Impact of Inoculation Practices on Microbiota Assembly and Community Stability in a Fabricated Ecosystem

Increasing pressure on agricultural systems and environmental degradation have led to examining new methods for developing sustainable agricultural practices in order to achieve global food security goals (United Nations 2022). Much effort is invested currently in developing microbial inoculants to replace or supplement fertilizers to improve crop productivity and environmental sustainability (Adesemoye and Kloepper 2009; Tyagi et al. 2022). The vast majority of conventional biological management strategies and scientific studies assessing the impact of biotic/abiotic stresses on biocontrol agents and/or native microbial communities use single microbial agents as inoculants (Khatoon et al. 2020; Oleńska et al. 2020). However, the activity of individual microbial species might differ when strains are applied as part of a community or consortium (Finkel et al. 2020; Matos and Garland 2005). This is the reason why in recent years the use of synthetic microbial communities (SynComs) has been explored (de Souza et al. 2020; Liu et al. 2019; McCarty and Ledesma-Amaro 2019). SynComs represent systems of reduced complexity that involve co-culturing multiple taxa under well-defined conditions to mimic the structure and function of a microbiome.

From an applied and biotechnological point of view, the SynCom approach has become popular because it can help assess how to develop versatile and productive multibacterial inoculants for the agriculture sector (McCarty and Ledesma-Amaro 2019). From an ecological point of view, the SynCom is a useful tool in understanding plant–microbe interactions in the complex, real-world environment (de Souza et al. 2020; Liu et al. 2019; Pradhan et al. 2022). SynComs provide functional and mechanistic insights into how plants regulate their microbiomes, and they can be used to assess the impact of biotic and abiotic stresses on microbial communities (de Souza et al. 2020; McCarty and Ledesma-Amaro 2019; Pradhan et al. 2022). Additionally, SynComs are useful for understanding the dynamic interactions within plant–bacteria ecosystems, which is applicable for tool development and identification of candidates for targeted microbiome manipulation (de Souza et al. 2020; Shayanthan et al. 2022).

A successful SynCom inoculant and inoculation methodology needs to address three criteria: persistent plant association, formation of a stable microbial community, and consistency in producing favorable plant phenotypes (Bhardwaj et al. 2014; Mitter et al. 2021). Soil–plant–microbe experimental systems are complex, and bridging the lab–field gap in a replicable and reproducible manner is laborious (de Souza et al. 2020; Parnell et al. 2016). A major challenge in identifying appropriate SynCom inoculants is the scarcity of comparable data sets. The lack of comparable data sets is in part due to the absence of standardized systems and protocols (Meisner et al. 2022; York et al. 2022). In an effort to address consistency and to achieve reproducibility, especially for plant traits, researchers have developed different types of fabricated ecosystems (e.g., EcoFAB, Rhizobox, and Rhizotron) (Busch et al. 2006; Ke et al. 2021; Oburger et al. 2013; Sasse et al. 2019; Yee et al. 2021). One of these ecosystems, EcoFABs, has enabled reproducible Brachypodium distachyon plant phenotypes in the absence of bacterial inoculant in a multilaboratory experiment across four different labs (Sasse et al. 2019). Experiments involving bacterial SynComs add an extra layer of complexity to these assays with fabricated ecosystems. Ensuring microbial colonization and achieving predictable long-term plant phenotypes are challenges to overcome in these experiments.

SynCom assays need to be accompanied by standardized methods that include SynCom preparation and inoculation procedures, the plant stage at which the SynCom is applied to plants, and data analysis pipelines. Comparable data on the impact of plant age, inoculation method, and sample harvesting time on microbial community composition are currently scarce.

Few studies have investigated the effects of inoculation methods on the dynamics of plant–microbe interactions. There are several methods for applying a SynCom to the plant: seed inoculation, transplant inoculation, and seedling root inoculation (Bashan 1998; dos Santos Lopes et al. 2021; Kumar et al. 2022). The seed inoculation method involves applying a bacterial culture (liquid or powdered) directly to the seeds, and it is widely used in agriculture (O'Callaghan 2016; Rocha et al. 2019; Sarkar et al. 2021). The inoculation upon transplant method involves microbial inoculation when transferring seedlings from the nursery plate to their permanent growth habitat (e.g., pot, field), and it is commonly used for vegetable growth and lab studies (Herrera Paredes et al. 2018; Hu et al. 2020; Schillaci et al. 2021). The seedling root inoculation method involves bacterial inoculation to the root after the plants have been growing for some time in their final habitat (dos Santos et al. 2019; Valetti et al. 2018). How these different inoculation methods affect the formation of a stable microbial community, persistent plant association, and consistency of the plant–microbe ecosystem is poorly understood.

With the aim of addressing the knowledge gap, we conducted a study using fabricated ecosystems consisting of sand-filled EcoFABs, a 17-member bacterial SynCom, and the model plant B. distachyon. The EcoFAB is a controlled modular growth system designed for reproducible plant–microbiome studies (Gao et al. 2018; Sasse et al. 2019; Zengler et al. 2019). We used quartz sand to provide the root with a physical attachment environment like that in soil (Gao et al. 2018; Sasse et al. 2020). The SynCom we used in this work is a well-characterized microbial community derived from the rhizosphere and adjacent soil of Panicum virgatum (switchgrass) plants (Coker et al. 2022). The 17 bacterial strains were selected from different genera, and their 16S rDNA sequences can be readily distinguished. The plant B. distachyon was chosen because it is an excellent model to study soil–plant–microbe interactions due to several advantages. First, it is a monocot model species for cereal and bioenergy crops, as it is phylogenetically closely related to switchgrass, sorghum, wheat, rice, and barley (Draper et al. 2001; International Brachypodium Initiative 2010). B. distachyon shares a similar root architecture with economically important grasses but with one-third the size (Hong et al. 2011; Kawasaki et al. 2016; Watt et al. 2009), making it possible to both observe and sample the whole root system in small lab-based experimental systems such as EcoFABs. Second, it has a small height, rapid life cycle, and reduced genome size, and a transfer DNA mutant collection is available (Bragg et al. 2012; Draper et al. 2001; Hong et al. 2011; Hsia et al. 2017; International Brachypodium Initiative 2010; Kawasaki et al. 2016; Vogel 2016). Third, the development of B. distachyon is well established and standardized using the Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie (BBCH) scale (Hong et al. 2011), which allows easy comparison between experimental setups. In addition to being an excellent grass model for the aforementioned reasons, understanding the relationships between B. distachyon and microbial communities could reveal novel biomass production enhancement strategies for this plant.

Using the EcoFAB−B. distachyon−SynCom system, we evaluated how different SynCom inoculation practices and plant harvesting time points affect the persistence, stability, and consistency of the sand–plant–microbe fabricated ecosystem. Our data showed that for at least 21 days postinoculation (DPI), the SynCom was highly persistent and stable regardless of inoculation practice, and the plant phenotypes were highly reproducible. The results demonstrate that our setup is a highly conserved system suitable for future plant–microbe interaction studies. This could include but is not limited to resilience studies under biotic and abiotic stress, SynCom strain selection and tailoring, modeling rhizosphere microbe–microbe and plant–microbe interactions, building more complex fabricated ecosystems that contain fungi/phages/archaea, and in situ microbial community gene editing.

MATERIALS AND METHODS

EcoFAB preparation.

EcoFAB devices (https://eco-fab.org/) were fabricated according to (Gao et al. 2018) with some modifications. Briefly, a siloxane elastomer base-curing agent mixture (1:10 vol/vol) (polydimethylsiloxane [PDMS]) (Ellsworth Adhesives) was poured onto a 3D-printed mold and allowed to solidify at 80°C for 4 h. The PDMS layer was separated from the mold, the edges trimmed, and the surface cleaned with a plasma cleaner and permanently bonded to a glass microscope slide by pressing the PDMS and glass slide together. The EcoFABs were then filled with substrates. Hydroponic medium consisting of a 50% strength Murashige and Skoog (0.5× MS) basal salt mixture (PhytoTech Labs) (Murashige and Skoog 1962) without additional carbon source was used to compare the development of B. distachyon grown in EcoFABs with conventional methods. Non-acid-washed quartz sand with a 50 to 70 mesh particle size (Sigma-Aldrich) was used for all other experiments. Finally, each device was placed in a Magenta GA-7 box (Fisher Scientific) with a customized watering system (Fig. 1) and sterilized by autoclave for 15 min.

Download as PowerPoint

Plant growth conditions.

B. distachyon Bd21-3 is an inbred diploid line with a full genome sequence (Vogel and Hill 2008) (hereafter referred to as Bd21-3). Bd21-3 seeds were dehusked and sterilized in 70% vol/vol ethanol (EtOH) for 2 min and 50% vol/vol sodium hypochlorite (NaOCl) for 5 min followed by five wash steps in sterile Milli-Q water (Millipore). The surface-sterilized seeds were stratified at 4°C in the dark for 2 days and then germinated on sterilized wet filter paper for 3 days. Bd21-3 seedlings were grown in a growth chamber maintained at 200 µmol m^–2 s^–1 with a 16-h light/8-h dark regime and 40% relative humidity at 24°C. After 3 days, using sterile forceps, seedlings of comparable size were transferred to EcoFABs with a root growth chamber volume of 3.1 ml. For seeds inoculated at 0 days after germination (DAG), sterilized seeds were placed directly into EcoFABs.

After being transferred into an EcoFAB, the root growth chambers were filled with 0.5× MS medium. Each EcoFAB was then placed into a sterile Magenta box to maintain sterile conditions during the experiment and maintained in the growth conditions described earlier. EcoFABs were supplemented with sterile water every 2 to 3 days until harvest through a customized watering system equipped with a 0.22-µm filter (Pall Life Sciences) to avoid opening the system (Fig. 1). EcoFABs without plants were used and incubated in the same conditions described earlier.

Bacterial strains and culture conditions.

The bacterial strains used in this study (Table 1) were obtained from a P. virgatum (switchgrass) field in Oklahoma, United States (Coker et al. 2022). Strains were cultured at 30°C at 180 rpm using Reasoner's 2A (R2A) medium (Reasoner and Geldreich 1985; van der Linde et al. 1999), 10% strength R2A (0.1× R2A), Luria-Bertani medium (Thermo Fisher Scientific), and yeast-malt extract medium as indicated in Table 1. Bradyrhizobium OAE829 was streaked onto several 0.1× R2A agar plates and grown at 30°C for 10 days and then resuspended in liquid medium. To test their ability to grow on the plant medium, bacteria were streaked simultaneously on 0.1× R2A agar and 0.5× MS salt agar.

TABLE 1 Bacterial strains of the 17-member synthetic community used in this study^a

Strain	Medium	Growth on 0.5× MS
Arthrobacter OAP107	R2A	Y
Bacillus OAE603	R2A	Y
Bosea OAE506	R2A	N
Bradyrhizobium OAE829	0.1× R2A	N
Brevibacillus OAP136	R2A	N
Burkholderia OAS925	R2A	Y
Chitinophaga OAE865	R2A	N
Lysobacter OAE881	LB	N
Marmoricola OAE513	R2A	N
Methylobacterium OAE515	LB	N
Mucilaginibacter OAE612	R2A	Y
Mycobacterium OAE908	R2A	Y
Niastella OAS944	R2A	N
Paenibacillus OAE614	R2A	N
Rhizobium OAE497	YM	Y
Rhodococcus OAS809	R2A	Y
Variovorax OAS795	R2A	N

^a LB = Luria-Bertani; N = No; R2A = Reasoner's 2A; Y = Yes; YM = yeast-malt extract; 0.5× MS = 50% strength Murashige and Skoog.

View as image HTML

Synthetic community preparation and inoculation.

Bacterial strains (except Bradyrhizobium OAE829) were grown for 48 h in 10 ml of culture medium as described earlier. Bradyrhizobium OAE829, which does not grow well in liquid medium, was grown on several 0.1X R2A agar plates for 10 days, and the bacterial mass was collected by sterile loops, pooled, and resuspended in 0.5× MS. Cultures were centrifuged at 4,000 × g for 20 min and washed twice with 0.5× MS. Then the OD₆₀₀ of each culture was measured and the 17 strains were pooled such that each had a final OD₆₀₀ equal to 1. This SynCom liquid mixture was diluted 100 times and then served as the inoculant. The SynCom inoculant (31 µl) was applied directly to each root growth chamber in the EcoFABs, which were prefilled with 0.5× MS medium.

B. distachyon plants were inoculated with the SynCom using one of three methods: seed inoculation (inoculation at 0 DAG), transplant inoculation (inoculation at 3 DAG in parallel with transferring the plant to an EcoFAB), and seedling inoculation (inoculation at 8 DAG when coleoptile node roots and lateral roots started to develop) (Supplementary Fig. S1). At least three biological replicates were used for each condition.

Sample harvesting.

At harvest, samples were collected in sterile conditions unless otherwise specified. The EcoFABs were removed from each Magenta box, the root growth chambers were sliced open with a sterile scalpel, and the plants were taken out. The fresh weight of aerial and root fractions was recorded. Three types of samples were collected: “sand” (bacteria isolated from sand not attached to the root), “rhizosphere” (bacteria from sand attached to the root), and “root” (bacteria from inside the surface-sterilized root). For the sand samples, the remaining sand in the EcoFAB after the plants were taken out was collected into a 50-ml tube containing 25 ml phosphate-buffered saline (PBS) (pH 7.2; Thermo Fisher Scientific). For the rhizosphere samples, the sand covering the root was obtained by placing the root in a 50-ml tube containing 25 ml PBS and vortexing for 10 s (twice). To obtain the root samples, the root was surface-sterilized by washing three times with Milli-Q water and then with 3% vol/vol hydrogen peroxide (H₂O₂) for 30 s followed by a wash with 100% vol/vol EtOH for 30 s, 6.15% vol/vol NaOCl with 0.01% Tween 20 for 3 min, and 3% vol/vol H₂O₂ for 30 s and five final washes with sterile Milli-Q water for 30 s. In the case of EcoFABs without plants, only the “sand” samples were harvested. For bacterial DNA extraction, samples were stored at –80°C until processed. For colony-forming unit (CFU) determination, samples were diluted serially, plated onto 10% R2A agar, and cultured at 30°C. Colonies were counted after 3 days.

DNA extraction and 16S rDNA amplicon sequencing.

The DNeasy PowerSoil HTP 96 Kit (Qiagen) was used to extract bacterial DNA according to the manufacturer's manual. Sand and rhizosphere samples were prepared by adding 250 mg of the collected substrate to a PowerBead plate (Qiagen) and homogenizing with a TissueLyser II (Qiagen) at a frequency of 20 Hz for 10 min (twice). Root samples were homogenized with a 1/4-inch precision ball using the TissueLyser II at 30 Hz for 3 min (twice). Homogenized root samples were transferred to a PowerBead plate, and DNA was extracted with the other samples.

DNA concentrations were quantified by a Quant-iT dsDNA high-sensitivity kit (Thermo Fisher Scientific) and normalized to 0.3 ng/µl. DNA template was added to a polymerase chain reaction to amplify the V4/V5 16S gene region using 515F/926R primers based on the Earth Microbiome Project primers (Parada et al. 2016; Quince et al. 2011) but with in-line dual Illumina indexes (Price et al. 2018; Sharpless et al. 2022). The amplicons were sequenced on a MiSeq with 600-bp MiSeq Reagent Kit v3 (Illumina).

Data analysis and visualization.

The 16S rRNA gene amplicons were analyzed as follows. Reads were processed with custom Perl scripts, implementing PEAR for read merging (Zhang et al. 2014). USearch was used to map reads to a database of SynCom V4/V5 16S region sequences using the ‘annot’ command (Edgar 2016). The mapped sequences were then used to identify and remove chimeras. The Ribosomal Database Project database (Cole et al. 2014) was used to assign taxonomy to unidentified amplicons. Data visualization was performed within Jupyter Notebook 6.3.0 using either the ‘seaborn 0.11.1’ and ‘matplotlib 3.3.4’ packages in Python 3.8.8 or within RStudio v2023.03.0 (Posit, PBC) using the ‘ggplot2 3.3.6’ package in R 4.2.2.

Statistical analyses were performed within RStudio v2023.03.0 using R 4.2.2. Statistical analyses on plant growth were performed using one-way analysis of variance (ANOVA) followed by Tukey's honestly significant difference (HSD) test (alpha = 0.05). Shannon diversity index was calculated using the ‘vegan 2.6.4’ package, and statistical analysis was performed using one-way ANOVA followed by Tukey's HSD test (alpha = 0.05). Beta diversity was calculated by Bray-Curtis distance using ‘vegan 2.6.4’, and permutational multivariate ANOVA (PERMANOVA) was calculated using ‘adonis2’ in the ‘vegan 2.6.4’ package. Pairwise comparisons were performed using DESeq2 v1.38.3. Reproducibility was calculated using the Euclidean distances among replicates within each group (0, 3, and 8 DAG).

RESULTS

The developmental stages of B. distachyon grown in an EcoFAB align with those of conventional methods.

To compare the development of B. distachyon grown in EcoFAB with the growth stages observed using conventional plant growth methods (Hong et al. 2011), we used hydroponic medium (0.5× MS) instead of quartz sand to better visualize root development. Three-day-old seedlings had a visible radical root and coleoptile that corresponded to day 2.9 in Hong et al. (2011) (Fig. 2; Table 2). At 8 DAG, the first leaf of the seedling had unfolded and two coleoptile node roots had appeared, which corresponded to day 7.6 (Hong et al. 2011) (Fig. 2; Table 2). On day 14, the shoot had three unfolded true leaves, the first tiller became detectable, and the root started to reach the far end of the EcoFAB root growth chamber (Fig. 2). Similarly, the “three unfolded true leaves” stage was observed at day 15.0 (Table 2). At 28 DAG, the plant began to enter the reproductive phase, with an extending flag leaf sheath, and the root system continued developing and occupied the root growth chamber, corresponding to day 24.9 (Fig. 2; Table 2).

Download as PowerPoint

TABLE 2 Brachypodium distachyon developmental stage in EcoFABs filled with hydroponic media with published data^a

DAG	Description	BBCH stage	Days according to Hong et al. 2011
0	Germination	0	0
3	Coleoptile emerged	7	2.9
8	First leaf unfolded	11	7.6
14	Beginning of tillering	21	15.8
28	Flag leaf sheath extending	41	24.9

^a The B. distachyon developmental stage was recorded at different days after germination (DAG) with a brief description. The description is compared with that used by the Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie (BBCH) numerical scale system.

View as image HTML

In the absence of a plant, Variovorax dominates the microbial community in the EcoFAB.

To understand how the SynCom assembled temporally without a plant, we inoculated sand-filled EcoFABs and determined their 16S rRNA gene profiles at different time points. The SynCom composition of the liquid inoculant control was similar to that of sand at 0 DPI (Fig. 3A). Variovorax OAS795 dominated the population as early as 6 DPI and remained a dominant species (>55% of the population) until 21 DPI. By contrast, the abundance of Lysobacter OAE881, Niastella OAS944, and Bacillus OAE603 significantly decreased to less than 0.01% from 6 DPI onward. Brevibacillus OAP136 and Bradyrhizobium OAE829 were not detected at 6 DPI and thereafter. Marmoricola OAE513 and Mycobacterium OAE908 could not be detected in the liquid inoculant despite being included in the SynCom at the same ratio as other bacterial strains based on OD₆₀₀.

Download as PowerPoint

To explore the impact of co-culture timing on bacterial composition, we performed a principal coordinate analysis (PCoA) of Bray-Curtis dissimilarity on all samples. The samples clustered primarily by DPI (PERMANOVA, F = 8.3, P = 0.0007) and sample type (PERMANOVA, F = 4.8, P = 0.0232), which together explained 77.6% of the variation (Fig. 3B). The dissimilarity within groups was smaller when the co-culture time increased and was most similar between 14 and 21 DPI. These results implied that the SynCom in quartz sand in the absence of plants had reached a stable community by 14 DPI.

The rhizosphere microbiome is significantly affected by inoculation practices.

To assess how inoculation practices shaped microbial communities in the presence of the plant, we inoculated B. distachyon with the SynCom and collected samples from sand, rhizosphere, and root at 14 DAG (Fig. 4). This enabled us to elucidate the effects of inoculation practices regardless of the plant developmental stage. The 14 DAG B. distachyon plants had at least three unfolded true leaves, and the root had reached the far end of the EcoFAB regardless of the SynCom inoculation practice (Supplementary Fig. S2), with a similar morphology to that observed in hydroponic EcoFABs (Fig. 3). There were no significant differences in fresh weight between different inoculation practices (ANOVA, P = 0.406 and 0.134 for shoot and root, respectively) (Supplementary Fig. S3). We determined the within-sample diversity (alpha diversity) of the SynCom using the Shannon diversity index (Fig. 4A). All root samples had significantly lower alpha diversity than their rhizosphere and sand counterparts (ANOVA, P = 1 × 10^–10), and different inoculation practices had similar alpha diversity (ANOVA, P > 0.05) in each sample type.

Download as PowerPoint

To further explore the between-sample diversity (beta diversity), we performed a PCoA test of Bray-Curtis dissimilarity. The PCoA1 (42.73%) was best explained by the variation within root samples, and PCoA2 (21.41%) clustered among sand and rhizosphere samples (Fig. 4B). The top two axes (PCoA1 and PCoA2) explained 61.41% (42.73% + 21.41%) of the variance. A PERMANOVA analysis showed that beta diversity was best explained by sample type (F = 21.9, P < 0.0001) and moderately (0.05 < P < 0.1) explained by inoculation practice (F = 1.9218, P = 0.058). To further evaluate whether different inoculation practices affected microbiome assembly within each sample type, we performed a PERMANOVA test among different inoculation practices. The microbial community was significantly affected by inoculation practice in the rhizosphere samples (P = 0.0059), moderately affected in sand samples (P = 0.08309), and not affected in root samples (P = 0.7154) (Supplementary Table S1).

To gain a better understanding of how inoculation practices affected the rhizosphere microbial community, we performed pairwise comparisons between inoculation practices for each SynCom member (Fig. 5). The Mucilaginibacter OAE612 rhizosphere in the 3 DAG group was significantly lower than that in the 0 DAG group (DESeq2, Benjamini-Hochberg corrected P = 0.0059) (Fig. 5A). When comparing 8 DAG rhizosphere microbial communities with those of the 0 DAG group, Burkholderia OAS925 and Lysobacter OAE881 were enriched (DESeq2, Benjamini-Hochberg corrected P = 0.0009 and 0.0001, respectively), whereas Variovorax OAS795 decreased (DESeq2, Benjamini-Hochberg corrected P = 0.0008) (Fig. 5B). When comparing 8 and 3 DAG groups, Burkholderia OAS925 and Lysobacter OAE881 were enriched (DESeq2, Benjamini-Hochberg corrected P = 0.0240 and 0.0092, respectively), whereas Methylobacterium OAE515 and Rhodococcus OAS809 decreased (DESeq2, Benjamini-Hochberg corrected P = 0.0063 and 0.0072, respectively) (Fig. 5C). In summary, the data demonstrated that inoculation practices significantly affected the rhizosphere microbial communities of 14 DAG plants.

Download as PowerPoint

The transplant inoculation is most reproducible when the microbial community reaches a steady state.

We analyzed samples harvested at 21 DPI across three inoculation practices (Supplementary Fig. S4). This prolonged co-culture setting enabled us to test the persistence and stability of the SynCom in association with the plant. Samples of sand, rhizosphere, and root were collected and resolved by 16S rRNA gene sequencing. At 21 DPI (corresponding to 21, 24, and 29 DAG plants), B. distachyon leaves had brown tips in every group, and the shoots of 21 DAG group plants were smaller than those of 29 DAG group plants (Supplementary Fig. S5). Similar to what was observed in hydroponic EcoFABs (Fig. 2), the root system occupied most of the root growth chamber (Supplementary Fig. S5). The noninoculated B. distachyon plants, which served as a control, had phenotypes similar to those of the SynCom inoculated groups (Supplementary Fig. S6).

For each inoculation method, the Shannon diversity index of the root samples was significantly lower than that of the rhizosphere and sand samples (Fig. 6A). Among the root samples, the Shannon diversity index of the seed inoculation group was significantly higher than that of the other inoculation methods; this difference in diversity was not observed in the rhizosphere or sand samples. To explore whether the different inoculation practices resulted in microbial composition differences, we analyzed their beta diversity by PCoA test using Bray-Curtis dissimilarity (Fig. 6B). The top two axes (PCoA1 and PCoA2) explained 68.32% (50.95% in PCoA1 + 17.37% in PCoA2) of the variance, and the samples separated by their sample type (PERMANOVA, F = 34.25, P < 0.0001) (Supplementary Table S2) but not by the inoculation practice (PERMANOVA, F = 1.5436, P = 0.1469). These results indicated that when reaching a steady state, the microbial communities were very similar within sample types regardless of the inoculation practice.

Download as PowerPoint

To assess how sample type shaped the microbiome, we combined all the data from each sample type and plotted the relative abundance of the bacterial species. The relative abundance of Burkholderia OAS925 was positively correlated with the proximity to the plant, whereas the relative abundance of Variovorax OAS795 and Mucilaginibacter OAE612 was negatively correlated (Fig. 6C). While all bacteria detected in the liquid inoculant were detected in the rhizosphere at 21 DPI, Bradyrhizobium OAE829 and Paenibacillus OAE614 were not detected in the sand. Regarding the root samples, Bradyrhizobium OAE829, Paenibacillus OAE614, Chitinophaga OAE865, Brevibacillus OAP136, and Niastella OAS944 were not detected. These results indicated that different bacteria established distinct niches in the steady state microbial community.

To better understand how the plant roots shaped their microbial community, we performed pairwise comparisons of bacterial abundance between rhizosphere and sand samples as well as between root and rhizosphere samples (Fig. 7). Burkholderia OAS925 was significantly enriched in the rhizosphere compared with sand, whereas Variovorax OAS795 and Mucilaginibacter OAE612 were significantly diminished (DESeq2, Benjamini-Hochberg corrected P = 0.0329, 5.052 × 10^–14 and 0.0152, respectively) (Fig. 7A). In addition to Burkholderia OAS925, Rhizobium OAE497, Lysobacter OAE881, and Rhodococcus OAS809 were significantly enriched in the rhizosphere compared with the sand (DESeq2, Benjamini-Hochberg corrected P = 2.3561 × 10^–11, 0.0152, and 0.0021, respectively). These results agreed with our observations regarding the merged relative abundance (Fig. 6C). When comparing the root and the rhizosphere, Mucilaginibacter OAE612, Variovorax OAS795, Rhodococcus OAS809, and Rhizobium OAE497 were significantly diminished in the root compared with the rhizosphere (DESeq2, Benjamini-Hochberg corrected P = 1.180 × 10^–21, 1.7171 × 10^–8, 3.9037 × 10^–8, and 7.9962 × 10^–4, respectively) (Fig. 7B). The results suggested the root was a more selective niche than the rhizosphere.

Download as PowerPoint

Next, we evaluated the best inoculation practice in terms of plant fresh weight and microbial community reproducibility. The SynCom inoculation had no significant effect on plant biomass except for the root weight of the 8 DAG group, which was negatively affected by SynCom inoculation (t test, P = 0.0140) (Supplementary Fig. S7). We calculated the microbial community reproducibility using Euclidean distance as described previously (Song et al. 2021), and the Euclidean distance was negatively correlated with the reproducibility. There were no significant differences between the sand and root samples (t test, P = 0.074 and 0.160, respectively) (Fig. 8A and C), indicating the reproducibility of the sand and root samples was not significantly different between inoculation practices. However, in rhizosphere samples, the 3 DAG group had the lowest median, which was significantly lower than that of the 0 DAG group (t test, P = 0.0046) (Fig. 8B). Therefore, the 3 DAG group had the highest reproducibility.

Download as PowerPoint

The microbial community in the EcoFAB had a persistent association with the B. distachyon root in the most reproducible setup.

After identifying the most reproducible setup of the EcoFAB−B. distachyon−SynCom system (i.e., inoculating SynCom at 3 DAG and harvesting it at 21 DPI), we addressed SynCom persistence with B. distachyon plants. We repeated the experiment with the most reproducible setup described earlier and calculated the CFUs at 0 and 21 DPI. As a control, EcoFABs without a plant were also included, and CFUs were collected at 6, 14, and 21 DPI in case the CFUs were too low to be detected at 21 DPI.

In EcoFABs with B. distachyon harvested at 21 DPI, the density of the SynCom in the rhizosphere (8.16 log CFU/g) and the root (9.02 log CFU/g) was significantly higher than that observed in the control EcoFABs without plants (7.57 log CFU/g) (P = 0.0104 and 8.000 × 10^–7, respectively) (Fig. 9). Moreover, the viable cells in sand from EcoFABs with (7.82 log CFU/g) and without plants showed no significant differences (P = 0.5604). The viable cell density from all samples collected at 21 DPI was significantly higher than that of the inoculant at 0 DPI (5.07 log CFU/g, P < 1.000 × 10^–8 in all pairs).

Download as PowerPoint

In the no-plant control group, the SynCom density increased more than 100 times from 0 DPI (5.07 log CFU/g) to 6 DPI (7.27 log CFU/g) and stayed at similar levels at 14 DPI (7.18 log CFU/g) and 21 DPI (7.57 log CFU/g) (Fig. 9). The control group did not have a plant to provide an additional carbon source. To evaluate whether the increased cell density could have resulted from bacterial growth on the 0.5× MS salts used in the EcoFAB, we tested the growth of each individual strain on such medium. Of the 17 strains, only Arthrobacter OAP107, Bacillus OAE603, Burkholderia OAS925, Mucilaginibacter OAE612, Mycobacterium OAE908, Rhizobium OAE497, and Rhodococcus OAS809 grew on 0.5× MS salts (Table 1). Therefore, cell growth in the no-plant control group held true only in these bacteria.

DISCUSSION

SynComs are valuable tools for exploring the complexity of interactions between microbes, plants, and the environment. In the last decade, numerous studies have given more insight into rhizosphere dynamics and structure using SynComs. However, more systematic and standardized methodologies are needed to harness the full potential of SynCom tools. A systematic review of the use of SynComs to understand the relationship between plants, microbes, and the environment was recently performed (Marín et al. 2021). The authors found that most SynCom studies published thus far use Arabidopsis thaliana as a plant model, with far fewer using economically important food or biofuel crop models such as B. distachyon. Additionally, little standardization and comparison between protocols were observed in the methodologies (Marín et al. 2021). Most SynCom studies focus on microbial abundance composition and in planta effect. Studies assessing the impact of inoculation method, plant age, or harvesting point in the microbial community are scarce.

Our current research provides a baseline and foundational information for using the EcoFAB−B. distachyon−SynCom as a model fabricated ecosystem for plant–microbe interaction studies, with a particular focus on how inoculation practices impact microbiota assembly and community stability in a small-scale fabricated ecosystem. The ideal system has persistent plant–microbe associations and stable microbial community structure and produces reproducible plant phenotypes. Using EcoFABs, we tested how harvesting time points (14 DAG or 21 DPI) and different inoculation methods (seed inoculation-0 DAG, transplant inoculation-3 DAG, and seedling inoculation-8 DAG) affect the microbial community assembly.

Our data showed that all bacteria from the initial inoculant were detected in the rhizosphere samples at 21 DPI, demonstrating that the SynCom had a persistent association with B. distachyon for 21 DPI regardless of inoculation method. Moreover, the microbial community of each sample type (i.e., sand, rhizosphere, root) was indistinguishable between inoculation practices. Among the samples collected at 21 DPI, the transplant inoculation (3 DAG group) showed the most reproducible microbial community and a neutral plant growth effect. Therefore, we concluded that transplant inoculation was the best inoculation practice for studying the steady-state microbiome in our system.

The sand-filled EcoFAB−B. distachyon−SynCom is a suitable benchtop fabricated ecosystem model.

The results obtained in this study show that the sand-filled EcoFAB−B. distachyon−SynCom system aligns with conventional plant growing methods and soil microbial community assembly data. On the plant side, we showed that the developmental stage of B. distachyon grown in an EcoFAB (Fig. 2) corresponded well to the conventional growth method (Hong et al. 2011) (Supplementary Fig. S2). When comparing the B. distachyon plants in sand-filled EcoFABs with the plants in the hydroponic setup, they both had three unfolded leaves at 14 DAG and a flag leaf sheath extending at 28 DAG (Fig. 2; Supplementary Figs. S3, S6, and S7). This demonstrated that the developmental stages of B. distachyon grown in sand-filled EcoFABs aligned well with both hydroponic EcoFABs and conventional growth methods.

It was challenging to directly compare the B. distachyon developmental stages between EcoFABs (hydroponic and sand) and soil systems at 21 and 24 DAG when the plants were in the stem elongation stage (BBCH 3, corresponding to 15.7 to 24.0 DAG in soil) (Hong et al. 2011). This was because BBCH 3 defined each substage by the length of an internode, but the internode elongation was largely restricted by the height of the Magenta box in which the EcoFAB was placed. Comparing the B. distachyon stage at 21 and 24 DAG using the number of tillers (BBCH stage 2) also was not practical. Hong et al. (2011) reported that secondary tillers occurred at approximately 30 DAG. While this aligned well with our observation that the first tiller emerged at 14 DAG and the second tiller emerged at 28/29 DAG (Fig. 2; Supplementary Fig. S6C and F), it was not a suitable factor for defining 21 and 24 DAG.

Regarding the SynCom population, CFU counting demonstrated that in the presence of B. distachyon, the SynCom population increased 1,000 times in 21 DPI compared with growth without a plant. Moreover, the SynCom population was strongly and positively correlated with its proximity to B. distachyon roots, and the numbers of viable cells in the root and rhizosphere were significantly higher than those of the control group at 21 DPI. The data not only suggest that the B. distachyon root supported the microbial community but also provide direct evidence that the SynCom established a persistent association with B. distachyon in our system for at least 21 days.

In terms of microbial community, our data showed that Burkholderia OAS925 (order Burkholderiales) and Lysobacter OAE881 (order Xanthomonadales) were enriched in rhizosphere samples at 21 DPI (Fig. 7). Our results agree with observations that Burkholderiales and Xanthomonadales are enriched in Brachypodium rhizosphere microbial communities collected from agricultural soil (Kawasaki et al. 2016). One reason why these two orders are usually abundant in the plant environment is because they are fast-growing and metabolically adaptable bacteria. Our data are also in line with a recent study using the same SynCom in hydroponic EcoFABs, in which it was shown that Burkholderia OAS925 dominated the rhizosphere microbial community at 7 DPI (Coker et al. 2022). The authors also showed that when the SynCom was mixed at a 1:1 ratio based on OD₆₀₀, Marmoricola OAE513 and Mycobacterium OAE908 had very low relative abundance, and both were undetectable 3 days after SynCom mixing. This indicated that Marmoricola OAE513 and Mycobacterium OAE908 were more difficult to detect and/or had a survival disadvantage under 1:1 mixing conditions, which may explain why both bacteria were not detected in our initial inoculant.

The SynCom microbial community composition changes in the absence of a plant are not merely due to the differential persistence of certain bacteria.

Our data showed that in the absence of a plant, the SynCom population increased more than 100 times within a week and reached levels comparable to those observed in plants at 21 DPI (Fig. 9). This growth demonstrates that the microbial community assembly in the EcoFABs without plants was not simply due to the differential persistence of certain strains, but rather to new bacterial growth.

Seven SynCom members were able to grow on 0.5× MS salts, suggesting the nutrient source came from the medium rather than the quartz sand. Four of the growing bacteria were among the top six abundant species in the microbial community: Burkholderia OAS925, Mucilaginibacter OAE612, Rhizobium OAE497, and Rhodococcus OAS809 (Fig. 3A; Table 1). Arthrobacter OAP107 and Bacillus OAE603 grew well on 0.5× MS salts, but their relative abundance in the SynCom was less than 0.2% at 21 DPI, and Mycobacterium OAE908 could not be detected in the liquid inoculant (Fig. 6C). It is possible that these strains can utilize the ethylenedinitrilotetraacetic acid (EDTA) in 0.5× MS salts as the carbon source, as previously reported (Bucheli-Witschel and Egli 2001; Chen et al. 2005; Oviedo and Rodríguez 2003). By contrast, Variovorax OAS795 and Methylobacterium OAE515 were among the top six abundant species in the microbial community at 21 DPI but did not grow on 0.5× MS salts alone (Table 1). The results strongly suggest that the SynCom assembly in the EcoFAB without a plant was not merely due to the ability of specific bacteria to survive on 0.5× MS salts.

The most likely explanation for the population growth in the SynCom without a plant is cross-feeding by other members in the community, which could support bacterial growth to a degree comparable to that observed with a carbon source (Goldford et al. 2018; Murillo-Roos et al. 2022). In the stable SynCom, Variovorax OAS795 (order Burkholderiales) dominated the population (>55% of the population) even though it was not able to survive on 0.5× MS salts (Fig. 3A). The ability of the Variovorax strain to outcompete the rest of the SynCom members must depend on its metabolic adaptability. Members of the genus Variovorax have indeed been described as metabolically versatile: Variovorax possesses extraordinary degradation abilities and can utilize natural compounds produced by other bacteria, such as acyl homoserine lactones, as carbon sources (Leadbetter and Greenberg 2000). Additionally, genomic analysis of several Variovorax strains revealed metabolic features of an autotrophic lifestyle (Han et al. 2011). We speculate that EDTA is the “keystone carbon source” for the community and the byproducts were consumed largely by the metabolic generalist Variovorax OAS795.

It is noteworthy that the number of viable bacteria in EcoFABs in the presence and absence of plants 21 DPI was comparable, but their microbial community composition differed dramatically (Figs. 3A, 6B, and 9). This suggests that the presence of the plants influences the composition of the microbial community and that SynCom without a plant is likely shaped by cross-feeding within the community, as discussed earlier. It would be interesting to further explore the metabolites that are critical in shaping this cross-feeding network.

The presence of B. distachyon plants influences the composition of the microbiome.

In the presence of a plant, Burkholderia OAS925 dominated the population in the rhizosphere and the root (Fig. 6C). Burkholderia species are reported among the dominant bacteria in the rhizosphere of several plants, including B. distachyon, wheat (Triticum aestivum), maize (Zea mays), and sorghum (Sorghum bicolor) (Donn et al. 2015; Hallmann et al. 1999; Kawasaki et al. 2016; Li et al. 2014; Parke and Gurian-Sherman 2001; Schlemper et al. 2017). The concentration of Burkholderia species in plant roots has been estimated to be 100 times higher than that in bulk soil (Pallud et al. 2001). Their versatile metabolism and adaptability allow them to degrade different compounds usually found in root exudates, such as organic acids, sugars, and polyalcohols (Compant et al. 2008; Morya et al. 2020).

To better understand how plant roots shape the microbial community, we performed pairwise comparisons between rhizosphere and sand as well as between root and rhizosphere. We found that Burkholderia OAS925 was significantly more enriched in the rhizosphere than in the sand, whereas Variovorax OAS795 and Mucilaginibacter OAE612 were significantly diminished (Fig. 7). Aside from Burkholderia OAS925, Rhizobium OAE497, Lysobacter OAE881, and Rhodococcus OAS809 were also significantly more enriched in the rhizosphere than in the sand. The results matched well with our observations regarding the relative abundance (Fig. 6C). These differences between sample types were expected because the Brachypodium plant produces compounds that affect soil microbe growth and metabolism (Kawasaki et al. 2016; Sharma et al. 2020). Coker et al. (2022) also observed that Burkholderia OAS925 and Rhizobium OAE497 increased in relative abundance after growing in the presence of B. distachyon. Due to the ability of Variovorax OAS795 and Burkholderia OAS925 to efficiently outcompete the rest of the SynCom members in the absence or presence of a plant respectively, these two strains are interesting targets for future research in order to identify the genomic traits that allow such competitive behaviors.

Inoculation practices affect microbial communities.

Inoculation methods are an important factor contributing to plant–microbe interactions, especially when harvest occurs early (Bashan 1998; dos Santos Lopes et al. 2021; Kumar et al. 2022). However, comprehensive comparisons between different inoculation methods are scarce. The present study demonstrates that inoculation practices significantly shaped the 14 DAG microbial community but not the 21 DPI samples. A recent study on sorghum (S. bicolor) also showed that inoculation practices impacted the total abundance of different bacterial species at an earlier harvesting point but not at a later harvesting point (Chai et al. 2022). This highlights the importance of incorporating different inoculation methods into experimental designs and the need to compare multiple inoculation methods when studying plant–microbe interactions.

Our results from 14 DAG plants suggest that microbial inoculation practices influence the dynamics of plant–microbe interactions, especially in the rhizosphere. The seedling inoculation method (8 DAG) demonstrated a significantly higher abundance of Burkholderia OAS925 and Lysobacter OAE881 and significantly lower Rhodococcus OAS809 compared with other inoculation methods at 14 DAG (Fig. 5). This pattern indicates that the seedling inoculation method had the most distinct microbial community compared with the other inoculation methods. The fact that inoculation methods significantly shaped microbial communities may be explained by the dynamic profiles of B. distachyon exudates and the response of the SynCom to root exudates (McLaughlin et al. 2023; Saleh et al. 2020; Sharma et al. 2020; Zhalnina et al. 2018). B. distachyon exudate profiles can vary significantly within hours (McLaughlin et al. 2023) and weeks (Zhalnina et al. 2018). Thus, the exudate concentration and profile of the 8 DAG group may be significantly different from those of the other groups. It has been shown that B. distachyon root exudates affect bacteria with regard to their chemotaxis and gene expression within 48 h of root exudate addition (Saleh et al. 2020; Sharma et al. 2020). Therefore, the differential abundance of SynCom members observed with the different inoculation methods could be attributed to SynCom–root exudate interactions.

Transplant inoculation coupled with a longer co-culture period is the best practice for the EcoFAB−B. distachyon−SynCom fabricated ecosystem.

This study aimed to establish a protocol that provides the most reproducible and reliable plant–microbe interaction results to serve as a baseline for standardized, small-scale fabricated ecosystem studies. Here we showed that the SynCom associated with B. distachyon persistently for 21 days, much longer than previously reported (Coker et al. 2022). Moreover, the microbial community reached the same steady state at 21 DPI regardless of the inoculation method, demonstrating the stability of the fabricated ecosystem used in this study. Future work could expand upon this result to test whether inoculation practices affect long-term microbiome assembly by collecting all samples at 29 DAG.

Among the inoculation methods tested in this study, transplant inoculation (3 DAG group) showed the most reproducible within-group microbial community and a neutral plant growth effect. Transplant inoculation is the most commonly used method for lab studies in part due to its being less labor-intensive (Herrera Paredes et al. 2018; Hu et al. 2020; Schillaci et al. 2021). We conclude that transplant inoculation coupled with a longer co-culture period (late harvesting point) is a good strategy for conducting benchtop, large-scale fabricated ecosystem studies.

Conclusions.

Our results for both plant growth and microbial community abundance suggest that the sand-filled EcoFAB−B. distachyon−SynCom system is highly comparable to conventional plant growing methods and thus is a suitable fabricated ecosystem for future studies. Additionally, we showed that transplant inoculation is the best application procedure for studying the steady-state microbiome in the sand-filled EcoFAB−B. distachyon−SynCom system. This research provides a baseline and foundational information for using our fabricated ecosystem to study plant–microbe–environment interactions.

Data availability.

The 16S rRNA gene fragment reads generated through this study can be found under National Center for Biotechnology Information BioProject accession number PRJNA975970. All code used in this study can be accessed through GitHub at https://github.com/JennyMortimer/EcoFAB-Brachypodium-rhizosphere.

ACKNOWLEDGMENTS

We thank John Vogel for providing Brachypodium distachyon Bd21-3 seeds and the QB3 Core Research Facility at the University of California, Berkeley for sequencing.