Synthetic bacterial community colonizes wheat roots grown in soil and mimics the assembly pattern of a field community in a cultivar dependent manner

Abstract

The root microbiome is important for plant development. The impact of the root microbiome is the result of multiple complex interactions among microorganisms, the plant and the environment. This complexity can be reduced by designing synthetic bacterial communities (SynComs) consisting of bacteria isolated from the roots, making it possible to study these interactions. However, the translational power from SynCom experiments to explain field observations is still very low, which demonstrates the need for development of SynComs that colonize plants comparable to what is observed in the field. Hence, we developed a SynCom consisting of 13 different strains from 13 genera with varying phenotypes originating from the roots of winter wheat (Triticum aestivum cv. Sheriff). The SynCom was inoculated into gamma-irradiated soil prior to sowing and community assembly was determined over 4 weeks using 16S rRNA amplicon sequencing. Winter wheat plants inoculated with the developed SynCom grew comparably to plants inoculated with a natural community (NatCom) obtained from a soil solution over the 4-week period. Furthermore, the temporal dynamics of the majority of the SynCom strains mimicked the development in relative abundance of their respective genera in field grown winter wheat of the same cultivar. However, this could not be translated to a different cultivar (Heerup). Our results demonstrate how SynComs inoculated into gamma-irradiated soil can provide a promising framework to study root microbiome assembly and relate the findings to field observations. At the same time, it highlights the plant-genotype specific impact on community assembly.

Introduction

Microorganisms interact with plants above- and belowground and are important for plant health and development, as they provide beneficial functions such as nutrient acquisition, disease suppression, or reduction of abiotic and biotic stresses [1–3]. However, the processes governing root microbial ecology and community assembly are not fully understood, despite an enormous effort being made to advance this field [4, 5]. In the rhizosphere, this is primarily due to the high chemical complexity [6] and microbial diversity [7].

The seed carries the starting members of the rhizosphere microbiome, yet the largest proportion originates from the soil [8, 9], with the initial assembly of the rhizosphere community being primarily driven by stochastic effects such as dispersal and drift [10]. At later developmental stages deterministic processes are becoming more important for community assembly [4, 10, 11]. While studies have shown that the variability in soil microbiomes leads to differences in the root microbiome of the same plant species across soil types, our knowledge on microbial community development is primarily based on observational studies that offer limited information on the causal effects. To leverage our understanding of root microbiome ecology and assembly, and establish causal relationships, it is crucial to develop and use reproducible model systems with reduced complexity, which still reflect the actual soil systems. Hence, synthetic communities (SynComs) have gained interest as model systems, providing the possibility to test hypotheses on causality in plant-microbe interactions [12, 13].

SynCom studies have been used to elucidate plant-microbe interactions on Arabidopsis thaliana [14–16], Lotus japonicus [15], maize [17] and tomato [18] revealing interesting results of priority effects [14, 15], the importance of bacterial-fungal interaction in pathogen suppression [18], and the identification of Enterobacter cloacea as a keystone species in the maize rhizosphere [17]. Despite the importance of these studies for fundamental understanding of mechanisms involved in root-microbe interactions, they have been performed solely in media [19], gnotobiotic systems [15, 20, 21] or artificial soil mixtures [18], thereby omitting the complexity of soil with varying niches, physical hindering for motility and chemical gradients.

SynComs applied to soil systems have been performed, but have focused solely on plant growth promotion [22, 23] with descriptive effects of SynCom introduction on microbial community dynamics in the root zone [24, 25]. Furthermore, the concept of SynEco [26] where the reductionistic principles of SynComs are used under increased complexity has recently been proposed to facilitate translation of ecological findings to natural systems. Still, soil-based systems with predictive power of microbial ecology and plant-microbe interactions at the mechanistic level under natural conditions are currently lacking.

While SynComs do not represent the full microbial diversity of a given environment, individual and rhizo-compatible strains can be used as representatives for a genus in studies of rhizosphere community assembly, as carbon utilization spectrum and growth rate are important factors governing bacterial success and deterministic outcomes in plant colonization [27, 28], and is strongly conserved at the genus level [29], even though high microdiversity is found within genera colonizing the rhizosphere [30]. This suggests that strains within a genus could be interchangeable in community assembly without affecting the genus abundance. This is supported by the notion that core bacteria are often being classified at the genus level across different soil types [31]. Accordingly, we hypothesized that SynCom members of a robust SynEco system would assemble on the roots of wheat in the same way as a natural community from a different soil type.

Recent studies have shown that even closely related wheat genotypes assemble rhizosphere microbial communities differently [30, 32]. Hence, we further hypothesized that the microbial dynamics of the SynCom under controlled conditions would resemble the dynamics in the same cultivar of field grown wheat plants, but differently from another cultivar.

To test our hypotheses, we designed a root-compatible SynCom consisting of 13 bacterial strains representing 13 bacterial genera isolated from the roots of winter wheat (Triticum aestivum cv. Sheriff). These genera comprised both rare and core members of the wheat rhizosphere [31]. We created a soil-based system, by inoculating the SynCom to gamma-irradiated soil prior to wheat seedling transplantation to enable natural assembly processes and active colonization of the bacteria to the growing roots. The microbial assembly of the SynCom on wheat roots was subsequently compared to the community assembly based on a natural community (NatCom) from a different soil type as well as field grown wheat plants.

Our work presents the development of a robust SynCom-based model system, where microbial dynamics on wheat roots mimic microbial assembly dynamics under field conditions.

Materials and Methods

Strain isolation and selection

Winter wheat, cv. Sheriff, was grown in a growth chamber under controlled conditions as previously described [33]. Three plants were harvested on day 142 at the flag leaf emergence stage (Feekes 8.0), and rhizoplane bacteria were isolated to generate a biobank of 1344 bacterial isolates (Supplementary methods). A top-down approach for designing our SynCom, with the aim to mirror the complexity of the community in the wheat rhizoplane, was then used for strain selection based on two criteria: (i) High abundance in rhizoplane/rhizosphere communities based on previously published data from the same wheat genotype [32, 33] (Table 1), (ii) rare taxa, due to their potential important role in microbial communities [34]. The two genera, Pseudomonas and Arthrobacter, were included as they were classified as core members of the wheat rhizosphere [31]. In addition Variovorax was included due to its importance for root growth as previously shown [35]. With the aim of determining if temporal dynamics of single strains would mimic the dynamics of genera under field conditions, only one strain from each selected genus was included. The resulting synthetic community (SynCom) consisted of 13 bacterial strains (Table S1). The SynCom strains’ abilities of phosphate solubilization, phytate degradation, nitrate reduction and siderophore production were determined (Supplementary methods).

Table 1.

Mean relative abundance of SynCom genera detected by 16S rRNA gene sequencing in two previous studies.

	Day 14 (Guan et al., 2024; n = 5)		Day 142 (Zervas et al., 2022; n = 10)
Genus	Relative abundance (%)	Rank abundance	Relative abundance (%)	Rank abundance
Flavobacterium	0.09	78	6.3	1
Pedobacter	ND		3.7	6
Pararhizobium	0.4	39	1.91	9
Bacillus	7.9	2	0.73	24
Stenotrophomonas	0.31	44	0.57	31
Pseudomonas	1.5	11	0.50	34
Microbacterium	0.27	52	0.20	77
Variovorax	ND		0.14	95
Arthrobacter	1.0	16	0.10	126
Paenarthrobacter	0.07	83	0.07	167
Ensifer	ND		0.04	259
Agrobacterium	ND		ND
Peribacillus	ND		ND
Total	11.5		14.2

Rank abundance of the SynCom genera found in the total community. Days indicate days after sowing. ND: Not detected.

Soil description

We collected soil from the Long-Term Nutrient Depletion Trial (LTNDT) field in Taastrup, Denmark (55° 40′N, 12° 17′E). The field is classified as a sandy loam soil, and composed of 164 g kg⁻¹ clay, 173 g kg⁻¹ silt, 333 g kg⁻¹ fine sand, 312 g kg⁻¹ coarse sand and 17 g kg⁻¹ organic matter [36]. Soil was collected for the SynCom plant experiment in the spring of 2018 from the nitrogen fertilizer only (N1) treatment. After collection the soil was lightly dried and exposed to gamma radiation (2 × 18 kGy, 10 MeV electron beam) at Sterigenics A/S, Denmark, to avoid the release of phytotoxins through heating, and stored at room temperature in sealed plastic bags until use. This reduced the DNA concentration by 79% in the gamma-irradiated soil compared to untreated field soil (Table S2), and resulted in a bacterial community primarily consisting of Streptomyces and Bacillus (Fig. S1). A bacterial community of 3.0 * 10⁵ CFU/g soil was detected post gamma-irradiation.

SynCom plant experiment

Wheat plants were grown in PVC pots as described in Guan et al. [32] (Fig. S2). Gamma irradiated soil (sieved 4 mm) was mixed with sand of grain size of 0.4–0.9 mm (Dansand, Denmark) in a 3:2 soil:sand ratio. The sand was dried for 24 h at 105°C and autoclaved twice at 121°C for 15 min prior to mixing. Soil samples (n = 3) were collected to determine the bacterial communities in the starting soil, and stored at −20°C until DNA extraction. Soil moisture content was adjusted to 17% w/v with autoclaved MilliQ water. We added 0.79 ml liquid 3–1-4 NPK fertilizer (Park® Drivhusgødning, Germany) to each pot by diluting it in the water used to adjust moisture content.

Seeds of the winter wheat cultivar Sheriff were sterilized using 70% ethanol and 5% sodium hypochlorite (Acros Organics, Belgium) for 3 min each wash, followed by three rinses with sterile MilliQ water. No colonies appeared after checking the final wash on TSA plates. Afterwards, we soaked the seeds in sterile MilliQ water for 1 h and left them to germinate on moist sterile filter paper at room temperature for 4 days in the dark prior to planting.

To prepare the SynCom inoculum, we washed overnight cultures of the SynCom strains twice, and adjusted OD₆₀₀ to 0.6 in 0.9% sterile NaCl. We then mixed the washed cells of the 13 SynCom strains to form the final SynCom inoculum. The NatCom inoculum was created by extracting a native soil microbial community from a non-agricultural soil, collected 3–5 cm below the surface at the Frederiksberg Campus (55°41′00.5”N 12°32′35.5″E, Frederiksberg, Denmark) and stored it at 5°C until use. The microbial community was extracted as in Yan et al. [37] with a few modifications. Soil was sieved (4 mm) and 20 g was suspended in 190 ml 0.9% NaCl and blended 2 × 2 min. A 1000-fold dilution was made in 0.9% NaCl to prepare the NatCom inoculum.

We added either 2.5 ml NatCom or SynCom inoculant to a two-centimeter hole in the soil in each PVC pot, and transplanted one pre-germinated seed to each pot and buried it, before adding one milliliter of sterile MilliQ. The plants were grown for 1–4 weeks in a non-sterile climate chamber with a 16/8-h day/night cycle at 400 μmol m-² s⁻¹ (BX-series LED bars, Valoya, Finland). Temperatures were 20/18.5°C (day/night), with a constant 70% relative humidity. We watered the plants from the top with sterile MilliQ water every second day. Pots were moved around in the climate chamber when watering.

We harvested five plants from each treatment for sampling of the rhizoplane communities 7, 14, 21, and 28 days after sowing. One plant comprised one sample. In total, we obtained 40 samples (2 treatments × 4 time points × 5 biological replicates). We measured the root and shoot length from the base of the roots. We sampled the rhizoplane as described in Guan et al. (2024) (Supplementary methods). To determine the grow-back in the gamma irradiated soil, three pots without seeds were sampled each week, and DNA concentration was measured (Table S2).

Field experimental setup

Seeds of winter wheat cultivars Sheriff and Heerup were sown at the LTNDT field, Taastrup, Denmark, in four spatially separated plots (3 × 12 m) (Fertility level: 120 kg N, 40 kg P, 240 kg K ha⁻¹ y⁻¹ (N₁P₂K₂) fertilizer) on September 26th, 2023. In 2023, the field plots had only received 120 kg N ha⁻¹. Each of the four plots were divided into two separate parcels, each parcel sown with several rows of one cultivar.

Nine days after sowing, plants were collected from each of the four plots at random for both cultivars. We separated the roots and shoots at their base with a sterilized scalpel, and subjected the roots to a two-step washing process to obtain rhizoplane samples. The rhizoplanes of five plants were pooled to one sample (4 replicates per cultivar). We repeated this sampling procedure at 16, 21 and 28 Days after sowing (DAS), except only three plants instead of five per treatment were pooled for one sample, due to the larger plants. In total, we collected three bulk soil samples prior to sowing and 32 rhizoplane samples (4 time points, 2 cultivars and 4 replicates). All equipment was cleaned with 70% ethanol between each sample. The samples were stored and freeze-dried as described for the SynCom experiment (Supplementary methods).

Bacterial community structure

We extracted DNA from all samples using the Fast-DNA™ Spin Kit for Soil (MP Biomedicals, Irvine, CA, USA) by following the manufacturer’s instructions, with minor modifications (Supplementary methods). We prepared sequencing libraries using the 799F and 1193R primers targeting the V5-V7 regions to reduce amplification of plant host DNA [38]. ZymoBIOMICS Microbial Community DNA Standards (Zymo Research, Irvine, CA, USA) was used to check for contamination during sequencing. Finally, indexing and sequencing was done by Eurofins Genomics (Cologne, Germany) on the Illumina MiSeq platform (2 × 300 bp).

Amplicon processing and data analyses

Data analysis was performed using R 4.2.1 [39]. Ampvis2 v.2.8.9 [40] and ggplot2 v.3.5.0 packages [41] were used for plotting. All of the 16S rRNA gene amplicons were processed using dada2 v.1.32 [42] as detailed in [32]. Taxonomy was assigned using the SILVA database v.138.1 [43]. We curated the taxonomic assignment through alignment of the sequences of amplicon sequence variants (ASVs) found in the SynCom inoculum with the 16S rRNA gene sequences of the strains. In the SILVA v.138.1 database, Peribacillus simplex is classified as Bacillus simplex, and thus we corrected this in our dataset. Non-bacterial, Mitochondria and Chloroplast ASVs were removed prior to downstream analyses. Mock communities and negative controls were inspected for potential contamination.

Samples were rarefied to even depth (minimum read number = 20 040 reads). This was repeated 100 times, as recommended [44]. Subsequently, the Alpha diversity (Shannon index) was determined using the plot_richness function phyloseq v1.40 [45]. Beta diversity was calculated using Bray–Curtis dissimilarities with the avgdist() function in vegan v. 2.6.4 [46], rarefying each sample to the minimum read number (20 040 reads) of rhizoplane samples 100 times (i.e. excluding bulk soil samples and controls). The difference in Bray–Curtis dissimilarities within groups between SynCom and NatCom inoculated plants were compared using a Wilcoxon rank-sum test, and corrected for multiple testing with Bonferroni correction. Principal coordinate analysis (PCoA) was performed using pco() in ecodist v.2.1.3 [47]. PERMANOVA was used to determine the importance of sampling time (as days after sowing) and cultivar on community composition with adonis2() in vegan, with 999 permutations and controlling for field plot using the permute v.0.9–7 [48]. Comparisons among different time points for the SynCom communities were performed using the pairwise.adonis2 function [49]. Differences in shoot and root length were compared with a t-test (P < .05).

To determine the origin of the ASVs (i.e. inoculum or soil), we identified all ASVs from the starting inoculum or gamma-irradiated soil. To track the SynCom strains, we sub-setted the ASVs found in the SynCom inoculum from the SynCom rhizoplane samples and summed their relative abundances for each SynCom strain. Upon comparison with NatCom and field samples, we extracted all ASVs belonging to the 13 SynCom genera.

Testing for equal signs/stats on temporal dynamics

The relative abundance was analyzed separately for each genus by ANCOVA models with categorical variable Treatment (SynCom, NatCom, Field Sheriff, Field Heerup) and numerical variable (Days) as the covariate. The first time-point (Days 7 or 9) was excluded due to the high dissimilarity among samples. The analyses were done on a logarithmic scale as this gave statistically valid models as validated by residual plots and normal quantile plots of the residuals. In summary, we used the statistical model

where the intercepts α(SynCom), α(NatCom), α(Field Sheriff), α(Field Heerup), the slopes β(SynCom), β(NatCom), β(Field Sheriff), β(Field Heerup), and the error variances are fitted separately for each genus. Some genera had a relative abundance of zero. As the logarithmic transformation cannot be used on zero, we estimated the ANCOVAs as Tobit models [50], where a zero is interpreted as the observation that this abundance is lower than all the non-zero abundances within that genus. There were rather few zero abundances, so the data would not be sufficiently rich to estimate separate models for the zeros, e.g. as required for hurdle models. Moreover, the Tobit model provides single slope estimates that allow us to study the overall direction of the community assembly in accordance with the objective of our study. In conclusion, we settle with the Tobit model for pragmatic reasons, although we acknowledge that there could be separate biological mechanisms underlying whether genera are present or not.

To establish statistical evidence that a genus is either increasing for both SynCom and NatCom or decreasing for both SynCom and NatCom, we tested the null hypothesis that the corresponding slope parameters β(SynCom) and β(NatCom) had different signs, i.e.

We have not been able to find a statistical test for this hypothesis in the literature. Hence, we have constructed a new test named the Conservative One-Sided Test (COST) to test this hypothesis. The COST procedure takes its onset in T-test statistics and for the null hypotheses and , and the associated two-sided P-values and . Based on this, the P-value for the COST is given as the maximum of and if the estimates for the slopes and are of the same sign, and the P-value equals 1 if the slope estimates have different sign in agreement with the null hypothesis for the COST. The name Conservative One-Sided Test arises since it is conservative (it takes a maximum) while using one-sided tests instead of two-sided tests. This is like the Two One-Sided T-test (TOST) [51] known from equivalence testing. We note that the P-value for the COST hypothesis can be computed by combining the P-values for the standard two-sided T-tests for and , and hence be computed in statistical software like R. Mathematical details on the COST procedure are provided in the Supplementary methods. This supplement also contains formal mathematical proof for the control of the Type-I error, i.e. that the probability that the null hypothesis is rejected is at most the claimed significance level if the null hypothesis is true.

Results

Designing the SynCom

We established a 16S rRNA genotyped diverse bacterial culture collection isolated from the rhizoplane of winter wheat (T. aestivum cv. Sheriff). Ultimately, the SynCom comprised 13 strains from 13 genera accounting for 12% and 14% of the genera detected in the rhizoplane community of soil-grown wheat after 14 and 142 days of growth, respectively [32, 33] (Table 1). The selected strains were tested for various phenotypic traits to ensure a diverse functional potential of relevance for plant-microbe interactions (Table S1, Figs S3–S5).

SynCom inoculation and plant development

The stable establishment and succession of the SynCom members on the wheat rhizoplane is crucial for further inferences on microbial community development and function, and derived effects (beneficial or detrimental) on plant health and performance. Here, we used gamma-irradiated soil, inoculated with the SynCom in the top soil, to strongly reduce the abundance of indigenous soil microorganisms, which would otherwise impede identification of the SynCom members during experiments. None of the ASVs in the starting SynCom, except for one ASV (Peribacillus simplex ASV13), were found in the gamma irradiated soil prior to inoculation, highlighting the applicability of the gamma-irradiated soil. In addition, none of the six genera, Agrobacterium, Microbacterium, Paenarhtrobacter, Pedobacter, Pseudomonas and Stenotrophomonas) were detected in the gamma-irradiated soil.

Plants inoculated with the SynCom developed in a uniform way across replicates, for both shoot and root length (Fig. 1), and no indications of nutrient limitations or pathogen attack were observed during the four weeks of growth. The shoot height and root length reached ~40 cm and ~17.5 cm after 28 days of growth, respectively. Further, we found no differences in shoot or root length between SynCom and NatCom (t-test, P > .05) nor were there any visual differences on plant growth and health between the inoculations (Fig. S6).

Figure 1.

Shoot (A) and root (B) length at the different sampling times for wheat plants inoculated with SynCom (blue) or NatCom (brown). Each point represents one plant. The error bars are the standard deviation.

SynCom strains successfully colonized the roots of wheat

SynCom strains were tracked in the rhizoplane at four time points, based on ASVs from 16S rRNA gene amplicon sequencing of the SynCom inoculum. Each SynCom member consisted of 1–5 ASVs (median = 2) (Table S1). All SynCom strains successfully colonized the rhizoplane at Day 7 and all strains were, with few exceptions, part of the rhizoplane community at all sampling times (Fig. 2, Fig. S7).

Figure 2.

Relative abundance of the SynCom strains on the rhizoplane after SynCom inoculation over a 4-week growth period. The relative abundance of each strain is the sum of ASVs from each strain, found in the starting SynCom inoculum (n = 5 plants). The relative abundances of each replicate plant are shown as shaded circles and the mean relative abundance is shown as a cross.

We determined the proportion of each of the 13 SynCom genera originating from the starting inoculum based on ASVs (Fig. 3). For most of the genera, the majority of the ASVs came from the SynCom inoculum. This is despite an observed high diversity of Pseudomonas seed endophytes (Table S3). The exceptions were Bacillus and Pedobacter, which could originate from the gamma-irradiated soil or the seed. However, Pedobacter was not detected in the gamma-irradiated soil samples without plants. Bacillus was primarily consisting of non-SynCom ASVs at all four sampling time points, while Pedobacter showed a high proportion of ASVs not originating from the SynCom at Day 21 and 28. In summary, this demonstrates that the added SynCom strains actively colonized the roots.

Figure 3.

Origin of SynCom ASVs detected in the rhizoplane samples. The bars indicate the mean relative abundance (per day) of each genus from the total community encompassing all ASVs affiliated with the specific genus. Peribacillus simplex ASV13, found in both SynCom and gamma-irradiated soil is grouped under SynCom ASVs. ASVs matching SynCom ASVs are combined (teal). Other sources include soil or seed (red).

The SynCom community decreases in variability over time

The alpha diversity (Shannon index) determined based on all ASVs (i.e. including ASVs of seed and soil origin) increased over time in the SynCom inoculated plants (Fig. S8). Despite the initial difference in diversity between SynCom and NatCom inocula, no difference in alpha diversity between NatCom and SynCom treated plants was observed at any of the sampling days, although the variation among replicates was higher for NatCom inoculated samples (Fig. S8). At Day 7, the SynCom and NatCom communities were highly dissimilar among replicates (Fig. 4A). At Day 14, the dissimilarity among replicates differed between SynCom and NatCom (Wilcoxon rank-sum test, P < .001), whereafter communities from both treatments stabilized.

Figure 4.

(A) BrayCurtis dissimilarities of 16S rRNA amplicons among replicates of SynCom (blue) or NatCom (brown) communities at the four sampling days. Each point represents a comparison. The compositional dissimilarity was compared between within- group-pairs SynCom (blue) and Natcom (brown) (Wilcoxon rank sum test, *** = P < .001, Bonferroni correction). The boxplot shows the median (horizontal line), the interquartile range (25th–75th percentile), and the whiskers extend from the hinges to the largest value no further than 1.5x the interquartile range. (B) PCoA plot showing beta-diversity of bacteria at 7 (circles), 14 (triangles), 21 (squares) and 28 (plus) days after sowing in SynCom (blue) or NatCom brown) based on Bray–Curtis dissimilarities (ASV level).

The PCoA ordination showed a separation between SynCom communities from Day 14, and those from Day 21 and 28 samples (PERMANOVA, P < .01) (Fig. 4B), which could be linked to the reduction in relative abundance of Massilia between Day 14 and 21 (Fig. S9). While Massilia was not found in the soil, we isolated seed endophytic Massilia (Table S3). Overall, our data show that despite a lower alpha diversity of the inoculated SynCom compared to the NatCom (Fig. S8), the diversity measures of the SynCom community followed the same overall trends as the NatCom rhizoplane communities. Furthermore, the data indicate that seed endophytes comprise a large proportion of the rhizoplane microbiome during early plant development, before SynCom members are recruited and colonize the roots.

We compared the assembly and succession of the SynCom community with the NatCom community development by tracking all NatCom ASVs belonging to the 13 SynCom genera over time (Fig. 5). Two genera, Bacillus and Peribacillus peaked in relative abundance at Day 14. Most other genera had the lowest relative abundance at Day 14, followed by an increase in relative abundance to Day 21. The genera Bacillus, Flavobacterium, Microbacterium, Pararhizobium, Pseudomonas, Agrobacterium and Pedobacter accounted for similar amounts of the total abundance in both communities in the later time points (Fig. 5), despite the differences in relative abundance in the starting community (Day 0) (Fig. S10). Arthrobacter and Paenarthrobacter contributed more to the total community at the first three time points in the SynCom compared to the NatCom, while being comparable at Day 28. These results suggest that the selected genera can colonize a specific proportion of the rhizoplane irrespective of starting concentration.

Figure 5.

Log10-normalized read counts of all ASVs assigned to the SynCom genera in SynCom, NatCom and field samples. The inserted regression lines show the ANCOVA models fitted within genera from Day 14 to 28. Regression lines are marked based on the outcome of the COST procedure. SynCom regression lines are dashed black, as are regression lines for NatCom and field samples with significant same signs as in the SynCom setup. Regression lines for NatCom and field samples with same signs, but non-significant as compared to the SynCom setup are black. Non-significant different signs as compared to the SynCom setup are represented as dashed grey lines. The genera are ordered according to statistical evidence for the same sign of the relative abundance for SynCom and Field Sheriff samples (see Table 3 below) using the COST procedure.

The temporal dynamics of the relative abundance of each SynCom member was compared to the corresponding genera in the NatCom with a new statistical method, COST (Conservative One-Sided Test). There was evidence (marginal P-values below 0.05) for the same direction of the temporal dynamics for four genera; Bacillus and Peribacillus were decreasing, and Microbacterium and Variovorax were increasing (Table 2). The decrease of Bacillus and the increase of Variovorax reflects their dynamics in previous studies (Table 1). A similar COST procedure can be used to test for different signs using the null hypothesis of equal signs. There was no evidence for different direction of temporal dynamics for any of the genera. This highlights the selective recruitment of soil bacteria by the wheat cultivar regardless of origin of the strains.

Table 2.

Comparison of the sign of the dynamics in SynCom and NatCom.

Slope signs	Genus	SynCom	NatCom	P-value	fdr
Equal	Bacillus	↓ (P = .0003)	↓ (P = .0001)	.0002	0.003
Equal	Peribacillus	↓ (P = .0001)	↓ (P = .0217)	.0109	0.071
Equal	Microbacterium	↑ (P = .0030)	↑ (P = .0851)	.0426	0.139
Equal	Variovorax	↑ (P = .0001)	↑ (P = .0855)	.0428	0.139
Equal	Ensifer	↑ (P = .0186)	↑ (P = .1332)	.0666	0.173
Equal	Agrobacterium	↑ (P = .0666)	↑ (P = .3930)	.1965	0.389
Equal	Pararhizobium	↑ (P = .0573)	↑ (P = .4184)	.2092	0.389
Equal	Pedobacter	↑ (P = .0030)	↑ (P = .4947)	.2474	0.402
Equal	Pseudomonas	↑ (P = .6892)	↑ (P = .2538)	.3446	0.498
Different	Arthrobacter	↓ (P = .3711)	↑ (P = .0166)	.1856	1.000
Different	Stenotrophomonas	↑ (P = .6653)	↓ (P = .0360)	.3327	1.000
Different	Flavobacterium	↑ (P = .0003)	↓ (P = .9324)	.4662	1.000
Different	Paenarthrobacter	↓ (P = .9728)	↑ (P = .7804)	.4864	1.000

The third and fourth columns specify if the relative abundances are increasing (↑) or decreasing (↓) in SynCom and NatCom, respectively, with P-value for the marginal two-sided T-tests stated in the parentheses. The fifth and sixth column specify P-values and associated false discovery rates for scientific hypothesis of equal signs (first 9 rows) and different signs (last 4 rows). Genera within the groups of equal and different signs are ordered according to increasing P-values.

Temporal dynamics show similar patterns in the field, but the trend is cultivar dependent

To determine whether the assembly of the SynCom under controlled conditions would reflect assembly of microbial communities in the field, and test our second hypothesis, we collected rhizoplane samples from the field of two winter wheat cultivars (cv. Sheriff, used for SynCom and NatCom experiments, and cv. Heerup) over a 4-week period in the fall of 2023. Seeds were sown following standard agricultural practices, without surface sterilization or pre-germination, in contrast to our SynCom and NatCom setup. Expectedly, the alpha diversity was higher in the field soil than in the SynCom or NatCom inocula (Fig. S8). This was also true for the alpha diversity if the rhizoplane communities of both cultivars grown in the field at all sampling time points (Fig. S8). The age of the plant explained 9% of the variation in the bacterial community composition (PERMANOVA, P < .001), while cultivar explained 6% of the variation (PERMANOVA, P = .002). The most dominant genera in the field were Bacillus, Paenibacillus and Massilia across the sampling period for both cultivars (Fig. S11). Except for Agrobacterium, Ensifer and Pedobacter, the rest of the SynCom genera were detected in the field soil prior to sowing (Fig. S10), with Ensifer and Pedobacter also showing very low abundance or were absent on the rhizoplane of both cultivars (Fig. 5). Despite this, the total relative abundance of SynCom genera comprised comparable proportions of the total communities regardless of experimental design (Fig. 5).

When comparing slopes using the COST procedure, we found that the majority of the SynCom members had slopes of equal signs when compared to field grown Sheriff (Table 3). There was evidence (marginal P < .05) for the same direction of the temporal dynamics for two genera; Flavobacterium and Variovorax that were both increasing. The increase in relative abundance with time of these two genera corroborate previous studies (Table 1). There was no evidence for different direction of temporal dynamics for any of the genera. This shows that the temporal dynamics of the SynCom strains mimic the dynamics of corresponding genera in the field of the same cultivar.

Table 3.

Comparison of the sign of the dynamics in SynCom and field-grown Sheriff.

Slope signs	Genus	SynCom	Field Sheriff	P-value	fdr
Equal	Flavobacterium	↑ (P = .0003)	↑ (P = .0019)	.0010	0.012
Equal	Variovorax	↑ (P = .0001)	↑ (P = .0184)	.0092	0.055
Equal	Bacillus	↓ (P = .0003)	↓ (P = .1730)	.0865	0.269
Equal	Pararhizobium	↑ (P = .0573)	↑ (P = .1794)	.0897	0.269
Equal	Ensifer	↑ (P = .0202)	↑ (P = .3370)	.1685	0.404
Equal	Peribacillus	↓ (P = .0001)	↓ (P = .4535)	.2268	0.441
Equal	Agrobacterium	↑ (P = .0666)	↑ (P = .5146)	.2573	0.441
Equal	Stenotrophomonas	↑ (P = .6653)	↑ (P = .6707)	.3354	0.459
Equal	Pseudomonas	↑ (P = .6892)	↑ (P = .2925)	.3446	0.459
Equal	Paenarthrobacter	↓ (P = .9728)	↓ (P = .2012)	.4864	0.584
Different	Microbacterium	↑ (P = .0030)	↓ (P = .7231)	.3616	1.000
Different	Arthrobacter	↓ (P = .3711)	↑ (P = .8315)	.4158	1.000

Evidence for equal (first 10 rows) and different (last 2 rows) signs was evaluated similarly as described in Table 2. Pedobactor was not included due to the very low abundance for field-grown Sheriff.

For Heerup, a winter wheat variety with a different microbiome [30, 33], there was evidence (marginal P < .05) for the same direction of the temporal dynamics for three genera; Pararhizobium, Flavobacterium, and Variovorax (Table S4). Notably these three genera also have the same direction for field grown Sheriff. For the remaining genera there were different directions between SynCom and field-grown Heerup, but none of these were statistically significant. To conclude, the SynCom can be used for studying interactions on the roots of wheat at least up to four weeks of grow. However, for the majority of the SynCom strains the assembly pattern was cultivar dependent.

Discussion

Here, we showed that a 13-member bacterial SynCom originating from the roots of winter wheat cv. Sheriff colonized the roots after inoculation in gamma-irradiated soil in a similar manner as observed for field grown Sheriff plants. These findings underpin the strong selective forces that drive rhizoplane community development, at least in the first month after sowing.

Previous studies focusing on plant-microbiome interactions have primarily used alternative substrates such as agar, artificial soil or EcoFab [18, 19, 21]. Using an experimental design with gamma-irradiated soil, provide a more realistic model system mimicking a natural soil system by maintaining soil structures to a large extend. At the same time, we acknowledge that gamma-irradiation alters soil nutrient levels, partly due to destruction of microorganisms, as well as impacts soil aggregate structure; however, the irradiation treatment has been shown to be less severe than autoclaving at resultant doses similar to ours (~35 kGy) [52]. While the soil was not sterile following gamma-irradiation, only one SynCom ASV was detected in the soil post-irradiation allowing for specific detection of the SynCom members on the rhizoplane. Hence, we present our model system as a valid platform for studying mechanisms in plant-microbe interaction in future research.

We found no differences in shoot biomass and length between the SynCom and NatCom inoculated plants suggesting that the SynCom provides the necessary functions for plant growth and development in the early stages. This complements the findings from maize, radish and tomato that have shown that 7–15-strain SynComs are sufficient for early plant development [17, 22, 53]. All SynCom strains colonized the rhizoplane over the four weeks of growth. Only one SynCom ASV (Peribacillus simplex ASV13) was found in the gamma irradiated soil prior to inoculation, and hence, we conclude that the SynCom ASVs found on the roots originated from the SynCom inoculum. None of the strains completely dominated the rhizoplane indicative of a niche preference for each SynCom member. Previous studies using single strains in gnotobiotic systems with wheat, barley and maize have shown that there is a maximum colonization potential, even without competition from other than seed endophytes [54]. In line with this, seed inoculation with Pseudomonas fluorescens SBW25 did not change the relative abundance of Pseudomonas on wheat roots compared to untreated seeds [32]. Taken together, this implies that the number of certain taxa that can colonize the root is constrained by specific root exudates, limitation of space on the roots and/or interactions among root colonizing microorganisms. Acknowledging this will be important in bioinoculant design, and within-genus competition on the roots should be tested due to similar niche preferences.

Despite surface sterilization and the use of gamma irradiated soil, we were not able to remove all other sources of bacteria. This was especially apparent for Bacillus and Pedobacter, as the majority of the ASVs from these genera were non-SynCom ASVs. Bacillus is a known wheat seed endophyte [55], but was also present in the gamma-irradiated soil. Similarly, Pedobacter, although not found in the soil, has been identified as a seed endophyte in barley [56]. This suggests that seed endophytes of Bacillus and Pedobacter had the ability to preempt niche space from the SynCom strains [57]. Contrastingly, Pseudomonas and Stenotrophomonas strains were not outcompeted by seed endophytes from the same taxa [58]. Seed endophytic and copiotrophic Massilia [59] are thus hypothesized to use early seed and root exudates resulting in niche pre-emption, and affect the colonization of fast-growing Pseudomonas, but this warrants further studies. These results substantiate that plant root community assembly is subject to priority effects, as previously demonstrated using SynCom inoculations on seeds [60]. Given the low abundance of bacteria in the gamma-irradiated soil, and that SynCom members were actively growing, we conclude that seed endophytes are more important for the development of the rhizoplane microbiome than taxa present in the gamma-irradiated soil in the current experiment.

In support of our first hypothesis, we found that several SynCom strains had temporal dynamics similar to the corresponding genera in the NatCom. This underscores the strong selective power by a cultivar across microbiome composition. Using genus-level aggregation for the NatCom data in the comparison to the SynCom community dynamics, neglects within-genus variation. While this might have implications for plant responses, a strong colonization stability at the genus level enables studies on more intimate intra-species variation and its impact on plant physiology possible.

We observed the largest variation in community composition (beta diversity) among replicates after 7 days for the SynCom communities. The high variation in beta diversity during early stages of plant growth was also shown for maize [11, 61]. This might be attributed to dispersal limitation [62]. Contrastingly, the lower beta diversity among replicates at later developmental stages for the SynCom and NatCom could be caused by homogenous selection across the plants. This supports a previously proposed model with deterministic processes become more dominant over time [10].

We compared single strains with the temporal dynamics of genera from the rhizoplane of field grown plants. Overall, the temporal dynamics of 10 strains in the SynCom, except Pedobacter, Arthrobacter and Microbacterium resembled those of the field grown Sheriff cultivar. Further, the sign of the slopes of Bacillus, Variovorax, Flavobacterium and Pararhizobium were significant and displayed a temporal development in accordance with previous findings (Table 1) [32, 33]. This is despite the lower diversity in the SynCom inoculum and lack of archaea and eukaryotes compared to the field soil and the differences in seed treatment and lower temperatures for the field grown plants.

In contrast to the high similarity in dynamics of the 13 SynCom genera for Sheriff across systems, fewer similarities were found for the Heerup cultivar. This is in line with our second hypothesis and supported by wheat-cultivar specific recruitment of members within a genus, as shown for Pseudomonas [30]. This finding of different temporal dynamics of bacterial genera between two cultivars emphasizes the difficulties in generalization based on plant-microbiome studies even within a plant species. In an applied perspective, it demonstrates the importance of cultivar-specific inoculant design, and could explain observed difficulties in translating results from lab to field.

Since the conducted growth chamber experiment only lasted for 28 days, further experiments are warranted for testing long-term related hypotheses. In fact, this is a general challenge in the research field [63]. Another aspect is the observations of zero abundance of some genera that can be due to either unsuccessful colonization or too low abundance to be detected. Since bacterial abundance is best modelled on a logarithmic scale these zeros pose a challenge. An often-used solution is to add a small positive number to all observations. However, this solution is ad hoc, and results may change with the number used. Here, we chose to interpret a zero as the observation that this abundance is lower than all the non-zero abundances within that genus. This solution still entails a modelling choice and hence impacts the precise P-values in the statistical computations. However, we believe that this is less ad hoc than adding an arbitrary small number. An alternative is to only investigate genera without zero observations, but this would exclude several genera (two for SynCom, six for NatCom, five for field grown Sheriff, and four for field grown Heerup).

Conclusion

In conclusion, we demonstrated the potential of applying a SynCom to study rhizoplane microbiome assembly on wheat in soil. Its members actively colonized the wheat roots and despite lacking eukaryotes and archaea, it developed in accordance with field communities of the same wheat cultivar. Overall, this supports the notion that there is a certain carrying capacity for each community on the roots due to niche preferences. Recognizing this will be helpful in advancing efforts on the application of beneficial microbes to seeds for root colonization. Our approach with inoculating the SynCom into gamma-irradiated soil provides a promising framework as a more realistic scenario, compared to seed inoculation and the use of hydroponics to identify determinant factors of microbial community assembly at the sub-genus level and to leverage studies of plant-microbe and microbe-microbe interactions at the root-soil interface.

Conflicts of interest

None declared.

Funding

This work was conducted within the INTERACT project funded by Novo Nordisk Foundation (grant no. NNF19SA0059360). FB was supported by Independent Research Fund Denmark (grant no. 10.46540/2031-00010B).

Data availability

Raw 16S rRNA gene amplicon sequences have been deposited in the NCBI Sequence Read Archive (SRA) database under Bioproject accession number PRJNA1189047. The scripts describing data treatment and the COST procedure are available at: https://github.com/BakDK/Wheat-SynCom.