red) with NT0133 (in green) colonized the same region: the zone distant from the tip (Figure 1C.3).

Bacterial abundance on the inoculated wheat roots was quantified by examining each strain separately and in pairs, using qPCR by targeting gfp or mCherry genes of each isolate. The plant gene tef (translation elongation factor), representing plant cell numbers, was quantified for normalization (Figure 2). Interestingly, qPCR results showed that even though the spatial distribution pattern of the pairs NT0124 with NT0133 and NT0128 with NT0133 did not change (Figure 1A), their abundance was affected significantly: NT0124 abundance on the roots decreased when inoculated with NT0133 (p = 0.006, Mann–Whitney U test, n = 7), whereas NT0133 abundance increased (p = 0.0006, Mann–Whitney U test, n = 7), compared to plants that were inoculated with each of these strains separately. Similarly, when NT0128 was inoculated on the roots together with NT0133, NT0128 abundance significantly decreased (p = 0.01, Mann–Whitney U test, n = 8), whereas NT0133 abundance was similar to its abundance as a monoculture. When NT0124 was coinoculated with NT0128, the abundance of NT0124 decreased (p = 0.005, Mann–Whitney U test, n = 6), whereas NT0128 abundance on the roots was not affected (Figure 2).

In order to rule out the effect of the physical and chemical properties of the plant roots on bacterial interactions and colonization, we studied monocultural and cocultural behavior on an inert surface (glass) when treated with either LB or root extracts. Bacterial surface colonization and attachment were examined by live imaging microscopy (Figures 3, 4). Pairwise interactions were also examined in liquid cultures, an environment wherein attachment to the surface is eliminated (see below). Live imaging was enabled following the dynamics of surface colonization, including attachment and aggregate formation on the glass surface, while monitoring the interactions over time. On the inert surface, in the presence of nutrients, the attachment of both monocultures, NT0124 and NT0128, was limited (Supplementary Figures 1A.1,A.2) (Supplementary Figures 1B.1,B.2). Under the same conditions, NT0133 was able to attach to the glass surface and to form aggregates (Supplementary Figure 1A.3). We also examined the ability of the three isolates to attach to glass in response to wheat or cucumber root extract (Supplementary Figures 1B.1–B.3,C.1–C.3). In the presence of wheat root extract, isolates NT0124 and NT0128 attached to the glass surface and formed micro-colonies (Supplementary Figures 1B.1,B.2, respectively) in a similar manner to that of NT0133 (Supplementary Figure 1B.3). In contrast, none of the isolates changed their attachment patterns to the surface in response to cucumber root extract (Supplementary Figures 1C.1,C.2).

We further followed bacterial colonization and aggregate formation dynamics on glass surface in the three pairs over time in the presence of wheat root extract. In the pair NT0124 and NT0128 (green and red, respectively; Figures 3A.1–A.5), both strains colonized the glass surface. Moreover, cells of isolate NT0124 appeared to attach to NT0128 aggregates from very early stages of their colonization, along all time points, regardless of aggregate size (Figures 3A.1–A.5). The pair NT0124 and NT0133 (green and red, respectively; Figures 3B.1–B.5) exhibited a differing pattern: NT0124 cells attached to NT0133 aggregates only after the latter grew bigger over time up until NT0133 covered the surface. In the pair NT0128 and NT0133 (green and red, respectively; Figures 3C.1–C.5), both colonized the glass surface and formed microcolonies without any observed interspecies preference.

To assess whether the spatial organization generated by each pair during the colonization process is indicative of interaction between the isolates, the PCC function g(r) was used (Figures 4A–C; see section “Materials and Methods” for more details). If g(r) at distance r is significantly higher than 1, then the two strains were colocalized at distance r more than expected by chance. If g(r) is significantly lower than 1, the two strains repel each other at distance r. Otherwise, if g(r) does not significantly deviate from 1, then the interaction cannot be distinguished.

The pair NT0124 and NT0128 exhibited the strongest co-localization at distances under 10 μm (Figure 4A.1). The pair NT0124 and NT0133 exhibited intermediate co-localization at distance of 5 μm (Figure 4B.1) up until NT0133 covered the surface (Figure 4B.3). In contrast, the organization of the pair NT0128 and NT0133 did not deviate from the expected by random at any range; therefore, no significant co-localization was observed. Thus, NT0128 and NT0133 spatially dependent interaction on the glass surface can be assumed to be weak (Figure 4C.1). In all three pairs, similar results were observed for later time points (Figure 4, panels 2–3).

Bacterial interactions were examined in liquid containing LB, wheat root extracts, or cucumber root extracts. Under these conditions, interactions are not related to surface attachment as they grow in planktonic mode rather than forming a biofilm. The abundance of each of the isolates in monoculture and in pairs was measured using qPCR (Figure 5). In LB cultures, population levels of all three pairs (NT0124 with NT0128, NT0124 with NT0133, and NT0128 with NT0133) were similar to their levels as monocultures (Figure 5, left-hand subplot). However, when grown in the presence of wheat root extract (Figure 5, right-hand subplot), in pairs, different interactions were observed. In a culture of NT0124 with NT0133, isolate NT0124 abundance decreased significantly (p = 0.001, Mann–Whitney U test, n = 9) compared to its growth alone. This interaction could be considered amensalism. On the roots, the interaction between these two strains may be considered antagonistic, as the population of NT0133 increased. Likewise, in co-culture NT0128 with NT0133, the abundance of NT0128 decreased significantly (p = 0.001, Mann–Whitney U test, n = 10). However, co-culture NT0124 with NT0128 may be explained as neutralism; that is, the growth of each isolate was unaffected by the other (p = 0.8, Mann–Whitney U test, n = 8). In a medium containing cucumber root extracts, the growth of all three strains in pairs was similar to their growth as monocultures (Figure 5, middle subplot).