Chitinophaga pinensis isolated from sorghum root endosphere, as well as Terrabacter sp. and C. rhizosphaerae isolated from soil where sorghum was growing, were used for inoculation in this study. These bacteria have been used in a previous study (Chai et al., 2021). The draft genome sequences and gene annotations of these bacteria are available through the IMG portal at the Joint Genome Institute under the taxon ID 2818991442, 2818991454, and 2818991462, for C. pinensis, C. rhizosphaerae, and Terrabacter sp., respectively.

A sweet sorghum variety, Grassl, was used throughout this experiment (Boyles et al., 2019). Grassl seeds were surface-sterilized for 6 h with chlorine gas generated by adding 3.3 ml of hydrochloric acid to 100 ml of sodium hypochlorite in a desiccator. Surface-sterilized seeds were then washed with sterile water and plated on YPD medium (Costanzo et al., 2001) to verify that there were no bacteria on the seed surface. The potting mix used in the greenhouse experiment consisted of two parts of peat and one part of vermiculite. To sterilize the pot and potting mix, 325 g of the potting mix were added to a pot with a diameter of 12.7 cm and autoclaved three times. After autoclaving, the potting mix was plated on YPD to ensure there were no viable microbes.

All bacteria were grown in R2A broth (Reasoner and Geldreich, 1985) except for C. rhizosphaerae, which was grown in peptone-yeast extract broth (Hottes et al., 2004) because it did not grow well in R2A. Two days before planting, each bacterial strain was grown on a rotary shaker at 180 rpm at room temperature (24°C). After a day of growth, a portion of each liquid culture was transferred to a fresh medium to allow for continued growth. On the day of the experiment, each bacterial culture was pelleted at 4,000 rpm for 10 min and resuspended in phosphate-buffered saline (PBS, 8 g/L NaCl, 0.2 g/L KCl, 1.44 g/L Na₂HPO₄, and 0.24 g/L KH₂PO₄). The optical density (OD) of each of the bacterial/PBS suspensions was measured at 600 nm with a spectrophotometer and adjusted to an OD₆₀₀ of 1 that corresponded to 10⁹ colony forming units (CFUs) for each of these bacteria before inoculation. The CFU number was derived by plating 200 μl of diluted bacterial cultures with an OD₆₀₀ of 1 on R2A medium.

Rhizosphere DNA was extracted using MagAttract® PowerSoil®DNA KF Kit (Qiagen) and root DNA using MagMAX™ Plant DNA Kit (ThermoFisher Scientific), with a KingFisher Flex Robot (ThermoFisher Scientific) following the manufacturer’s protocol. DNA concentration was quantified using QuantiFluor® dsDNA System (Promega) with CLARIOstar® Plus microplate reader (BMG LABTECH) following the manufacturer’s protocol.

Different primer pairs were used for each bacterium to provide adequate specificity for each of the three bacteria used in this study. These primer pairs were constructed from the corresponding genome sequence of each isolate (Table 1). Standard curves were constructed by serial dilutions of the genomic DNA of each bacterium from 10⁶ to 10¹ pg DNA μl⁻¹ using molecular grade water. The genome copy number of each bacteria was computed using their genome sizes, (C. rhizosphaerae: 5563326 bp, Terrabactor sp.: 4320267 bp, and C. pinensis: 8318214 bp), DNA molecular weight of 650 Da bp⁻¹, and Avogadro’s constant of 6.022 × 10²³. The detection limit of C. rhizosphaerae, C. pinensis, and Terrabacter sp. were 13, 7, and 7 genome copies, respectively. All qPCR was carried out using CFX Connect (Bio-Rad Laboratories Inc., Hercules, CA, United States) in a final volume of 10 ml, which contained 5 ml of Power Sybr Green PCR Master Mix (Applied Biosystems, Foster City, CA, United States), 0.5 ml of each of the forward and reverse primers (10 pM each), 1 ng of template DNA, and water. The same amplification conditions were used for all three bacteria with an initial incubation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 s, annealing at 60°C for 30 s, and extension at 72°C for 15 s. The specificity of amplification was determined using a melting curve analysis at the end of the amplification by ramping the temperature up to 95°C for 1 min followed by a 0.5°C s⁻¹ increment from 60 to 95°C. Three technical replicates were performed for each sample in the qPCR. The average of the three Ct values from the technical replicates was calculated and reported.

All the statistical analyses in this study were performed using R v3.6.0 (R Developmental Core Team, 2018). A one-way ANOVA was performed to determine whether the colonization of the inoculated bacteria (logarithm of bacterial copy number) in the rhizosphere and root endosphere was influenced by inoculation method. Tukey’s HSD post hoc pairwise comparison was then conducted to compare the mean difference between inoculation methods. These analyses were performed on both the greenhouse and field datasets.

Linear models were constructed using lm function to determine the changes in sorghum shoot and root dry biomass for each combination of inoculation method, bacterial strain, sampling time point, and degree of colonization (log copy number). Prior to model construction, root and shoot dry weight were power-transformed by 0.222 and 0.303, respectively, which were determined using boxcox function in “MASS” package (Venables and Ripley, 2002) to homogenize the residual variances. Backward selection was performed to eliminate the interactions that were not significant in affecting sorghum biomass from the global models. The marginal means for each treatment and strain combination were computed and subjected to Tukey’s HSD pairwise comparisons using the emmeans function in “emmeans” package (Lenth, 2021). Plots were generated using ggboxplot and ggplot function in “ggpubr” (Kassambara, 2020) and “ggplot2” package (Wickham, 2016), respectively.

To construct primer pairs specific for each bacterial isolate, we first mapped each genome sequence to the NCBI database to identify the genomic regions that were unique to each of the three bacterial isolates and not found in their close relatives. As a result, we identified the genomic regions that exhibited zero matches when searched using the BLAST alignment tool. Using these unique genomic regions, we constructed three primer pairs for these isolates (Table 1). We further confirmed the specificity of these primers on the targeted strains by performing specificity tests on their closely related isolates in our culture collection from sorghum in the same field, some of which have a perfect match (100% similarity) with our targeted bacteria in their full-length 16S rRNA regions (Table 2; Supplementary Table 1).

All three bacteria were detectable in the rhizosphere of the inoculated plants up to 8 weeks after planting (Figure 1). The colonization of C. rhizosphaerae in the rhizosphere was greater (10⁴–10⁵ copies per ng of rhizosphere DNA) when the seedling priming and soil drench method were used as compared to the seed coating approaches (10²–10³ copies per ng of rhizosphere DNA; Figure 1A). Similar trend was also found for C. pinensis where seedling priming and soil drench method promoted its colonization in the rhizosphere (Figure 1B). Prolonged seed coating for 12 h enhanced the colonization of C. pinensis (Figure 1B) but did not improve the colonization of C. rhizosphaerae in the rhizosphere (Figure 1A). The colonization of Terrabacter sp. in sorghum rhizosphere was consistent in all five inoculation methods; although its abundance was lower (10² copies per ng of rhizosphere DNA) as compared to the other two strains which reached as high as 10⁵ copies per ng of rhizosphere DNA (Figure 1C). C. pinensis and C. rhizosphaerae but not Terrabacter sp. were detected in the rhizosphere of the uninoculated control at week 8 after planting (Figures 1A,B).

C. pinensis was the only strain that could robustly colonize the root endosphere starting from week 6 after planting (Figures 2A–C). The colonization of C. pinensis in the root endosphere was greater when inoculated with seedling priming and soil drench as compared to the seed coating methods at week 6 after planting (Figure 2B). C. pinensis was detected in the root endosphere of the uninoculated plants at week 8.

Linear models were used to determine the changes in sorghum root and shoot dry biomass for each combination of bacterial strain, inoculation method, sampling timepoint, and degree of colonization (log copy number). Overall, all three bacteria exhibited a certain amount of root growth promotion (Figures 3A–C), with Terrabacter sp. being particularly stronger than the other two at enhancing root growth on week 6 after planting (Figure 3B; Table 3). The degree of growth-promotion from these bacteria was affected by the inoculation methods. Although significant root growth-promotion was measured when inoculating Terrabacter sp. with all five inoculation methods, the degree of growth enhancement was greater when the three seed coating methods were used (Figure 3; Table 3). On the other hand, greater root growth-promotion was detected when inoculating C. rhizosphaerae and C. pinensis with the seedling priming compared to other inoculation methods (Figure 3). In fact, root growth-promotion from C. rhizosphaerae was only detectable with seedling priming despite this effect being marginally significant. For C. pinensis, significant root growth-promotion was also observed with alginate coating and marginally significant for 12 h coating. No significant root growth-promotion was measured when inoculating C. rhizosphaerae and C. pinensis with soil drench method.

Among the three bacteria used, only C. pinensis and Terrabacter sp. exhibited significant shoot growth enhancement (Figures 4A–C; Table 4). Significant shoot growth-promotion from C. pinensis was measured when inoculated with seedling priming, alginate, and 12 h coating methods. Significant shoot growth-promotion was also observed when Terrabacter sp. was inoculated with the same seed coating methods but not the seedling priming.

Alginate and 12 h coating method were further tested in the field to assess their efficacy for delivering the three bacterial inoculants to sorghum rhizosphere under non-sterile conditions in which there would be competition from the native microbial communities. While C. rhizosphaerae and C. pinensis were detected in the rhizosphere up to 12 weeks after inoculation in the field, DNA copy numbers in the rhizosphere of the inoculated plants were lower as compared to that of the greenhouse experiment and not significantly different from the uninoculated control (Figures 5A,B). Terrabacter sp. was not detected in either of the sampling timepoints (Figure 5C). No significant improvement in shoot dry weight was measured for all bacteria and inoculation treatment combinations at both sampling timepoints (Figures 6A,B). Bacterial colonization in the root endosphere was not quantified due to the lack of biomass difference between the inoculated plants and the uninoculated controls.

In this study, the compartmental specificity of the three inoculated bacterial isolates was demonstrated, with C. pinensis being the only strain that robustly colonized both the rhizosphere and root endosphere, whereas the other two were only able to colonize the rhizosphere. This result was in agreement with our expectations and other studies (Bai et al., 2015; Maggini et al., 2019) that show inoculated bacteria tend to colonize the plant compartments from which they are isolated. The degree of plant growth-promotion from C. pinensis was not greater than the other two bacterial isolates, although endophytic colonization could theoretically allow bacteria to interact directly with the host plant and potentially deliver the plant growth-promoting effects more efficient than the bacteria in the rhizosphere (Santoyo et al., 2016). Nonetheless, the compartmental specificity of inoculants may be crucial in determining their downstream impact on plants. For example, root nodule colonization of rhizobia is crucial for enhancing the nitrogen nutrition in the host plant because the root nodule restricts the entry of oxygen that can inhibit biological nitrogen fixation (Lindström and Mousavi, 2020).

Seed inoculation is currently the most widely used approach for the introduction of bioinoculants because it is the most practical and cost-effective compared to other approaches (e.g., in-furrow inoculation; O’Callaghan, 2016). Nevertheless, we found that seed coating methods may not be suitable for the inoculation of Gram-negative bacteria, which may be due to their thinner cell wall structure that renders them vulnerable to desiccation in the seed coating process (Schimel et al., 2007). In accordance with our hypothesis, the seed coating methods were less effective in delivering the Gram-negative bacteria (C. rhizosphaerae and C. pinensis) to the sorghum rhizosphere compared to seedling priming and soil drench whereas the Gram-positive strain (Terrabacter sp.) could be delivered successfully with seed coating. Interestingly, we also observed increased colonization from C. pinensis but not C. rhizosphaerae with the 12 h seed coating method. Since C. pinensis is a root endophyte, we speculate that it may have colonized the seed endophytically during the longer seed coating process (Kandel et al., 2017), thereby enhancing its survival under desiccation. Similar results were also demonstrated in another study in which 12 h seed coating promoted the initial rhizosphere colonization of an endophytic biocontrol bacterium, Serratia plymuthica HRO-C48, on oilseed rape and enhanced the survival of this bacterium on seeds (Müller and Berg, 2007). Different inoculation methods may further affect the downstream impacts of inoculants on plants. For instance, seed inoculation of Gram-positive Bacillus strains on cowpea and mash bean has been shown to suppress root-infecting phytopathogens more effectively than soil drench (Dawar et al., 2010). The improved biocontrol abilities from these inoculants with seed inoculation may be attributed to the fact that seed inoculation allowed them to establish inside the root prior to pathogen infestation. On the other hand, soil drench has been shown to be more effective in delivering inoculants to Italian ryegrass growing on soil contaminated with diesel oil than the 12 h seed coating method, and improved plant growth-promotion and hydrocarbon degradation (Afzal et al., 2012). This was probably due to the greater density of inoculant being applied from the soil drench than the 12 h coating method in which the amount of bacteria applied to seeds was constrained by the seed size. These findings suggest that it may be important to tailor inoculation methods for inoculants with specific characteristics and functionalities.

Although all the inoculation methods tested successfully delivered bacterial isolates to the sorghum rhizosphere in the greenhouse study, the impacts on sorghum growth were variable. Despite being equally effective at delivering C. pinensis to sorghum rhizosphere, no growth-promotion was observed with the soil drench method while both shoot and root growth promotions were detected with seedling priming. Similarly, Burkholderia ambifaria MCI 7 was reported to improve maize growth when coated on seed but was detrimental to growth when applied into the soil (Ciccillo et al., 2002). Although it was unclear why the same bacterium had contrasting effects on plant growth, the authors noted that the root adjacent to the stem was mainly colonized when the seed was coated whereas the entire root system was colonized with the soil drench method (Ciccillo et al., 2002). In our studies, Terrabacter sp. was in lower abundance in the rhizosphere but was still able to promote sorghum root growth comparable to or better than the other two bacteria. This is in line with another study that observed similar biomass-promoting effects on banana from Pseudomonas fluorescens Ps006 and Bacillus amyloliquefaciens Bs006 despite P. fluorescens Ps006 being a less efficient root colonizer compared to B. amyloliquefaciens Bs006 (Gamez et al., 2019). These findings highlight that the degree of colonization by inoculated bacteria of host plants may not necessarily be directly related to the level of growth enhancement induced.