Analysis of the alpha-diversity of the active microbiota revealed contrasting trends in bacterial and fungal communities between the root and BS compartments, consistent with previous studies on plant diversity effects on soil microbiota (Tkacz et al., 2020; Li & Wu et al., 2018; Schlemper et al., 2018).
Using the Bray-Curtis taxonomic dissimilarity index, we showed a clear separation between samples from BS, RAS, and root compartments, in both bacterial and fungal communities, as reported in other plant studies (Yeoh et al., 2017). Notably, PM line samples showed clustering and differentiation based on their aggregation capacity, particularly in the root compartment, with distinct separation of HAL lines from IAL and LAL lines (Fig. 1A). The rhizosphere of HAL PM lines exhibited specific enrichment of Verrucomicrobia OTUs, indicating their potential contribution to soil aggregation through increased root exudate availability (Ndour et al., 2022) These findings align with previous reports of Verrucomicrobia's ubiquity and low relative abundance across soil nutritional gradients (Bergmann et al., 2011).
The ß-NTI values elucidate that the assembly process of bacteria in the RT and RAS compartments of the four PM lines was markedly influenced by deterministic processes, predominantly characterized by a prevalence of homogeneous selection. This phenomenon, potentially driven by the presence of plant-root exudates, underscores the pivotal role of these compounds in shaping bacterial community dynamics (Figure 2C). An interesting parallel can be drawn from the research by Fan et al., revealing that the significance of deterministic processes in shaping diazotrophic communities diminishes as one moves away from wheat roots (Fan et al., 2018). In a contrasting manner, the assembly of the fungal community within the pearl millet rhizosphere is primarily governed by stochastic processes (Figure 2C). This outcome harmonizes with recent findings concerning the rhizosphere of Typha orientalist (Wang et al., 2023), as well as earlier studies (Powell et al., 2015; Lekberg et al., 2011).
The analysis of co-occurrence networks provides valuable insights into ecosystem functioning, plant nutrition, and resilience to biotic and abiotic stresses (Berendsen et al., 2012; Vandenkoornhuyse et al., 2015; Benidire et al., 2020). Our study revealed relatively minor differences in the organization and complexity of microbial networks among PM lines, suggesting that interactions are highly intricate in the rhizosphere of each line. The majority of specific connector hubs in each PM line network exhibited a combination of generalist and specialist characteristics, highlighting the role of the host plant in shaping the structure and function of its microbial community through root exudation (Zhou et al., 2010; Saad et al., 2010) (Fig. 2B). For instance, the EPS-producing Mesorhizobium (bac_283) functioned as a generalist in the HAL network (Fig. S3 and 2B), potentially providing benefits to the plant, but shifted to a specialist role with different ecological implications in the LAL network (Fig. 2B).
Interactions among different species shape the assembly and functions of microbial communities, with potential beneficial, neutral, or detrimental effects on community members.
Metabonomic profiles of PM line compartments
The influence of plant compartments on the microbiome, driven by their distinct physical and chemical properties, has been well-documented (Chen et al., 2020; Ehlers et al., 2020). Previous studies have primarily focused on hydroponic or sterile plant systems, analyzing specific compounds related to root exudation, soil aggregation, and associated metabolites (Khorassani et al., 2011; Van Dam et al., 2016; Luo et al., 2017; Huang et al., 2019; Korenblum et al., 2020). While these approaches successfully quantified specific chemicals of interest (e.g., phenolics, antioxidants) influencing plant growth and health (Mokgotho et al., 2013; Altemimi et al., 2017; Rajniak et al., 2018; Pudziuvelyte et al., 2020), they often overlooked the complexity of plant-microbe interactions and the biochemical diversity present in real-world conditions, where chemical compounds segregate among different plant parts (Kuijken et al., 2015).
In our study, we established an in situ system using native soil and employed a sensitive and untargeted analytical approach (FT-ICR-MS) to profile the diverse metabolites present in the various compartments of four PM lines with contrasting soil aggregation capacities. This allowed us to capture both root exudates and microbial metabolites activated by root exudation in the RAS compartment, as well as differentiate them from soil organic matter-related metabolites observed in the BS compartment. By considering the intricate interplay between plants and microbes under natural conditions, our approach provides a comprehensive understanding of the metabolic profiles associated with different plant compartments and their implications for plant-microbe interactions.
After assigning the molecular compositions, we observed a slightly lower number of assigned formulae (1851) in the root-adhering soil (RAS) compared to the bulk soil (BS) control (1875 molecules) (Fig. S4A-B). This finding is consistent with studies conducted on Arabidopsis thaliana (Witzel et al., 2017), and suggests that the microbial rhizosphere effect, influenced by root exudation, leads to a reduction in biochemical diversity in the rhizosphere (Bais et al., 2006; Jones et al., 2009; Bakker et al., 2013; Whalley et al., 2005). The compounds detected in the BS and RAS compartments mainly comprised low-mass (150-400 m/z) CHO- and CHNO-molecules, identified as carbohydrates, aliphatics, and amino acids (Fig. S4B-C) (Whalley et al., 2005). In contrast, the compounds detected in the root tissues (RT) exhibited a lower number but a higher mass range (up to 700 m/z ratio) compared to the RAS and BS, with an increase in the CHOS molecular composition observed in the phenolic and carboxyl-rich aliphatic molecule (CRAM) zones, along with an increase in aliphatics (Fig. S4A-C). The shoot compartment of the PM lines showed the highest number of assigned molecules (2450) (Fig. S5B), which aligns with findings from studies on other plants such as rice and banana (Pudziuvelyte et al., 2020; Tawaraya et al., 2018). The compounds detected in shoots predominantly belonged to the CHNO- and CHO-molecules, with fewer CHOS compounds, resulting in an abundance of condensed and highly unsaturated compounds in the CRAM and phenolic classes (Fig. S4A-C). These differences in molecular composition between above- and below-ground plant parts can be attributed to factors such as biomass distribution, compartment-specific functions, cellular structures, and the production, translocation, and storage of primary and secondary metabolites (Pyankov et al., 2001; Oliveira et al., 2007; Shabala et al., 2016).
Metabolic profiles of PM lines in relation to their aggregation capacity
Molecular variance between the compartments of the four contrasting PM lines was observed in the number, intensity, and molecular composition of detected compounds. Specifically, in the root-adhering soil (RAS) compartment, the HAL lines (L132 and L253) exhibited the highest number of assigned carbohydrate (CHO) compounds compared to IAL-3 and LAL-220 (Fig. 3A). This indicates an increased presence of metabolites originating from root exudation and the RAS-associated microbiota, potentially contributing to enhanced soil aggregation in these HAL lines. Notably, there was a distinct differentiation between the two HAL lines, with HAL-132 displaying a higher abundance of aliphatic CHO compounds compared to HAL-253 (Fig. 3D). Previous studies on root exudation mechanisms have emphasized the inter- and intra-specific variability in root exudate composition as shown in Helianthus and Quercus ilex (Bowsher et al., 2016; Cargallo-Garriga et al., 2018).
For both the root and shoot compartments, the HAL lines exhibited a distinct molecular composition compared to the other two lines (Fig. 3D), characterized by an enrichment of CHNO molecules (Fig. 3B). Root exudation mechanisms often involve a coupling of passive and active processes (Jones et al., 2019; Shabala et al., 2016; Limmer & Burken, 2014; Sasse et al., 2018).
The relationship between root exudate composition and soil aggregation in the rhizosphere was further supported by statistical analysis (PCA), which revealed clustering of samples and compartments of the four PM lines based on their aggregation capacity (Fig. 3D). Additionally, a heat map analysis of the 1000 most significant compounds (using Pearson distance), based on similarity values, demonstrated a clear correlation between the spectra of the 64 samples and the soil aggregation ratios of the PM lines in each compartment, particularly in the root and shoot compartments (Fig. 3C).
Determination of unique metabolites in each compartment of the PM lines
The BS compartment has the greatest number of unique low-mass CHNO- and CHO- molecules (200-500 m/z) compared to the RAS of the PM lines (Fig. S5C, Fig. S5ARAS-BRAS). These compounds correspond to the composition of low-mass range molecules found in the Bambey arenosol, which is characterized by low organic matter content (0.4% wt). To identify common root exudates and metabolites associated with microbial activities in the rhizosphere of each PM line, the composition of BS was subtracted from that of RAS. A total of 99 molecules ranging from 199 to 495 m/z mainly composed of carbohydrates/carbohydrates conjugates, amino acids/peptides and analogues, hydroxycinnamic acids and derivatives, fatty acids and conjugates, fatty acid esters as well were found in all PM lines but absent in the BS.
Among a list of 37 molecules specific to HAL-132 (Table S3), several were of interest, including coniferyl alcohol, known as a nod gene inducer in Bradyrhizobium japonicum (Kape et al., 1991); rhodojaponin, described as an insecticide produced by Rhododendron molle (Cheng et al., 2011); hydroxyoxytetracycline, an antibiotic analog (Valcavi et al., 1981); cerulenin, an antifungal antibiotic isolated from Cephalosporium caerulens (Tomoda et al., 1984); and two homoserine lactones (N-(3-oxododecanoyl) homoserine lactone and N-(3-hydroxy-heptanoyl)-homoserine lactone), known to be involved in bacterial quorum sensing (Miller and Bassler 2001).
Fewer specific molecules were identified in the other three PM lines: 10 for HAL-253, 6 for IAL-3 and 12 for LAL-220. LAL-220 harbored sieboldin, a dihydrochalcone compound typically found in Malus species (Rivière, 2016), as well as lucidone A, a plant diterpene secondary metaboliteFungal compounds predominantly produced by Aspergillus were retrieved from the RAS of the HAL line. Specifically, shoyuflavone A and a dihydropyranone called aspyrone were identified in the RAS of L253. Additionally, the RAS of L132 contained a sesquiterpenoid compound called asperugin and a carbonyl compound known as phomaligin. Notably, the RAS compartment of IAL-L3 and LAL-L220 contained two benzoxazinoids, DIMBOA-Glc (IAL-3) and HDMBOA (LAL-220) which have been found in maize (Cambier et al., 2000) and more widely in Poaceae (Oikawa et al., 2002), knowing that pearl millet belongs to Poaceae. These secondary metabolites are known to trigger rhizosphere colonization by the plant-growth promoting bacterium Pseudomonas putida (Neal et al., 2012) and inhibit host recognition and virulence of the phytopathogen Agrobacterium tumefaciens (Maresh et al., 2006). More recently, using a metabonomic approach of root exudation in a non-sterile soil, it was shown that large amount of benzoxazinoids (including DIMBOA) and flavonoids were detected in the maize rhizosphere (Pétriacq Maresh et al., 2017), and that benzoxazinoids (especially MBOA) shaped the bacterial and fungal diversity in the maize rhizosphere (Mönchgesang et al., 2016). DIMBOA has also been shown to have multiple effects on rhizosphere microbiota, especially Proteobacteria and Chloroflexi, such as plant-soil feedback, metabolic regulation, and gatekeeper effects that will all lead to a change in the microbial community structure and functions (Kudjordjie et al., 2019; Cotton et al., 2019), and more recently, Wang et al., (2022) showed that the exudations of GABA and DIMBOA are involved in shaping the rhizosphere and endosphere microbiomes. The absence of DIMBOA-Glc or HDMBOA-Glc in the RAS compartment of HAL lines may suggest that these molecules could negatively control the activity of soil-structuring bacteria in the rhizosphere of IAL-3 and LAL-220.
Co-inertia analysis of root and RAS omics datasets
Co-inertia analysis of the microbial populations and the specific metabolic compounds datasets from the root and RAS compartments revealed a correlation between the microbial populations and specific metabolic compounds in the PM lines, as depicted in the 3D plots (Fig. 4). The separation of the correlated omics data sets of each PM line and their alignment with the soil aggregation ratios further supported the link between plant-microbiota interactions and soil aggregation. Similar approaches combining metabonomics and metagenomics have been employed in various plant species, including A. thaliana (Huang et al., 2019; Witzel et al., 2019; Mönchgesang et al., 2016), Avena barbata (Zhalnina et al., 2018), British bluebells (Raheem et al., 2019), rice (Li et al., 2020), tomato (Korenblum et al., 2020), potato (Gotthardt et al., 2016), poplar (Kaling et al., 2018) highlighting the role of root exudation in shaping the root/rhizosphere-associated microbiota and their collective impact on soil aggregation. All of these studies mentioned above, as well as many others have evidenced the effect of the plant through root exudation on the root/rhizosphere-associated microbiota and their combined role in soil aggregation (Alami et al., 2000; Trivedi et al., 2017; Baumert et al., 2018; Costa et al., 2018; Saleem et al., 2018, Erktan et al., 2020, Lehmann et al., 2020; Haichar et al., 2014), which is under complex genetic control in pearl millet (de la Fuente Cantó et al., 2022). In the future, these omics approaches will continue to evolve and improve, particularly in terms of statistical and bioinformatics analysis, (Meng et al., 2014, Deng et al., 2012; Lucaciu et al., 2019), and combined with more complementary omics tools such as metaproteogenomics, metatranscriptomics and metaproteomics to strengthen the analysis of the plant-soil-microbiota continuum and shed light on this black box (Kaling et al., 2018; Deyholos 2010; Knief et al., 2012; Aguiar-Pulido et al., 2016; Liu et al., 2017). This integration of diverse omics data will provide a more comprehensive understanding of the plant-soil-microbiota continuum, helping to unravel the complexities of this intricate relationship.
The root exudate composition plays a crucial role in shaping the assembly and interaction networks of the rhizosphere microbiota, thereby influencing the structuring of the soil surrounding the roots (Figure 5). A significant portion, up to 20%, of the photosynthetates produced by the plant is allocated as root exudates to recruit the rhizosphere microbiota. Some microorganisms transform these exudates into exopolysaccharides (EPS), which contribute to soil particle aggregation by increasing soil adherence to the roots. This process improves water and mineral availability for the plant and enhances carbon storage in the soil as the fresh carbon is not completely mineralized (Alami et al., 2000; Ndour et al., 2022). Interestingly, the presence of DIMBOA and H-DIMBOA, known for their antimicrobial activity, is exclusively detected in the rhizosphere of PM lines with lower aggregation capacity. We hypothesize that these compounds may inhibit EPS synthesis by bacteria or selectively suppress certain EPS-producing bacterial populations. Further investigations are warranted to elucidate the mechanisms underlying the interplay between root exudate composition, EPS synthesis, and microbial communities, shedding light on their combined influence on soil aggregation and carbon sequestration in soils (Fig. 5).