Defined synthetic microbial communities colonize and benefit field-grown sorghum

Abstract

The rhizosphere constitutes a dynamic interface between plant hosts and their associated microbial communities. Despite the acknowledged potential for enhancing plant fitness by manipulating the rhizosphere, the engineering of the rhizosphere microbiome through inoculation has posed significant challenges. These challenges are thought to arise from the competitive microbial ecosystem where introduced microbes must survive, and the absence of adaptation to the specific metabolic and environmental demands of the rhizosphere. Here, we engineered a synthetic rhizosphere community (SRC1) with the anticipation that it would exhibit a selective advantage in colonizing the host Sorghum bicolor, thereby potentially fostering its growth. SRC1 was assembled from bacterial isolates identified either for their potential role in community cohesion through network analysis or for their ability to benefit from host-specific exudate compounds. The growth performance of SRC1 was assessed in vitro on solid media, in planta under gnotobiotic laboratory conditions, and in the field. Our findings reveal that SRC1 cohesion is most robust when cultivated in the presence of the plant host under laboratory conditions, with lineages being lost from the community when grown either in vitro or in a native field setting. We establish that SRC1 effectively promotes the growth of both above- and below-ground plant phenotypes in both laboratory and native field contexts. Furthermore, in laboratory conditions, these growth enhancements correlate with the transcriptional dampening of lignin biosynthesis in the host. Collectively, these results underscore the potential utility of synthetic microbial communities for modulating crop performance in controlled and native environments alike.

Introduction

In nature, microbes associated with plants and other eukaryotic organisms self-assemble into communities with remarkable genetic and physiological diversity, often impacting their hosts in complex ways based on forces and factors that are not yet fully understood [1–4]. Synthetic communities (SynComs), in which strains are assembled together following novel compositional formulas that either mimic or build upon patterns observed in nature, can be used in multiple ways to explore these forces and to reveal additive and synergistic properties that emerge from these communities as they interact with their host. For example, SynComs can serve as testbeds for exploring hypotheses regarding the molecular underpinnings of plant-microbe and microbe-microbe communication [5], or alternatively to explore the evolution of the microbiome structure and function over time. Additionally, they can also be used to augment the colonization ability of individual strains, and to support the manipulation of complex phenotypes in their hosts that are challenging to produce with single organisms [6–8].

Whereas interest in the use of rhizosphere engineering to support plant performance in agricultural contexts has grown in recent years, development of products capable of robust phenotypic change in field environments has proven challenging. Many such amendments comprised of single isolate strains have appeared promising in laboratory conditions, where competition from native microbes and environmental perturbation are minimal. However, upon testing in a native setting, both colonization efficiency and gained benefit are often lost [9–11]. The use of designed SynComs has been proposed as a promising alternative to single strain inocula [6] due to the potential for positive interactions within the community that bolster colonization of the host. However, additional experimentation is needed to explore plant-associated SynCom performance across a range of environments and the specific biotic and abiotic conditions necessary to foster beneficial impact on their host.

In this study, we aimed to create a defined synthetic rhizosphere community (SRC1) with the ultimate goal of evaluating 1) the degree to which growth conditions impact community composition of the community and 2) the impact of the community on host phenotype. The bioenergy crop Sorghum bicolor was chosen as the host for this work given the extensive sequencing-based characterization of its microbiome under a range of abiotic conditions and environments [12–16], and a recently developed bacterial isolate collection available for its root associated microbiome [17]. Using these datasets and culture collection, we selected a total of 57 strains for inclusion in SRC1 following two complementary but orthogonal guiding principles: 1) identification of strains related to more abundant and common hub taxa within the sorghum rhizosphere microbiome using network analysis, and 2) identification of strains capable of using sorghum-specific root exudate compounds as a carbon source. We assembled SRC1 and grew it in three environments (media-based, gnotobiotic plant-based, and native field-grown plant-based) and explored properties of community assembly in each context. We observed that whereas colonization by SRC1 was undetectable above background levels in the field environment, sorghum phenotypes in both growth chamber and field environment were positively influenced by SRC1 treatment. Additionally, we find evidence that SRC1 application leads to changes in lignin biosynthesis and immunity within the root, suggesting a potential role for these processes in the improved growth phenotypes. Collectively, these results demonstrate the potential of SynCom-based strategies for modulating agricultural productivity and demonstrate the need for an improved understanding of forces controlling the activity and persistence of plant growth effectors in native environments.

Materials and methods

Network analysis of 16S rRNA data for SRC1 formulation

Strains for the SRC1 formula were identified using two methods. First, we selected strains based on a network analysis of a previously published 16S rRNA dataset [12], which originated from an analogous field experiment conducted in 2016 at the same location as the current field experiments (described below). A subset of the previously published 16S rRNA dataset [12] that included only the watered rhizosphere samples across the complete growing season (n = 126) was processed as previously described to deliver an ASV occurrence table for downstream analysis that contained approximately 12 000 ASVs (Fig. S1). Prior to network analysis, the ASV table was filtered to include only ASVs found in at least 50% of samples (n = 1762), and then again to include ASVs found only in the top 50th percentile based on average abundance across all samples (n = 881). The resulting table was subjected to a Pearson Correlation analysis using CSS normalized count data and the cor function [18] in R-v4.2.1 [19], and ASVs with strong correlations (absolute abundance of greater than 0.8) were maintained in the list of candidate lineages (n = 440). A comparison of the taxonomic assignment of the remaining ASVs to an existing sorghum root isolate collection developed in house [17] produced isolate level matches at the genera level for approximately 70% of selected ASVs (n = 293). This list was further reduced to a subset of these strains for inclusion in SRC1 (n = 42) by down selection to better colonizing taxa based on relative abundance (Table S1, Fig. S2).

Isolation of exudate utilizing microbes for SRC1 formulation

Additional strains for the SRC1 formula were selected based on the ability to utilize sorghum exudates, such as sorgoleone, an allelopathic compound produced at high abundance within the root exudates of sorghum [20, 21]. To identify such microbes, field soil was collected from the Kearney Agricultural Research and Extension (KARE) Center in Parlier, California collected at the time when no sorghum was growing and stored at 4°C. MOPS Media Enhanced (MME) medium supplemented with 2 mM single sorghum exudate was utilized for initial enrichment of microbes capable of utilizing sorghum exudates as a sole carbon source (Table S1). Glass test tubes were used for all cultivation containing any sorghum exudate to minimize adherence of the substrate to the walls of the test tubes. Five mL of this medium was inoculated with 100 mg soil and incubated for 96 h. The initial enrichment culture was diluted 1:100 into fresh MME containing 2 mM sorghum exudate supplemented with Wolfe's vitamins and incubated for 72 h. Following cultivation, 100 μL of the enrichment culture was spread onto an MME agar plate containing 2 mM sorghum exudate and incubated for 96 h at 25°C. Well-separated, single colonies were picked and streaked onto an R2A agar plate for single colony isolation. Selected isolates were inoculated into MME containing 1 mM sorghum exudate. Isolates that appeared to utilize sorghum exudates were serially passed (1:100 dilution) 6 times into fresh MME containing 1 mM sorghum exudate. These cultures were further spread onto R2A agar plates, single colonies were picked, inoculated into R2A, and finally pure cultures were stored at −80°C. The 16S rRNA sequences of the isolates were determined via Sanger sequencing. A total of 15 additional strains were selected from this resource for inclusion in the SRC1 formulation (Table S1, Fig. S2).

SRC1 assembly and media-based assays

To generate SRC1 material for use in media-based experiments, we collected approximately the volume of one inoculating loop (~12 mg) of biomass per strain after each strain was grown on plates for 24–96 h. The incubation time for each strain was based on preliminary experiments designed to determine the time required to produce sufficient colony biomass for loop inoculation. A total of 57 strains were used to make SRC1, 42 strains from network analysis and 15 strains from the isolation of sorghum-exudate utilizing microbes, including some strains of the same species. The strains were plate-inoculated on the days indicated in Table S1. Each loop inoculation was then resuspended in 200 μL of sterile 1x Phosphate Buffered Saline (PBS). For each resuspended strain, 50 μL was collected and combined into a master mix of all 57 strains, 1 mg/mL of biomass per strain (SRC1 formulation). From this mixture, 500 μL was plated onto each of two agar plate types: SSM agar and SSM agar with 1:5 diluted sorghum exudates (SSM-Ex) collected as described above. The resulting SRC1-inoculated plates were grown at 20°C for one week, followed by biomass assessment and collection for passaging and 16S rRNA analysis (Fig. 1A). Biomass from each plate was resuspended in 2 mL of 1x PBS. The resuspension was used to make 10-fold and 100-fold dilutions. SSM and SSM-Ex plates were independently inoculated with 500 μL of each dilution from the first growth passage and from like condition (e.g., SSM-Ex dilutions on SSM-Ex plates). This incubation, collection, dilution, and reinoculation process was repeated for a total of eight passages.

Figure 1

SRC1 colonizes the rhizosphere and roots of sorghum and increases the shoot biomass. (A) Schematic diagram of the SRC1 cohesion and colonization analyses. SRC1 dilutions of 10- and 100-fold were plated on agar plates with synthetic soil media (SSM) or SSM with sorghum exudates (SSM-ex) and passed over eight weeks. 16S rRNA profiling was performed of at least two replicas across passages. Surface disinfected sorghum seeds (cultivar RTx430) were germinated and treated with mock or SRC1 in 5-L-capacity microboxes under a sterile environment. Plants were irrigated every three days (26 days) or drought stressed (no irrigation added, 11 days). Plant phenotyping, host transcriptome, and 16S rRNA profiling were performed at the time of harvest. (B) Heat map of the abundance of the SRC1 members identified across treatments in vitro and in planta experiments. The abundance is presented in log₂ scale for n = 174 samples for in vitro experiments and n = 154 samples for the in planta experiment. (C) Representative pictures of watered or drought stressed sorghum plants inoculated with mock or SRC1. (D) Shoot fresh weight. (E) Shoot dry weight. W = normal irrigation; D = drought stress. Letters above represent statistical differences among treatments by Brown-Forsythe and Welch ANOVA one-way, holm-Šídák post hoc test P < 0.05 of 18–20 plants per treatment (78 plants in total).

Lab-based in planta experiments

To explore SRC1 growth in planta, sorghum seeds from cultivar RTx430 were sterilized by 10% bleach (sodium hypochlorite) treatment for 15 min, washed five times with sterilized water, and germinated on sterilized wet filter paper at 30°C without light. To produce enough SRC1 inoculum, each strain was grown on TSA plates (15 g pancreatic digest of casein; 5 g peptic digest of soybean meal; 5 g sodium chloride; 15 g agar; autoclaved once at 120°C for 15 min) at 30°C according to its growth rate (Table S1). We collected approximately six inoculating loopfuls of growth for each strain and suspended in 1.2 mL of sterile 1x PBS. Cell suspensions were pooled and split into 50-mL beakers with 6 mL each. To heat-kill SRC1, we autoclaved the pooled cell suspension one time at 120°C for 25 min before inoculation. Microboxes of 5-L-capacity (Combiness, Nevele, Belgium) were used to create a sterilized environment for growth of the sterile seedlings and SRC1 inocula. Microboxes containing 1 kg of calcined clay (Sierra Pacific Supply http://www.sierrapacificturf.com) per box were autoclaved and placed into a Laminar Flow hood. A total of 400 mL of autoclaved water with Hoagland’s solution (1.6 g/L, Catalog No. H2395-10 L, Sigma, St. Louis, Missouri) was added into the calcined clay and mixed thoroughly. Our pre-sterilized germinated seeds were placed in each microbox following a 10 min incubation with either the SRC1 cell suspension, heat-killed SRC1, or mock treatment (1x PBS). Each seedling was transplanted 5 cm beneath the soil surface and an additional 1 mL of the SRC1 inoculum or mock treatment buffer was pipetted on top of each seed. Seedlings were then covered with soil. A total of 20 microboxes were used for each experiment: 5 or 4 replicates x 2 water treatments x 2 or 3 inoculation treatments (mock, heat-killed SRC1, and live SRC1). Plants were irrigated with 3 mL of sterile distilled water per plant three times per week (Fig. 1B). After two weeks the lids were removed, and the drought treatment was applied for 11 days. Growth conditions were 26–28°C, 16 h:8 h light/dark cycle, 50% humidity for 26 days before harvesting. At collection, shoot phenotyping was performed and roots were used for microbiome profiling and transcriptomics analysis.

Field-based in planta experiment

To explore SRC1 impact in field-grown plants, a field experiment was conducted at the Kearney Agricultural Research and Extension (KARE) Center in Parlier, California (36.6008°N, 119.5109°W). Sorghum cultivar BTx642 seeds were planted (June 9, 2022) within a block design and two different types of irrigation (drought stressed and watered plants) with three replicate blocks per treatment (Fig. S3). Prior to planting, all sorghum seeds were inoculated with either SRC1 or mock treatment. Sets of 200 seeds were incubated into 50-mL conical tubes with 15 mL of SRC1 or mock (PBS 1X buffer) for 45 min at room temperature and then dried for 3 h in a laminar hood. SRC1 preparation was carried out under the same conditions as described for the microbox experiments above, with one modification: we added 0.01% Tween 20 to 1x PBS as an adherent. Each block had four rows treated with mock (R2-R5) and four rows treated with the SRC1 (R6-R9). One block from each water treatment presented a low rate of germination, leaving only four blocks in total for the experiment. Five weeks after planting (WAP), a second inoculation (SRC1 or mock) was applied to 10 selected plants per row from each block, and these plants were labeled with numbered stakes. Plants of a similar size and with only one to two tillers were selected; inoculation consisted of 1 mL of SRC1 diluted into 200 mL of tap water. At 5 WAP, a drought treatment was imposed on one half of the blocks, consisting of a complete lack of irrigation until rewetting following the microbiome sample collection (Fig. S3). Watered blocks, which account for the remaining half of the blocks, received irrigation for the entire growing season beginning in the third WAP; no water was applied in the first three weeks to allow seedling establishment. At the end of the drought treatment (9 WAP), a total of 160 plants were sampled for microbiome profiling and plant phenotyping. These consisted of 5 plants inoculated from each treatment group (2 inoculation types x 2 inoculation times x 2 water treatments x 2 replicate rows x 2 replicate blocks). At 16 WAP, an additional 10 plants from each treatment group (n = 320) were bagged to protect the seed development from birds, and later harvested for shoot phenotyping and yield quantification at 25 WAP.

The amount of water applied to watered plants (3–25 WAP) and drought blocks (3–5 and 9–25 WAP) was 80% of calculated evapotranspiration each week. Daily potential evapotranspiration (ET_o) was determined from an on-site weather station located approximately 1200 m from the field site. A locally-derived crop coefficient was matched to the crop growth stage and multiplied by the ET_o values to determine a calculated daily estimated crop evapotranspiration (ETc) for a non-stressed grain sorghum crop. A drip system was utilized to apply all irrigation water during the growing season, consisting of drip lines placed on the surface of each furrow (0.76 m row spacing), with 0.3 m emitter spacing and 2 L/h emitter output. The drip system was only operated for part of one day out of each 7-day period, with treatments being irrigated for that period receiving an amount equivalent to 100% of the estimated ETc calculated for the prior 7-day period. Surface drip lines were used for irrigation to provide accurate water application amounts and a high level of water application uniformity. The irrigation treatment designed to be a non-water-stressed block was irrigated at 7-day intervals between the first within-season irrigation (June 9) and the final irrigation (December 1). Measured rainfall during the June 9 to December 1 period was less than 7.11 mm.

Sample collection

Plants from the lab-based in planta experiments were harvested at the end of the drought treatment. Plants were carefully placed and separated in a disinfected tray with new aluminum foil. Rhizosphere samples were collected using 8 mL of epiphyte removal buffer (0.75% KH₂PO₄, 0.95% K₂HPO₄, 0.1% Triton X-100 in ddH₂O; filter sterilized at 0.2 μM) with manual agitation and centrifuged to pellet the resulting rhizosphere soil at 3700 rpm for 10 min. Shoots and roots were separated with sterile scissors, then roots were washed manually in 8 mL of cold epiphyte removal buffer for 10 seconds and frozen immediately in liquid nitrogen. Frozen roots were divided into two parts and used for DNA and RNA extractions, respectively. Rhizosphere soil was centrifuged to pellet at 3700 rpm for 10 min. Root tissues and rhizospheres were stored at −80°C until nucleic acid extraction.

Plant samples from the field trial were collected by manually extracting whole plants with root systems using a shovel to a depth of approximately 30 cm. Root samples and rhizosphere samples (soil tightly adhering to the root surface) were collected for each plant by pooling multiple roots from that plant; a total of 20 roots and 20 rhizosphere samples per treatment type were collected in this way at 9 WAP. Roots were frozen on dry ice and then transported to the laboratory (all samples were collected between 9 a.m. and 1 p.m. during a single day). In the lab, roots were vortexed in 30 ml of cold epiphyte removal buffer for 5 min and centrifuged to pellet the resulting rhizosphere soil at 3700 rpm for 10 min after removal of the root tissue. Root endosphere samples were obtained by washing the vortexed roots three times in fresh sterile water. To aid in the DNA extraction process, roots samples were first ground in a mortar and pestle with liquid nitrogen and then stored at −80°C.

Plant phenotyping

Plant phenotypic data was recorded at 9 and 25 WAP of the field trial. For the 9 WAP time point, plant height was recorded two days prior to the microbiome sample collection day. At sample collection, shoot tissue was separated from harvested root tissue with pruning shears disinfected with 70% ethanol and shoot fresh weight biomass was measured. The shoot tissue was subsequently dried in paper bags at 50°C for 14 days and weighed again for dry weight. Additional root phenotyping of the root ball was performed after root and rhizosphere sampling using the RhizoVision Crown system following the protocols established in [22]. As described above, at 16 WAP 10 plants per row were bagged for 10 weeks to avoid birds and animal predation. At 25 WAP, the plant height of bagged plants was recorded and each bagged plant was then harvested. Shoots were separated from roots and panicles were weighed in a balance, dried in paper bags at 50°C for 14 days, and weighed again. To quantify seed production, the seeds of each panicle were threshed a week after harvest and weighed. A thousand seeds of each panicle were then counted and weighed to establish 1000-seed weight.

Statistical analyses of the shoot phenotyping data were performed in R-v4.2.1 [19]. Significance was tested by ANOVA one-way, Brown-Forsythe and Welch ANOVA one-way, and Kruskal Wallis according to their distribution and variance homoscedasticity using the aov, oneway.test, and kruskal.test functions in the R stats package [18]. Distribution and variance homoscedasticity were tested using the shapiro.test and bartlett.test functions also in the R stats package [18]. Post-hoc pairwise comparisons tests were conducted using HSD.test function in the R package Agricolae [23], pairwise.t.test function in the R stats package [18], and kwAllPairsDunnTest function in the R PMCMRplus package [24]. Root phenotyping data was analyzed using the pca function in R PCAtools package [25] and the biplot function in R stats package [24], and treatment significance was determined by PERMANOVA test with 999 permutations using adonis function in R vegan package [26].

Linin quantification and monomeric composition

Lignin and lignin S/G ratio were determined on cell wall residues as previously described [27, 28] with some minor modifications; detailed descriptions of this protocol can be found in the Supplementary Materials and Methods. Statistical analyses were performed in R-v4.2.1 [19] as described above.

DNA extraction and 16S rRNA library preparation

DNA extraction of root and rhizosphere samples was performed using Qiagen DNeasy PowerSoil DNA Isolation Kit (Catalog No. 12888–100; QIAGEN, Venlo, Netherlands) following the directions of the manufacturer. Samples were quantified, amplified, and sequenced using a MiSeq System (Illumina) as previously described [29] with some detailed modifications in Supplementary Materials and Methods.

Amplicon sequence processing and analysis

16S rRNA amplicon sequencing reads were demultiplexed in QIIME2 [30] and then passed to DADA2 [31] to generate amplicon sequence variants (ASVs). In both the media-based and in planta experiments, ASVs not alignable via BLAST and phylogeny tree to the 57 16S rRNA sequences of the SRC1 members were removed. Alpha and Beta diversities were determined as previously described [29] using R-v4.2.1 [19]. The bacterial ASV heatmaps were generated using the R package pheatmap [32]. Differentially abundant ASVs were conducted using R package edgeR [33], the ASVs with a log₂FoldChange > 1.5, P < 0.05 were considered differentially significant.

RNA extraction

Frozen bulk root tissue (~100 mg) from each plant was ground to a powder with 5 mm zirconium beads using a Mixer Mill MM400 (Retsch, Haan, Germany) by shaking at 30 Hz for 3 min. RNA extraction was performed as previously described [34] with some minor modifications described in the Supplementary Materials and Methods.

Results

In order to develop a tool that would allow us to explore rhizosphere microbiome assembly and interactions with host phenotype, we chose to formulate a synthetic community from individually cultured bacterial isolates. To identify a combination of strains well-suited to collectively support rhizosphere and root colonization, we used a combination of two distinct strain selection strategies. First, we identified bacterial lineages that we hypothesized would play a role in the community cohesion by applying a network-based analysis of an existing 16S rRNA microbiome dataset [12, 17, 29] from the sorghum rhizosphere environment. This analysis was performed using a Pearson correlation on 16S rRNA data for 126 rhizosphere field samples collected over the entire growing season [12]. Using this approach and additional selection parameters based on ASV abundance and occupancy across the dataset (Fig. S1), we identified a total of 293 abundant and well-connected ASVs belonging to 16 genera. A final selection based on 16S rRNA matches to an existing sorghum root isolate collection and high total abundance within the 16S rRNA dataset was used to narrow the list of target strains to 42 bacterial isolates (Table S1). As a second strategy for engineering a robust community capable of colonizing the sorghum rhizosphere, we additionally included strains capable of using sorghum exudates for growth. To achieve this, we isolated bacterial strains out of soil obtained from the sorghum field using a set of sorghum exudates as the sole carbon source (see supplementary Materials and Methods). These isolations yielded 15 individual strains, belonging to 10 additional genera, each of which was included in the final SRC1 formulation (Fig. S2).

In vitro experiments

As all SRC1 strains were isolated from the plant root system, we hypothesized that growth in the presence of the plant host would significantly and positively influence community growth and stability. To test this, we chose to first explore SRC1 community performance in a context that did not include the plant root system. To this end, we created a mock sorghum exudate media formulation (Sorghum Synthetic Medium, or SSM) composed of metabolites identified within sorghum exudates (see Materials and Methods and Table S2) and grew SRC1 on agar plates comprised of this media type. Each of the 57 individual strains were first grown on TSA plates and then mixed together using approximately equal volume of cells per strain as input; this mixture was then plated on SSM media. After incubating, samples were collected weekly to assess community composition and diversity through 16S rRNA (V3-V4 region) sequencing (Fig. 1A). As a subset of members in the community (n = 24) could not be distinguished from one another by their 16S rRNA sequence across the V3-V4 region, these isolates were grouped and tabulated together as listed in Table S1. The resulting data demonstrate that approximately 55% of the SRC1 strains were no longer detectable on the media after the second week of growth (Fig. 1B). In particular, the plate-based environment failed to support the growth of the majority of Gram-positive strains selected using the network analysis. To test whether inclusion of sorghum exudates collected directly from the plant root could recover growth of these strains, the experiment was repeated using SSM media supplemented with sorghum exudates collected from 5- to 10-week old plants (SSM-Ex) (described in Supplementary Material and Methods); inclusion of sorghum root exudates in the media did not appear to improve survival rates of these root-derived strains, nor did it seem to impact community dynamics over the course of the experiment. However, the majority of strains (12 out of 14) selected based on their ability to utilize sorghum exudates as a sole carbon source remained detectable throughout the eight-week experiment. Collectively, these results demonstrate that the intended overall community structure for the SRC1 is not well supported by in vitro growth on solid media.

In planta experiments

We hypothesized that many of the strains that did not persist in the media-based experiments would grow better in the native habitat from which they were isolated, namely the sorghum rhizosphere. To test this hypothesis, we developed a lab-based in planta growth system using gnotobiotic containers (microboxes of 5-L capacity) filled with calcined clay. To these containers we added sorghum seeds that were inoculated with either our SRC1 or a mock treatment (Fig. 1A). Following a period in which all plants were irrigated (15 days), we divided the plants into two irrigation treatments to explore the performance of SRC1 colonization under a range of conditions. One group remained well-watered throughout the experiment and the other group was subjected to a period of drought stress, a condition previously shown to support growth of Gram-positive strains [12, 17, 29]. As hypothesized, the majority of the SRC1 strains (56 out of 57) were detectable in both root and rhizosphere samples of SRC1 treated sorghum under both watering treatments. Only Brevibacillus agri TBS_043 was not detected in any sample (Fig. 1B). These data demonstrate that whereas sorghum-exudate solubilizers dominated the community composition in media-based growth, a more even distribution of SRC1 strains is evident when the plant host serves as growth substrate for the community.

Apart from SRC1 treatment, the other experimental factors appeared less influential in determining community composition. Likely due to the short duration of the drought stress applied in this experiment, watering treatment did not significantly impact the composition of the community in terms of alpha and beta diversities (Fig. S4, Table S3). In addition, plant compartment led to differences in composition, with impacts primarily in some members of SRC1; for example, both Kribella shirazensis SAI_225 and Flavobacterium sp. SO6b were observed at higher relative abundance in roots compared to rhizosphere. Overall, we observed that Gram-negative strains of SRC1 remained better colonizers than Gram-positive members; this was true both for strains selected based on the network analysis and those selected based on sorghum exudates catabolism. We also observed that application of SRC1 led to significant changes in plant phenotype in a treatment dependent manner. Specifically_, under well-watered conditions application of SRC1 was correlated with increased shoot fresh weight and dry weight compared to mock inoculated controls (Fig. 1C–E). To test whether the improvement to plant phenotype requires a live (bioactive) SRC1 formulation, we repeated this in planta experiment once more using three inoculation treatments: mock, heat-killed SRC1, or live SRC1. Only the live SRC1 treatment produced increased shoot biomass under well-watered conditions (Fig. S5). Collectively, these results indicate that SRC1 grows more uniformly in the presence of the plant root and has the ability to impact plant phenotype when it is metabolically active.

To explore whether the inclusion of sorghum-exudate utilizing strains impacted colonization efficiency of the other SRC1 members, we performed another in planta experiment once again using a modified formula in which the sorghum-exudate utilizers were removed from the community. We noted that many other community members showed increased representation in the resulting 16S rRNA datasets (Fig. 2A), including the strains Streptomyces ambofaciens SAI_104, Streptomyces albogriseolus SAI_173/190, Paenibacillus lautus TBS_091, and Pseudomonas_E pudica TBS_049 (Fig. S6). We also noted that the previously observed positive influence on plant phenotype was conserved in the absence of the sorghum-exudate utilizers; this was true not only under watered conditions, but also appeared during drought conditions under the new formulation (Fig. 2B–E). Collectively, these results show that the inclusion of sorghum-exudate solubilizers impacts the composition of the SRC1 community in sorghum roots, but is not necessary for the observed impact on plant phenotypes under normal irrigation.

Figure 2

Sorghum-exudate solubilizers modulate the SRC1 composition and increase the overall shoot biomass of sorghum plants. (A) Heat map of SRC1 member abundance with or without sorghum exudate utilizers identified in lab-based in planta experiments. Abundance is presented in log2 scale of n = 218 samples. Arrows highlight the SRC1 members potentially modulated by the sorghum exudate utilizers after a Brown-Forsythe and Welch ANOVA one-way and holm-Šídák post hoc test P < 0.05. Sample types (roots and rhizosphere) and water treatments (watered and drought) are shown at the top of the heat map. Genera and members present in SRC1 are shown at the left and right sides, respectively. (B) Representative pictures of watered or drought stressed sorghum plants treated with mock or SRC1. (C) Shoot fresh weight. (D) Shoot dry weight. (E) Root fresh weight. W = normal irrigation; D = drought stress. Letters above represent statistical differences among treatments by ANOVA one way test and Tukey post hoc test with P ≤ 0.05 of 16 plants per treatment (n = 64 plants in total).

Field experiments

The colonization by SRC1 under controlled and presterilized conditions suggests that this community can effectively assemble on the sorghum root in the absence of competition from other environmental microbes. To explore how SRC1 performs in a native context, we conducted experiments in the field through inoculation of SRC1 onto seeds and growing seedlings. In this experiment, sorghum seeds were first subjected to either a preincubation with SRC1 or a mock inoculation prior to sowing in the field at the Kearney Agricultural Research and Extension Center (Material and Methods). A second round of inoculation (SRC1 or mock) was performed on a subset of 4-week old seedlings in each plot to explore the impact of an alternative method of SRC1 deployment. As with our lab based in planta experiment, following a period of normal irrigation, half of the plants in all inoculation treatment types were subjected to a period of drought stress (5–9 WAP) (Fig. 3A). At the end of the drought treatment (9 WAP), we collected root tissue and measured biomass from all treatment groups and performed 16S rRNA community profiling of root and rhizosphere (see Material and Methods). In accordance with results from previous studies of the sorghum microbiome conducted in the field [12], alpha diversity differed between roots and rhizosphere, with lower Shannon index in roots and overall reduced diversity in droughted samples compared with watered treatments (Fig. S7A). It was also noted that the SRC1 inoculation significantly impacted alpha diversity of watered root samples (under seed inoculation), increasing the Shannon index by approximately 2%, from 5 to 5.15 in mock- and SRC1-treated plants, respectively (no effect was observed in the remaining treatments).

Figure 3

SRC1 inoculation impacts the beta diversity of the sorghum root microbiome in the field. (A) Schematic diagram of the field trial design. Sorghum cultivar BTx642 plants were treated once (seeds) or twice (seeds and 4-week-old plants) with mock or SRC1. Two irrigation treatments were applied, a drought stress, no irrigation, for four weeks after the second inoculation (5 WAP - 9 WAP), or irrigation periodically. Microbiome profiling was analyzed at the end of drought stress (9 WAP) and the plant phenotyping was recorded also at 9 WAP and after three months of the restored irrigation at 25 WAP. (B) Beta diversity analysis of the sorghum microbiome of the field trial. Constrained Analysis of Principal Coordinates (CAP) ordination plots of the microbial community in the sorghum (B-C) rhizosphere and (D-E) roots samples under (B,D) normal irrigation and (C,E) drought stress inoculated twice with mock or SRC1.

Figure 4

Long-term lack of engraftment of the SRC1 in the sorghum root microbiome under field conditions. (A) Heat map presents the abundance of 16S rRNA ASVs identified to each SRC1 member across samples treated with mock or SRC1 once (seeds only) or twice (seeds and 4-weeks-old plants). Sample types (roots and rhizosphere), inoculation (mock and SRC1), and water treatments (watered and drought) are shown at the top of the heat map. Genera and members included in the SRC1 are shown at the left and right sides, respectively. Abundance is presented in log₂ scale of the sum of 20 samples per treatment (n = 320 samples in total). (B) Proportions of differentially abundant ASVs (log₂FoldChange > 1.5, P < 0.05) either decreasing or increasing in response to SRC1 application including both root and rhizosphere compartments across both drought and watered conditions, drought only, and water only. (C) Results when root and rhizosphere compartments are separately compared (n = 320 samples in total).

In order to evaluate the effect of SRC1 treatment on the structure of the microbial community, we analyzed the beta diversity through unconstrained Principal Coordinates Analysis (PCoA) using the Bray–Curtis dissimilarity. Four clusters were observed in the analysis across the two axes (Fig. S7B), where the primary axis clustered samples principally by water treatment (explaining 32.8% of variation), and the second axis by sample type (explaining 10.6% of total variation), in agreement with similar previous studies [12]. A permutational multivariate analysis of variance (PERMANOVA) analysis was performed using the Bray–Curtis dissimilarity for sample type, water treatment, SRC1/mock treatment and inoculation time points; these analyses indicate that the microbial community was significantly influenced by all four experimental factors (Table S4), but mostly by sample type and water treatment. To analyze the effect of SRC1 inoculation on community composition in more detail, we performed Constrained Analysis of Principal Coordinates (CAP) using Bray Curtis dissimilarity and Permutation tests of multivariate homogeneity of group dispersions (permutest) for each sample type, water treatment, and SRC1 inoculation time points. Collectively, these analyses indicate that SRC1 inoculation had a significant but small impact (10.2–6.9% for watered and 6.5–3.7% for drought) on the plant microbiome across both treatments (Fig. 3B–E, Fig. S8, and Tables S4 and S5).

To explore the impact of each experimental factor on specific taxonomic fractions of the sorghum microbiome, we analyzed the data at the phylum-level and genus-level using relative abundance plots. In agreement with previous studies of root microbiome response to drought stress [12], we observed an enrichment of Gram-positive bacteria in drought-stressed samples together with a depletion of Gram-negative bacteria. More specifically, the phylum Actinobacteriota and the genus Streptomyces showed higher relative abundance in both roots and rhizosphere of drought-stressed samples (Fig. S9 and S10), whereas the phyla Proteobacteria and Bacteroidota showed lower relative abundance.

We explored the impact of inoculation on microbial community composition. Whereas we had observed a relatively small impact of inoculation timing (seeds versus seedlings) on beta diversity, these microbial compositional analyses were conducted on combined data treating the two inoculation type points as additional replicates. The data reveal that SRC1 inoculation did not appear to lead to significant increases in the relative abundances of SRC1 community members within the root or rhizosphere above background levels observed in the mock control (Fig. 4A). Given that all SRC1 members were originally isolated from soil and roots obtained from this field environment, it was anticipated that mock inoculated samples would exhibit some level of SRC1 ASVs. However, despite a lack of enrichment of any SRC1 ASV between SRC1 inoculation and the mock control, we did observe significant impacts of SRC1 treatment on many other taxonomic groups (Fig. 4B–C). At the phylum level, a greater fraction of Firmicutes ASVs showed significant decreases in SRC1 treatment compared to the mock control (Fig. 4B–C and Fig. 5A). By contrast, ASVs belonging to the phylum Bacteroidota were more likely to have increased relative abundance following SRC1 treatment (Fig. 4B–C and Fig. 5B). For Actinobacteriota ASVs, which tend to be enriched during drought stress, a greater percentage showed gains in relative abundance following SRC1 treatment under watered conditions, but the opposite pattern was observed under drought conditions (Fig. 4B–C and Fig. 5C). At finer taxonomic resolution, each of these phylum level changes appeared driven by enrichment or depletion of a small and select number of target families (Bacillaceae within the Firmicutes, Chitinophagaceae within the Bacteroidota, and Streptomycetaceae within the Actinobacteriota) (Fig. 5D–F). Collectively, these results demonstrate that impact of SRC1 inoculation on root microbiome structure lies largely in rewiring existing community abundance patterns rather than environmental persistence of SRC1 members in the existing soil microbiome.

Figure 5

SRC1 impacts bacterial community member abundances in a lineage-dependent manner. Responses to SRC1 when comparing across all samples, including both root and rhizosphere samples, under both drought and water regime. Log₂CPM (log₂ counts per million mapped reads) represents the mean abundance across both treatments. Negative and positive log₂FoldChange values indicate decreased and increased relative abundance upon SRC1 treatment, respectively. Results separated by phylum (A-C). Significant (log₂FoldChange > 1.5, P < 0.05) differentially abundant ASVs colored by family. Shape corresponds to changes in response to SRC1 application. Boxplots for notable family-wise trends (D-F) include divergent responses among Streptomycetaceae, increases in abundance in Chitinophagaceae, and decreases among Bacillaceae.

Despite a lack of evidence of SRC1 persistence within the root environment of our field study, positive impacts on plant phenotype were again observed following SRC1 treatment. As was observed in the lab-based in planta experiments, the shoot fresh weight, dry weight, and shoot length all increased with SRC1 treatment (both seed and 4-week-old seedling inoculations) compared to mock-treated plants under both water and drought treatments (Fig. 6A–D). In addition, we also observed a positive impact on the root phenotypes of watered plants that had been inoculated twice with SRC1 (Fig. S11, Tables S6 and S7); specifically, total surface area, volume diameter, and projected area diameter were identified as the top three root features (out of 36 features explored) exhibiting the greatest degree of change between SRC1 treated and untreated samples [22]. Whereas no significant effect was observed in either the shoot or root phenotypes on those plants that were inoculated only onto seeds (Fig. S11 and S12, Tables S6 and S7), the number of tillers increased both for plants inoculated once (seeds) or twice (seeds and seedlings) with SRC1 under normal irrigation (Fig. 6E, Fig. S12C). After a period of three months of restored irrigation (25 WAP), the shoot fresh weight and dry weight were measured for all treatment groups; we observed increases in biomass yield both for plants inoculated with SRC1 once or twice (Fig. 6 F–G, Fig. S12D–E) under the watered treatment, although no significant effect was observed on droughted plants at this stage of development. Finally, we noted that SRC1 treatment also positively impacted seed yield for plants under the normal irrigation treatment (Fig. 6 H–I, Fig. S12F–G). Altogether, these results demonstrate that treatment with SRC1 led to increased biomass and yield-related traits in the field, with some dependence on environmental conditions, recapitulating results previously observed in the laboratory setting.

Figure 6

SRC1 enhances the shoot biomass and seed production of sorghum plants in the field. Shoot phenotyping of 8-week-old sorghum plants inoculated twice (seeds and 4-week old plants) with mock or SRC1 (A-E). (A) Representative pictures of watered or drought stressed plants, (B) shoot fresh weight, (C) shoot dry weight (C), (D) shoot length, and (E) tiller number of 8-week-old plants. Shoot phenotyping of sorghum plants after three months of restored irrigation inoculated with SRC1 twice (F-I). (F) Shoot fresh weight, (G) shoot dry weight, (H) total seed weight, and (I) 1000-seed count weight. W = normal irrigation; D = drought stress. Letters above represent statistical differences among treatments by Brown-Forsythe and Welch ANOVA one-way, holm-Šídák post hoc test P < 0.05 for B-C, ANOVA one-way, Tukey post hoc test P < 0.05 for D and I, and Kruskal-Wallis test, Dunn's test of multiple comparisons P < 0.05 for E-H. A total number of 80 and 317 plants were harvested at 9 WAP and at the end of the season, respectively.

Transcriptomic analysis of in planta experiments

Whereas these lab-based in planta and field data demonstrate that SRC1 is able to alter above- and below-ground plant phenotypes compared to uninoculated controls, it does not suggest an underlying cause for these changes. In order to investigate the molecular mechanisms of the impact of SRC1 on plant fitness, we performed a transcriptomic analysis of plant roots treated either with SRC1 or with mock inoculation from the lab-based in planta experiment. The impact of SRC1 treatment on plant root gene expression relative to mock treatment was analyzed via differential expression (DE) analysis. Among the top 30 DE genes were 5 members of the core phenylpropanoid pathway involved in lignin biosynthesis; all of these exhibited down-regulation under SRC1 treatment, suggesting a possible reduction in monolignol content or other phenylpropanoid-derived metabolites following inoculation. Further inspection revealed general down-regulation amongst many other identifiable members of the lignin biosynthesis pathway (Fig. 7A–B).

Figure 7

Transcriptomic analysis reveals downregulation of the lignin biosynthesis pathway in sorghum plants treated with SRC1 relative to control. (A) MA plot comparing transcription of genes in SRC1 treatment relative to mock. (B) Lignin biosynthesis pathway adapted from Barros et al. [70], DE genes represented with black lines. (C) Lignin content quantified with the thioglycolic acid method and expressed as absorbance at 280 nm per mg of cell wall. (D) Differentially expressed gene clusters are shown with cluster names annotated above each heatmap and gene count within each cluster annotated below. Normalized expression values are group-wise means of z-scored counts-per-million, normalized by gene. Letters above bars in C represent statistical differences among treatments by ANOVA one way test and Tukey post hoc test with P ≤ 0.05 of at least 3 pools of 16 plants per treatment (n = 64 plants in total). Metabolites are named as Phe: Phenylalanine, CinAc: Cinnamic acid, p-CoAc: p-coumaric acid, p-CorCoA: p-coumaroyl-CoA, cade: Caffealdehyde, Caol: Caffeyl alcohol, p-code: p-coumaraldehyde, p-CoAl: p-coumaryl alcohol, p-CoShi: p-coumaroyl shikimate, CaShi: Caffeoyl shikimate, CaCoA: Caffeoyl CoA, FeCoA: Feruloyl CoA, Conde: Caffealdehyde, cool: Coniferyl alcohol, 5H-Conde: 5-hydroxyconiferaldehyde, 5H-cool: 5-hydroxyconiferyl alcohol, side: Sinapaldehyde, Siol: Sinapyl alcohol, H: p-hydroxyphenyl, S: Syringyl, G: Guaiacyl, 5H: 5-hydroxyguaiacyl, C: Catechyl. enzymes are named as PAL: Phenylalanine ammonia-lyase, C4H: Cinnamate 4-hydroxylase, COMT: 5-hydroxyconiferaldehyde O-methyltransferase, F5H: Coniferaldehyde 5-hydroxylase, 4CL: 4-hydroxycinnamate:CoA ligase, HCT: Hydroxycinnamoyl-CoA:Shikimate hydroxycinnamoyl transferase, C3’H: Coumarate 3′-hydroxylase, CCoAOMT: Caffeoyl-CoA 3-O-methyltransferase, CCR: Cinnamoyl-CoA reductase, CAD: Cinnamyl alcohol dehydrogenase.

Lignin content and monomeric composition

In light of results from the transcriptomic analysis, we next investigated the direct impact of SRC1 on root lignin content by repeating an additional lab-based in planta experiment with SRC1 under the same experimental conditions as used in the previous experiments (Fig. 1A). Following sample collection, the Thioglycolic Acid (TGA) method was used to measure the lignin content from all treatment groups. These data show that SRC1-treated sorghum roots from drought treatments exhibited significantly reduced lignin content compared to the mock-treated samples; SRC1-treated root samples under drought treatment had 20% lower lignin content compared to mock, whereas those under watered conditions showed no significant reductions (Fig. 7C). To further characterize the differences in lignin, cell wall material from roots of SRC1 and mock-treated plants were analyzed by pyrolysis–gas chromatography–mass spectrometry to determine lignin monomer composition. For each treatment, identification and relative quantification of the pyrolysis products derived from guaiacyl (G) and syringyl (S) units allowed determination of the S/G ratio, a trait that is directly related to digestibility and rigidity of the plant cell wall [35] (Table S10). This analysis showed that SRC1-treated samples exhibited altered S/G ratios compared to the mock, though the changes were water treatment dependent; increased ratios of S/G monomers (1.2-fold) compared to mock samples were observed under water treatment, whereas SRC1 droughted samples exhibited the opposite trend (0.9-fold).

Connections to immune responses

In addition to its role in cell wall processes, the lignin biosynthesis pathway is also known to be connected to plant immunity; lignin deposition is an essential process that blocks pathogen invasion and arrests the access or diffusion of pathogenic molecules [36] and its pathway-associated phytoalexins have direct involvement in plant defense signaling and action [37]. Considering this, and the increased biomass of SRC1-inoculated plants in lab-based experiments and the field, we hypothesized that the observed downregulation of the lignin biosynthesis pathway may co-occur with a general suppression of immunity, leading to increased plant biomass through a growth-defense trade-off [38]. To explore this hypothesis, we defined DE gene clusters that responded to the SRC1 treatment, to the watering treatment, or to both treatments (Fig. 7D). As there are known antagonistic interactions between drought stress and immunity that could influence our ability to detect SRC1-related patterns in the data, we chose to focus on Cluster II, which includes genes impacted by SRC1 treatment but not by watering regime (Table S8). A hypergeometric test was used to test for enrichment of the 878 genes within 12 immunity-related Gene Ontology (GO) terms among the 1101 genes observed in Cluster II (SRC1-DOWN - (Drought-UP ∪ Drought-DOWN)) relative to the full genome (Table S9). The result was significant (P = 0.038), suggesting a downregulation of immunity under SRC1 application that accompanies the downregulation of lignin biosynthesis. Subsequent testing of Cluster II for enrichment of the individual GO terms revealed significant enrichment for GO:0009751 response to salicylic acid (P = 0.026) and GO:0002218 activation of innate immune response (P = 0.009). Collectively, these data suggest that the impact of SRC1 on plant phenotype may be driven in part by a suppression of immune activity through shifts in the lignin biosynthesis pathway.

Discussion

In this study, we developed a defined synthetic community comprised of individual cultured isolates from 16 genera representative of taxa commonly found within the sorghum rhizosphere. One clear observation from our studies is that the community structure of the assembled community SRC1 varied significantly depending on both growth environment and the presence or absence of specific community members or additional native microbes. In plate-based assays, despite our efforts to include in the media complex metabolites typically provided by the plant, Gram-positive bacteria were less successful at persisting across multiple passages compared to their Gram-negative counterparts. This may in part be explained by differences in growth strategies of these two groups. Gram-negative bacteria are commonly r-strategists [39], which favor resource-rich environments; the media used in this experiment likely presents a comparatively resource-rich environment with compounds that can be rapidly metabolized [39]. Similar results have been observed in other studies, with favor falling to Gram-negative bacteria when simple carbon sources are plentiful [40]. This explanation is unlikely to fully account for the results, however, as the Bacillus strains in SRC1 are both fast-growing r-strategists and members which fared poorly in the media based growth environment.

Given other results from this field of research [10, 11, 41], it is perhaps not surprising that, when SRC1 was grown in the presence of native microbial diversity and wider environmental heterogeneity typical of field environments, we were not able to detect long-term enrichment of 16S rRNA signatures from any SRC1 community members. This outcome has been observed in many other systems where applied microbes that colonize well in gnotobiotic conditions or greenhouse settings did not exhibit persistent colonization in the root microbiome in the field [10, 11, 41]. The root cause of this has been hypothesized to be the result of comparatively poorer fitness of the introduced strains for the specific environmental conditions present in the targeted environment. However, in our specific case, all SRC1 members were originally cultured from sorghum roots or soils taken from this same field location, suggesting the likelihood of other factors at play. One possible explanation could be that colonization is developmental stage dependent; root and rhizosphere microbiomes have been shown to be dynamic, with taxonomic variation occurring throughout the growing season [42]. This could allow for the possibility that SRC1 may have exhibited early colonization followed by replacement with alternative soil-derived taxa in response to plant metabolism, plant root development, or edaphic changes later in the experiment. An alternative explanation could be that an established balance between our SRC1 strains, which were derived from this field, and the rest of the resident soil microbiome was already established prior to inoculation, and prevented or buffered against additional shifts in favor of SRC1 strains. Future studies that include either broader longitudinal sampling of microbial communities with tagged SCR1 members, or alternatively with SRC1 introduced into a different field site, could capture early post-inoculation SRC1 colonization dynamics and help differentiate their engraftment in the rhizosphere from that of other native strains.

Despite a lack of evidence of long-term engraftment of SRC1 in our field study, there was observable impact on the broader root microbiome composition. Other studies have found that SynCom treatment can lead to rewiring of existing community structure without significant increases in the relative abundances of SynCom community members [9, 10, 41]. In our experiment, at the time of harvest (nine weeks after planting) we observed an enrichment in the family Chitinophagaceae taxa in SRC1 treated plants. It has been reported that the relative abundance of the parent lineage Bacteriodota is positively correlated with the total C, N, and P in soil [43], and that members of this group can also degrade complex organic matter such as chitin and cellulose [44]. These processes can provide C and N to the plant, possibly helping explain improved plant phenotypes in our SRC1 treated plants. By contrast, the family Bacillaceae was mostly depleted in the SRC1-inoculated plants under both irrigation treatments despite being one of the dominant taxonomic groups in our SRC1 formulation (n = 18 Bacillus strains). This decrease in Bacillus relative abundance may in fact be related to the observed SRC1-mediated decreases in immune response. Lebeis et al. [45] demonstrated that SA-insensitive Arabidopsis mutants with diminished immune signaling had decreased relative abundance of many Bacillus within their rhizospheres. In fact, much of the microbiome modulation following SRC1 treatment may be tied to the observed shifts in immunity. Numerous recent studies have found strong connections between defense signaling and bacterial colonization in plants [46], with the activation or suppression of salicylic acid and jasmonic acid signaling pathways impacting the colonization of a wide range of bacterial lineages in a variety of plant systems [47–49].

Equally interesting to the observed modulation of the native root microbiome structure is the clear impact of SRC1 treatment on the lignin biosynthesis pathway. As one of the most important polymers within cell wall barriers, lignin governs the diffusion of water and solutes into and out of the vascular system in roots [50]. Recently, Kawa et al. [51] showed that at least some configurations of the root microbiome can influence root cellular anatomy traits of sorghum, specifically leading to less endodermal lignification in crown roots compared to a sterilized treatment group. In addition, Salas-González et al. [52] demonstrated strong connections between pathways associated with root diffusion barriers, root microbiome composition, and shoot biomass. This research demonstrated that Arabidopsis plants treated with SynComs had negatively altered suberin deposition, another important component of water diffusion barriers in the root [53], and noted a positive correlation with nutrient accumulation in shoots.

SRC1 treatment was also shown to alter the ratios of the S and G monomers of lignin within the root tissue. Recent work has demonstrated that alterations in the S/G ratio directly impact the digestibility of biofuels feedstocks [54–57], potentially influencing the conversion of plant tissue into bioethanol. As bioenergy genotypes of sorghum represent a growing section of the biofuels market [58–60], it will be interesting in the future to explore whether treatment of SRC1 or related microbiome amendments could be used to increase the efficiency of biofuel production in above-ground tissues as well. Whether the decreased lignin and changes in the S/G ratio following SRC1 treatment observed in our study is directly responsible for improved phenotypes remains unclear; at present, we suspect it is unlikely to be the primary driver of biomass improvements, as the most significant changes were observed under drought stress, whereas phenotypic changes were found under both watering conditions.

An alternative, and potentially more plausible, explanation for the SRC1-driven improvements in growth phenotypes lies in the observed transcriptional changes to plant immune response and the growth-defense trade-off. The growth-defense trade-off theory suggests that due to limitations in resources, plants are required to prioritize either growth or defense, and that an integration of both external and internal factors [38] is used to prioritize one over the other. Based on this model, a suppression of plant immune signaling could remove or reduce basal levels of growth inhibition [61] leading to more robust growth phenotypes. Recently, several studies have shown an important crosstalk between lignin pathways and growth and defense trade-off systems [62–66], offering further support for this hypothesis. Future work to investigate the direct impact of SRC1 synthetic community treatment and individual community members on plant immune activation will help fully evaluate this possibility.

The results from this study, along with other recently published work [9], suggest that it may be helpful to distinguish between cases in which persistence of a microbial inoculum is likely to be required versus helpful. In the context of our field trial, persistence of the members of the SRC1 did not appear to be necessary for long term impact on the plant performance. We suspect that it is possible that other microbial amendments in agricultural settings may also be able to influence crop phenotypes without long-term colonization of the host. However, in some specific scenarios, persistence may be a prerequisite for success. For instance, in the case of microbially-mediated soil remediation, microbial presence may be necessary for the long-term services it provides. Phytoremediation studies of diesel-contaminated soils have shown a positive correlation between the persistence of the population of inoculated microbes in the rhizosphere and endosphere of the host plant and the improvement of plant growth, hydrocarbon degradation, and toxicity reduction [67]. Similarly, for the long-term prevention of a specific pathogen agent through biocontrol, short term colonization may not be sufficient for pathogen suppression. The mode of action of many of these disease suppressive agents is often direct microbe-microbe interaction, principally mediated by allelochemicals including siderophores, antibiotics, cell-wall-degrading enzymes, volatile organic compounds, alkaloids, and many other compounds [68]. As an example, Sheoran et al. [69] demonstrated that black pepper-associated endophytic Pseudomonas putida BP25 could inhibit, by volatile emission, the proliferation of multiple pathogens including fungi and oomycetes, and a plant-parasitic nematode. For this method of disease suppression to be effective, the biocontrol agent would likely need to remain in the system because volatile compounds typically have short half-lives in soil environments. We propose that whereas a better understanding of the forces that act to limit the persistence of introduced microbes is critical to designing ones that can persist, it is also worth considering whether persistence is in fact required for the intended outcome.

Conclusion

There is increasing commercial interest in the use of microbial amendments in agriculture environments to improve yield and fitness outcomes. Consistent challenges in achieving these desired outcomes following inoculation are often attributed to failure of these products to compete and colonize their environment. In this study we simultaneously demonstrate that: 1) community engraftment operates as a function of the microbial diversity in the system and growth environment parameters; and 2) that engraftment may not be strictly necessary over long periods of time for realization of specific outcomes. Collectively, these results point to the promise of SynCom-based strategies for field deployment and use in improving agricultural activity.

Acknowledgements

We thank Alex Styer, Bradie Lee, Jesus Sanchez, Joelle Park, Claudia Castro, and the staff of the Kearney Agricultural Research Center for their help in sample collection and field preparation. We also thank Claudia Castro for her constructive revision of this manuscript.

Conflicts of interest

The authors declare no conflict of interest.

Funding

This work was supported by US Department of Agriculture (CRIS 2030-21430-008-00D), USDA-NIFA (2019-67019-29306), and is a contribution of the Pacific Northwest National Laboratory (PNNL) Secure Biosystems Design Science Focus Area “Persistence Control of Engineered Functions in Complex Soil Microbiomes” (operated by the U.S. DOE under contract DE-AC05-76RL01830). The work at Lawrence Berkeley National Laboratory was conducted by the DOE Joint BioEnergy Institute (http://www.jbei.org) supported by the U. S. Department of Energy, Office of Science, Biological and Environmental Research Program, through contract DE-AC02-05CH11231 between Lawrence Berkeley National Laboratory and the U.S. Department of Energy.

Data availability

All datasets and scripts for analysis are available through GitHub (https://github.com/CitlaliFG/SRC1-PerCon) and all short read data can be accessed through NCBI BioProject PRJNA1097579. Raw data of the host transcriptome can be accessed and downloaded via JGI Genome Portal (https://genome.jgi.doe.gov/portal/) using the Project ID: 1425625.