Metabolome-driven microbiome assembly determining the health of ginger crop (Zingiber officinale L. Roscoe) against rhizome rot

Abstract

Background

Plant-associated microorganisms can be found in various plant niches and collectively comprise the plant microbiome. The plant microbiome assemblages have been extensively studied, primarily in model species. However, a deep understanding of the microbiome assembly associated with plant health is still needed. Ginger rhizome rot has been variously attributed to multiple individual causal agents. Due to its global relevance, we used ginger and rhizome rot as a model to elucidate the metabolome-driven microbiome assembly associated with plant health.

Results

Our study thoroughly examined the biodiversity of soilborne and endophytic microbiota in healthy and diseased ginger plants, highlighting the impact of bacterial and fungal microbes on plant health and the specific metabolites contributing to a healthy microbial community. Metabarcoding allowed for an in-depth analysis of the associated microbial community. Dominant genera represented each microbial taxon at the niche level. According to linear discriminant analysis effect size, bacterial species belonging to Sphingomonas, Quadrisphaera, Methylobacterium-Methylorubrum, Bacillus, as well as the fungal genera Pseudaleuria, Lophotrichus, Pseudogymnoascus, Gymnoascus, Mortierella, and Eleutherascus were associated with plant health. Bacterial dysbiosis related to rhizome rot was due to the relative enrichment of Pectobacterium, Alcaligenes, Klebsiella, and Enterobacter. Similarly, an imbalance in the fungal community was caused by the enrichment of Gibellulopsis, Pyxidiophorales, and Plectosphaerella. Untargeted metabolomics analysis revealed several metabolites that drive microbiome assembly closely related to plant health in diverse microbial niches. At the same time, 6-({[3,4-dihydroxy-4-(hydroxymethyl)oxolan-2-yl]oxy}methyl)oxane-2,3,4,5-tetrol was present at the level of the entire healthy ginger plant. Lipids and lipid-like molecules were the most significant proportion of highly abundant metabolites associated with ginger plant health versus rhizome rot disease.

Conclusions

Our research significantly improves our understanding of metabolome-driven microbiome structure to address crop protection impacts. The microbiome assembly rather than a particular microbe’s occurrence drove ginger plant health. Most microbial species and metabolites have yet to be previously identified in ginger plants. The indigenous microbial communities and metabolites described can support future strategies to induce plant disease resistance. They provide a foundation for further exploring pathogens, biocontrol agents, and plant growth promoters associated with economically important crops.

Video Abstract

Background

Plant-associated microorganisms can be found in various plant niches and collectively comprise the plant microbiome [1]. Plant microbiomes contain beneficial and pathogenic microbes [2]. Advances in high-throughput sequencing techniques have deepened our knowledge of the relationship between microbiomes and hosts [2].

Plant microbiome assemblages separated into above- and belowground constituent parts have been extensively studied, primarily in model species, including the soil microbiome [3,4,5], rhizosphere [6,7,8,9,10], root [11,12,13], and phyllosphere [14, 15]. Microbial communities associated with several plant niches have also been analyzed [16]. Fungus-induced changes are correlated with changes in the wheat leaf microbiome [17]. However, understanding the variation in the microbiome is imperative for determining how microbiome assembly affects overall plant holobiome health.

Conversely, plant secondary metabolites (PSMs) perform many functions, including defense against pathogens [18]. PSMs capable of broadly changing plant microbiomes have been described [2]. Phytohormones such as jasmonic acid (JA), salicylic acid (SA), ethylene (ET), and abscisic acid (ABA), among the most studied pathogenesis mediators, have also been shown to have an impact on the microbiome of plants [7, 11, 19].

The plant microbiome and metabolome are closely correlated, which indicates that endophytes can promote the accumulation of secondary metabolites that are relevant to active medicinal properties [20, 21]. The rhizosphere microbiome was shown to drive the systemically induced root exudation of metabolites [22]. Less attention has been given to the effects of the metabolome–microbiome relationship on plant health, although the interactive effect of host plant defense and root-associated microbiota is evident after Fusarium oxysporum infection in Arabidopsis thaliana [12].

Although little research has been conducted on ginger (Zingiber officinale L. Roscoe) compared to other agricultural plants [23], ginger is a perennial monocotyledonous herb with underground rhizomes and a long history of use as a fresh vegetable, spice, and herbal medicine. However, this crop is vulnerable to various plant pathogens [24], and rhizome rot has been a significant limiting factor for ginger’s yield and marketing potential in China.

Rhizome rot is a highly destructive disease that has been found to reduce ginger production by 50–90% [25]. The disease causes significant losses, especially in warm and humid conditions, with severe outbreaks observed in recent years. In 2020, rhizome rot led to an average yield loss of 20 to 25% in the Tangshan region, posing a significant threat to local ginger farming [26]. This disease has increasingly become one of the most devastating issues for ginger cultivation in Shandong Province, a key ginger production area in China [27].

Further research on the disease’s epidemiology and potential management options is necessary. Ginger rhizome rot can be attributed to multiple causal agents, including Fusarium oxysporum f. sp. Zingiberi [25], Pectobacterium brasiliense [26], Bacillus pumilus [27], Pythium myriotylum [28], and Enterobacter cloacae [29]. This complex pathosystem is worth studying to determine the microbiome and the metabolome assembly that keeps plants healthy. Here, we performed metataxonomic analyses using bacterial and fungal amplicon sequencing and untargeted metabolomics analysis to identify the metabolome-driven structure and function of microbial communities associated with rhizome rot and ginger plant health.

Methods

Sample collection and preparation

Samples were collected in the Laiwu district of Jinan, Shandong Province (1.36°19′50" N, 117°29′29" E; northern China), which has optimal growing conditions, but rhizome rot is a factor limiting the yield and marketability of ginger [28]. The sampling area is in a typical warm-temperate humid/semihumid climate zone, with an annual mean temperature of 12.5 °C, annual mean precipitation of 688.9 mm, and 62% relative humidity. The frost-free period is 191 days, and the annual sunshine hours are 2629 h [30]. Almost 70% of the total precipitation occurs from July to September. The soil in the area is classified as sandy loam [31].

The ginger variety used, Zingiber officinale var. officinale, was the same as that planted by local farmers. The size of each plot was approximately 666 m², and 7000 to 8000 plants were grown in each plot. The plots were subjected to the same irrigation and fertilization regimes. These plants were watered ten times during the crop growth cycle. Approximately 100 kg of compound and organic fertilizer (chicken manure) were applied to the soil at various times during the crop cycle, including during soil preparation, sowing, and crop growth. Sample collection was performed on September 12, 2021. In September, the mean temperature is 25°C during the day and 18°C at night. The relative humidity of the soil is 75–85%, and the area is exposed to 9 h of sunshine on average. Only ginger was grown within a radius of at least 1500 m in the sampled area. The area where samples are collected is also utilized for planting garlic. The crop rotation cycle occurs every 2 years, and this area has been dedicated to ginger farming for approximately 40 years. The diseased plants were stunted with yellowish, dry lower leaves that turned brown. Additionally, their rhizomes were rotted or spongy, which aligns with the symptoms of rhizome rot previously described [25]. Endophytic bacteria were identified from asymptomatic plant tissues, but there was a notable increase in Pectobacterium_carotovorum_subsp._brasiliense (Supplementary Table 1) in diseased plants compared to healthy plants. Three replicates of healthy and diseased plants were collected from three adjacent plots. Each replicate consisted of a composite sample obtained by mixing three samples collected from the same niche (leaf, stem, root, rhizome, rhizosphere soil, and bulk soil; Fig. 1D) from three symptomatic or asymptomatic plants per plot for a total of 36 composite samples. The rhizosphere is the microbial habitat around the root [32], although we also applied this term to the soil adjacent to the rhizome. Approximately 30 g of bulk soil sample was collected at a distance of 20 cm from the root and at a depth of 0 to 15 cm, and the rhizosphere soil attached to the roots and rhizomes was collected by manual shaking. The samples were subsequently transferred to collection bags and transported to the laboratory on dry ice. Plant samples were washed immediately upon arrival at the laboratory with tap water until they appeared to be free of debris and then rinsed three times with distilled water (dH₂O). To sterilize the surface of the plant organs and remove exogenous bacteria and fungi, the samples that were used for endophytic diversity analyses were immersed in 70% ethanol for 5 min, 2.5% sodium hypochlorite solution for 1–2 min, and 70% ethanol for 1 min and then rinsed vigorously three times with sterilized Millipore water. To verify the efficacy of the sterilization process, a sample from the last portion of the water used for washing was inoculated on potato dextrose agar (PDA) plates, which were incubated at 28 °C for 10 days, and on LB plates, which were incubated at 37 °C for 5 days before checking for the appearance of colonies [33]. The surface-sterilized plant organs constituted the endophyte samples. Samples for molecular analysis were stored in a − 80°C freezer until DNA extraction.

Fig. 1

The plant disease rhizome rot drives changes in the assembly of the microbiota associated with the entire ginger plant. Occurrence of specialist bacterial (A) and fungal (B) genera in healthy and diseased ginger plant niches. C Number of core/specialists bacterial and fungal microbes. D Sampling diagram of various ginger microbial niches. Each pie chart shows the number of specialist microbes inhabiting a specific microbial niche, and the most abundant microbes (> 5%) per niche are indicated by letters. The microbial niches are numbered beside each one, and the specialist microbes per niche are listed in Supplementary Table 3. The numbers shown below each abbreviation equal the total number of microbes and the number of specialist microbes for each microbial niche. An empty circle is shown for HPRh, indicating the absence of specialist microbes inhabiting that microbial niche. HPBS healthy plant bulk soil, DPBS diseased plant bulk soil, HPRhS healthy plant rhizosphere soil, DPRhS diseased plant rhizosphere soil, HPRh healthy plant rhizome, DPRh diseased plant rhizome, HPR healthy plant root, DPR diseased plant root, HPS healthy plant stem, DPS diseased plant stem, HPL healthy plant leaf, DPL diseased plant leaf

DNA extraction and PCR amplification

All laboratory protocols were performed at Shanghai Majorbio Bio-pharm Technology Co., Ltd. The samples were processed under normal experimental conditions. Illumina metagenomic library preparation guidelines were followed to create 16S and ITS rRNA gene amplicon libraries. DNA extraction from 0.5 g of rhizosphere and bulk soil samples or 5 g of plant tissues was performed using a DNeasy PowerSoil Kit (Qiagen, MD, USA) according to the manufacturer’s instructions. After the genomic DNA extraction was completed, 1% agarose gel electrophoresis was carried out to detect the extracted genomic DNA. DNA was quantified using a NanoDrop spectrophotometer. Each sample was tested three times and kept at − 20℃ until PCR amplification was performed. The V5–V7 hypervariable region of the bacterial 16S rRNA gene was amplified using the universal primers 799F (5′-AACMGGATTAGATACCCKG-3′) and 1193R (5′-ACGTCATCCCCACCTTCC-3′), which provided a more accurate picture of the bacterial community structure and very low amplification of nontarget DNA [10], while the fungal ITS2 region was amplified using the primers ITS3F (5′-GCATCGATGAAGAACGCAGC-3′) and ITS4R (5′-TCCTCCGCTTATTGATATGC-3′) [34], which proved to be the most appropriate for the characterization of fungal communities with metabarcoding [35]. An AxyPrep DNA Gel Recovery Kit (AXYGEN) was used to excise the products from the gel and recover them according to the manufacturer’s instructions. PCR products were assessed and quantified with the QuantiFluorTM-ST Blue Fluorescence Quantitative System (Promega Corporation, Madison, WI, USA). Replicates of the same sample were pooled in equimolar proportions for sequencing.

Amplicon sequencing and bioinformatic analysis

The bacterial and fungal amplicon sequences of the 36 analyzed samples were independently sequenced. Negative controls were used (sterile water was used instead of template DNA) to exclude contamination by PCR amplification. Amplicon libraries were sequenced on the Illumina MiSeq PE300 platform (Illumina, USA) according to the manufacturer’s protocols, and 250 bp paired-end reads were generated. The 16S rRNA and ITS gene sequences generated were analyzed using the online Majorbio Cloud Platform [36] based on the QIIME pipeline [37] version 1.9.1 using recommended parameters. Paired-end reads obtained from the Illumina platform were assembled, and the primer sequences and low-quality reads with scores less than Q30 were trimmed using USEARCH v.11.0 software [38] with default parameters. The sequencing run produced 2,645,244 high-quality reads across the 36 input libraries. Operational taxonomic units (OTUs) were assigned based on 97% similarity among clustered reads and then checked for chimeras using the UPARSE (v.7.0.1090, https://drive5.com/uparse/) pipeline [39] in USEARCH v.11.0 software [38] with default parameters before generating an OTU count table. OTUs were taxonomically annotated using the SILVA reference database (v.138, https://www.arb-silva.de) and I database (v.8.0, http://unite.ut.ee/index.php) for bacteria and fungi, respectively. The Shannon rarefaction curve was calculated (Supplementary Fig. 1A and 1B) by randomly resampling each sample several times, plotting the rarefied number of OTUs defined at a 97% sequence similarity threshold relative to the number of samples (Mothur v.1.30.2, https://www.mothur.org/wiki/Download_mothur), and the minimum number required for subsequent analysis was validated. We performed a single rarefaction at a depth of the shallowest sample to control for variable sequencing effort between representatives. Then, we chose a subsampling depth of 27,618 sequences per bacterial sample and 45,861 per fungal sample, which yielded a final rarefied dataset for all 36 models. Bacterial and fungal sequences were assigned to each sample based on their barcodes using the SILVA v138 16S (http://www.arb-silva.de) and UNITE v8.0 ITS (http://unite.ut.ee/index.php) databases, respectively.

Microbial diversity analysis

Although different indices showed very similar results (source data Fig. 2), plant health was related to both diversity and microbial composition [40]. Thus, two alpha diversity indices were considered at the genus level: observed richness (Sobs), which provides a direct measure of population complexity by counting the number of different species in a sample (observed OTUs), and the Shannon H’ index, which is an estimator of taxon diversity, combining richness, and uniformity [41] with the Kruskal–Wallis test for all pairwise combinations. Principal coordinate analysis (PCoA) was conducted with the vegan package v.2.4.3 in R software v.3.3. [42] based on the Bray–Curtis distance algorithm to visualize the β diversity pattern of microbial communities between samples from different microbial niches of healthy and diseased plants. Permutational multivariate analysis of variance (PERMANOVA) was performed using 999 permutations computed from the rarefied dataset (n = 36) to test the relative contribution of both disease and plant compartment microhabitats to community dissimilarity. The core or generalist taxa in the ginger microbiomes were defined as OTUs present in 100% of the plant samples, while the specialists were present in only one plant niche.

Fig. 2

Diversity analysis at the genus level for microbial communities from the niches of healthy and diseased ginger plants. Box plots of alpha diversity showing the Sobs index (A) and Shannon index (C) for archaeal/bacterial communities at the genus level and the Sobs index (B) and Shannon index (D) for fungal communities at the genus level for both healthy and diseased ginger plants. The bars represent the average of three composite biological replicates for each microbial niche, and the error bars indicate the standard variation in the mean. Significance was tested using a Kruskal–Wallis rank sum test; *P < 0.05, **P < 0.01, ***P < 0.001. Principal coordinate analysis (PCoA) plot visualizing variation in the bacterial (E) and fungal (F) community compositions of healthy and diseased plant samples in two-dimensional space based on Bray‒Curtis dissimilarity. Permutational multivariate analysis of variance (PERMANOVA) by Adonis was performed to test the significance of microbial community dissimilarity. Each dot represents a single composite sample. Lines are drawn to connect three replicates of each composite sample to each other. The different colors in the figure represent different microbial niches

Predictive and statistical analysis

The data are displayed as the average of at least three independent replications and the standard deviation. P values less than 0.05 were considered to indicate statistical significance. We summarized the distribution of the annotated OTUs based on the species results to reveal the general species distribution patterns of the different samples. In particular, pie diagrams were generated to indicate the numbers of shared (core) or unique (specialist) microbial genera among compartments for healthy and diseased ginger plants. Clustering heatmaps reflecting differences in the abundance of different samples through color changes were generated (ggplot2’ package v3.2.1 in R Studio v3.5.3). Microbial functional assemblages from 16S rRNA gene sequences were predicted by FAPROTAX [43] and were compared using the Kruskal–Wallis rank sum test, while fungal OTUs were classified into ecological guilds using the online application FUNGuild [44]. A confidence ranking of “highly probable” or “probable” was retained for high accuracy, whereas those with “possible” confidence rankings were considered unclassified. Undefined guilds: undefined pathogens, defined as nonspecific pathogens of fungi, plants, or animals; undefined saprotrophs, defined as nonspecific saprotrophs of wood, plants, or litter soil. Linear discriminant analysis (LDA) effect size (LEfSe) was applied to determine the features (differentially enriched microbial taxa and functions) most likely to explain differences between healthy and diseased ginger plants. The samples were pooled to analyze the soil and endophyte microbiomes of plants that appeared healthy or diseased. Taxa with an LDA effect size greater than 4.0 (P < 0.05) were considered significant.

Metabolomics analysis

We analyzed changes in the endophyte microbiome of plants driven by the metabolome and implications for plant health. The same 24 samples of leaves, stems, roots, and rhizomes from healthy and diseased plants that were used for the microbiome analysis were analyzed using an untargeted metabolomics approach. Fifty milligrams of each sample was added to a 2-ml centrifuge tube, and a 6-mm diameter grinding bead was added. For the extraction of the metabolite, 400 μL of methanol:water (4:1 (v:v)) containing 0.02 mg/mL internal standard (L-2-chlorophenylalanine) was used. The samples were ground with a Wonbio-96c frozen tissue grinder (Shanghai Wanbo Biotechnology Co., Ltd.) for 6 min (− 10°C, 50 Hz), followed by ultrasonic extraction at a low temperature for 30 min (5°C, 40 kHz). The samples were kept at − 20°C for 30 min and then centrifuged for 15 min (4°C, 13,000g), after which the supernatant was transferred to an injection vial for LC‒MS/MS analysis in positive and negative ionization modes. A pooled quality control sample (QC) was prepared by mixing equal volumes of all the samples. The QC samples were disposed and tested in the same manner as the analytic samples. LC‒MS/MS analysis of the samples was conducted on a SCIEX UPLC-Triple TOF 5600 system equipped with an ACQUITY HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm; Waters, USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). The mobile phases consisted of 0.1% formic acid in water:acetonitrile (95:5, v/v) (solvent A) and 0.1% formic acid in acetonitrile:isopropanol:water (47.5:47.5, v/v) (solvent B). The flow rate was 0.40 mL/min, and the column temperature was 40°C. The UPLC system was coupled to a quadrupole time-of-flight mass spectrometer (Triple TOFTM5600⁺, Sciex, USA) equipped with an electrospray ionization (ESI) source operating in positive mode and negative mode. The optimal conditions were set as follows: source temperature, 550°C; curtain gas (CUR), 30 psi; Ion Source Gas1 and Gas2, 50 psi; ion-spray voltage floating (ISVF), − 4000 V in negative mode and 5000 V in positive mode; declustering potential, 80 V; collision energy (CE); and 20–60 eV rolling for MS/MS. Data acquisition was performed in information-dependent acquisition (IDA) mode. Detection was carried out over a mass range of 50–1000 m/z. Pretreatment of the raw LC/MS data was performed with Progenesis QI (Waters Corporation, Milford, USA) software, and a three-dimensional data matrix was exported in CSV format. Internal standard peaks, as well as any known false-positive peaks (including noise, column bleeding, and derivatized reagent peaks), were removed from the data matrix, and the peaks were pooled. Moreover, the metabolites were identified by searching databases, and the main databases used were the HMDB (http://www.hmdb.ca/), Metlin (https://metlin.scripps.edu/), and Majorbio databases. The data were analyzed through the free online platform Majorbio cloud (cloud.majorbio.com). At least 80% of the metabolic features detected in any set of samples were retained. To reduce the errors caused by sample preparation and instrument instability, the response intensity of the sample mass spectrum peaks was normalized by the sum normalization method, after which the normalized data matrix was obtained. The RSD was less than 0.3 for the overall dataset, and the peak ratio was more than 70%, so the overall data were suitable for subsequent analysis (Supplementary Fig. 3). The total set of metabolites identified was annotated using public databases, including KEGG (Kyoto Encyclopedia of Genes and Genomics, http://www.genome.jp/kegg/) and HMDB (Human Metabolome Database, www.hmdb.ca). Pearson’s correlation based on the Bray‒Curtis distance algorithm was used to evaluate the abundance of endophyte microbiome at the genus level and metabolites in the ginger plant compartments. Correlation analysis heatmaps were drawn, and KEGG enrichment analysis of the differentially abundant metabolites was performed using SciPy v.1.0.0 (Python) software. The differentially abundant metabolites were screened with the orthogonal projections to latent structures discriminant analysis (PLS-DA) model using the default criteria, with a variable importance value (VIP) ≥ 1 and a significance threshold of P < 0.001, using ropls v.1.6.2 (R software). Procrustes analysis of the Euclidian distances of eigenvalues for both the bacterial or fungal microbiome and metabolome datasets was executed to analyze the congruence of two-dimensional shapes produced from the superimposition of principal component analyses (PCAs) [45].

Results

Overview of the sequencing and de novo assembly

Data analysis of 36 composite samples from 6 microbial niches of healthy and diseased plants was carried out to characterize the microbial communities associated with the sampled ginger plants. Supervised taxonomic classification of all high-quality reads was performed using the SILVA and UNITE databases to examine the taxonomic structure of the bacterial and fungal communities, respectively. A total of 994,248 archaeal/bacterial and 1,650,996 fungal high-quality reads were obtained and sorted into 5353 archaeal/bacterial and 1793 fungal operational taxonomic units (OTUs). Archaeal/bacterial OTUs were assigned to 2 domains, 2 kingdoms, 44 phyla, 125 classes, 304 orders, 517 families, 1073 genera, and 2101 species (Supplementary Table 1), and the fungal OTUs were assigned to 15 phyla, 51 classes, 112 orders, 233 families, 426 genera, and 667 species (Supplementary Table 2). The saturated rarefaction curves (Supplementary Fig. 1A, B) and species diversity (Shannon index) for both the archaeal/bacterial (Supplementary Fig. 1C) and fungal communities (Supplementary Fig. 1D) indicated that the sampling efforts were adequate to reflect the microbial communities within each sample. Proteobacteria (57.40%), Actinobacteriota (16.74%), Bacteroidota (8.24%), and Firmicutes (6.83%) were the dominant bacterial phyla, while unclassified_k_Fungi (62.08%) and Ascomycota (32.53%) were the dominant fungal phyla.

Despite the great diversity, the ginger ecosystem’s bacterial and especially fungal communities were dominated by a few phyla among all samples, and we next examined how differences in the microbiome assembly can impact plant health.

Assemblage of plant-associated bacterial and fungal microbiota

Microbial composition in plant niches associated with plant health

To determine the microbial composition and relative abundance in the niches of healthy and diseased plants, we constructed pie diagrams to represent the number of generalist (shared) and specialist (inhabitants of a single niche) microbes between the niches of all the plants. The total number of microbes per niche included generalists, specialists, and those inhabiting more than one niche, known as satellites. A greater number of bacterial genera was found in the microbial communities of healthy plants compared to diseased plants, except in the rhizosphere soil, where the number of bacterial genera was higher in the diseased plants. A total of 3331 bacterial genera were identified in the healthy plants, with 138 unique to that group. In contrast, 2512 genera were detected in the diseased plants, with 58 unique to that group (Fig. 1). Only two representatives of archaeal microorganisms (g_norank_f_Nitrososphaeraceae and g_Candidatus_nitrocosmicus) were present in the analyzed soil samples analyzed, and their relative abundances were very low to be included in further analyses.

Eighty-three bacterial genera were identified as members of the core (generalist) bacterial microbiota (Fig. 1C, Supplementary Table 3). The most abundant bacterial genera were Flavobacterium (10.48%), Acidovorax (8.78%), Sphingomonas (7.92%), Methylobacterium-Methylorubrum (6.38%), and Bacillus (5.07%). Compared with the same niches in the diseased plants, all the organs of the healthy plants except for the stem harbored the largest number of endophyte bacteria (total; specialist); this trend was more notable in the leaves and rhizomes than in the other organs (Fig. 1A).

Healthy plants’ roots (588; 11) and rhizomes (550; 25) harbored the most significant number of endophyte bacterial genera. However, rhizome rot strongly reduced the number of these in the roots (220; 1) and rhizomes (332; 3). Diseased plants exhibited fewer specialist bacteria in all plant organs except the stems.

The most abundant fungal generalist genera were unclassified_k_Fungi (85.89%) and Gibellulopsis (5.20%) (Supplementary Table 3). The greatest number of fungal genera (total; specialist) was detected in the niches of healthy plants (922; 61) compared with diseased plants (833; 38) so the presence of the specialists was noticeably more affected by the disease. A greater abundance of fungal genera was observed in the rhizomes of diseased plants (70; 1) than in those of healthy plants (53; 0) (Fig. 1B).

Most of the taxa with relatively high abundances inside the diseased ginger plants were also detected in the soil, indicating that these taxa might have colonized the plants from the ground. Interestingly, the rhizomes of the diseased plants harbored a greater diversity of fungi than did those of the healthy plants, while the opposite occurred for bacteria.

Rhizome rot drives microbial community assembly in diverse plant niches

To quantify the diversity and summarize the structural changes in the microbial community, we first used the Kruskal‒Wallis test to calculate the microbial alpha diversity across all niches of healthy and diseased plants. The soil samples showed the highest diversity of bacteria (Fig. 2A, C) and fungi (Fig. 2B, D). The microbial communities in the rhizosphere were similar to those in the bulk soil, except for the fungal community in healthy plants, which was notably richer in the rhizosphere (Sobs index: 180 ± 21.6). Significant differences were observed in the bacteria and fungal populations among healthy plants.

The disease significantly reduced bacterial richness in both roots (healthy, 396.3 ± 56.2; diseased, 130.0 ± 26.2; P = 0.0439) and rhizomes (healthy, 353.7 ± 44.5; diseased, 185.0 ± 35.0; P = 0.0431) (Fig. 2A). Healthy plants had higher bacterial diversity in rhizomes (Shannon H': 4.30 ± 0.6). Diseased plants showed significant differences in fungal richness (Sobs index: 24.33 ± 9.3, Fig. 2B) and diversity (Shannon H': 0.04 ± 0.0, Fig. 2D) in rhizomes compared to healthy plants.

To assess the microbial community dissimilarity between the niches of healthy and diseased plants, principal coordinate analysis (PCoA) based on Bray–Curtis distance was performed (Fig. 2E, F). The closer the distance between samples in the PCoA diagram, the more similar their community composition. The analysis revealed differences in bacterial and fungal microbiome compositions between healthy and diseased plants. The first two axes account for about 50% and 47.5% of the variation for bacterial microbiomes (PERMANOVA, R = 0.70, P < 0.001; ANOSIM: R = 0.73, P < 0.001) and fungal microbiomes (PERMANOVA: R2 = 0.63, P < 0.001; ANOSIM: R = 0.39, P < 0.001), respectively. Different plant niches displayed distinct microbial communities, suggesting a potential link to plant health. These findings indicate that plant health is connected to unique microbial communities in various parts of the plant.

Additionally, functional signatures related to plant health status were predicted via FAPROTAX analysis based on the classification results from 16S amplicon sequencing. Testing for significance was performed using a Kruskal–Wallis rank sum test (Fig. 3A, Supplementary Table 4). The analysis predicted that the bacteria inhabiting the stems of diseased plants would have the highest functional potential for nitrogen (9.29 ± 2.19%), nitrate (8.05 ± 3.70%), and nitrite (8.79 ± 2.28%) respiration; nitrite (7.55 ± 1.85%) and nitrate (7.55 ± 1.55%) ammonification; nitrate reduction (10.47 ± 4.16%); and plant pathogens (8.37 ± 2.04%), presumably associated with the highest relative abundance of Pectobacterium (Fig. 3B).

Fig. 3

Functional analysis of the bacterial and fungal genera from the various niches of healthy and diseased ginger plants. Genus-level distribution of the bacterial (A) and fungal (C) microbiome associated with different ginger plant niches. The length of the bars represents the percentage of each microbial genus per sample. Top genera with an abundance > 1% in at least one sample are shown. Clustered heatmap of top thirty predicted bacterial functional profiles (B). The samples are grouped according to their similarity to each other, and the clustering results are arranged horizontally according to the clustering results. The color bar indicates the relative abundance of microbial functions from lowest (blue) to highest (red). Variations in the composition of the top fungal functional groups (> 5%) inferred by FUNGuild analysis (D). BS bulk soil, RhS rhizosphere soil, Rh rhizome, R root, S stem, and L leaf

The most common functional groups of fungi were undefined saprotrophs in the bulk (25.19 ± 2.21%) and rhizosphere (37.09 ± 2.90%) soils of healthy plants, while in the diseased plants, these functional groups were dominant in the bulk soil (13.91 ± 1.03%), roots (18.12 ± 1.92%), and rhizomes (17.12 ± 1.28%). Interestingly, the highest levels associated with the ecological guild of plant pathogens were observed in the rhizosphere (27.75 ± 2.08%) and rhizomes (27.75 ± 2.19%) of diseased plants (Fig. 3C), associated with an increased relative abundance of various potential pathogens in these microbial niches (Fig. 3D, Supplementary Table 5).

The alpha diversity analysis indicates that plant roots and rhizomes harbor a significant number and variety of bacterial microbes. However, the presence of rhizome rot disease reduced these indices. In contrast, plant disease increased the diversity of the fungal microbes. Beta diversity analysis revealed changes in the composition of microbial communities in rhizosphere soil due to plant disease. Additionally, the composition of the bacterial microbiome in rhizomes and roots differed from that in stems and leaves in healthy plants, but the disease nullified this difference.

Bacterial and fungal taxa potentially involved in plant health

We used linear discriminant analysis (LDA) effect size (LEfSe) to identify discriminative features at taxonomic levels for overall plant health regardless of the microbial niche. This study focused on potentially pathogenic and disease-suppressive microbes in soil and endophytes in plant tissues. In total, 105 taxa (from phylum to species) were identified with a log10 (LDA) score > 4.0 and a P value < 0.05.

In the LEfSe analysis, we found seven plant-endophyte bacteria (Fig. 4A) and five soilborne bacteria (Fig. 4B) that are biomarkers for plant health. Specifically, we observed that bacteria such as s_unclassified_g_Sphingomonas, Quadrisphaera granulorum, and Methylobacterium komagatae were significantly enriched in healthy plants. On the other hand, bacteria like P. carotovorum subsp. brasiliense, s_unclassified_f_Alcaligenaceae, Alcaligenes faecalis, and Klebsiella aerogenes were found to be significantly increased in diseased plants. Additionally, we found certain bacteria enriched in the soil of healthy and diseased plants.

Fig. 4

LEfSe (LDA effect size) multilevel species hierarchy tree diagram (cladogram) and latent Dirichlet allocation (LDA) discrimination results diagram for endophytic (A) and soilborne (B) bacterial community biomarkers for plant health. The cladogram demonstrates the classification of taxa at the five levels, and the different colors indicate the differences in relative abundance for microbes that inhabit healthy (red) or diseased (blue) plants. Nonsignificant differences are represented by yellow circles. The LEfSe bar chart shows the biomarkers with significant differences between healthy and diseased plants, and the lengths of the bars indicate the influence of the species. The LDA score threshold was log10 (LDA score) > 4.0. The higher the LDA score is, the greater the impact of species abundance on the difference effect

Four species of endophyte plant fungi (Fig. 5A) and ten soil-borne fungi (Fig. 5B) were identified as potential biomarkers. In healthy plants, s_unclassified_k_Fungi (from phylum to species) was significantly enriched. Biomarkers associated with s_unclassified_g_Cheilymenia (from class to species), Pseudaleuria sp. (from genus to species), Lophotrichus sp. (order to species), Pseudogymnoascus sp. (from class to species), Gymnoascus sp. (order to species), Mortierella polycephala (phylum to species), and Eleutherascus cristatus (from family to species) were significantly increased in the soil of healthy plants. In diseased plants, Gibellulopsis piscis (from phylum to species), Pyxidiophorales sp. (from class to species), and Plectosphaerella cucumerina (from phylum to species) were enriched, serving as potential biomarkers of disease. However, only three fungal biomarkers were characteristic of the soil of diseased plants: P. cucumerina (from genus to species), Trichoderm longibrachiatum (species), and Fusarium nematophilum (species).

Fig. 5

Fungal biomarkers for plant health based on the LEfSe (LDA effect size) multilevel species hierarchy tree diagram (cladogram) and latent Dirichlet allocation (LDA) discrimination results diagram for endophytic (A) and soilborne (B) microbes. The cladogram demonstrates the classification of taxa at the five level, and the different colors indicate the differences in relative abundance for microbes that inhabit healthy (red) or diseased (blue) plants. Nonsignificant differences are represented by yellow circles. The LEfSe bar chart shows the biomarkers with significant differences between healthy and diseased plants, and the lengths of the bars indicate the influence of the species. The LDA score threshold was log10 (LDA score) > 4.0. The higher the LDA score is, the greater the impact of species abundance on the difference effect

A probabilistic graph model related to a co-occurrence Bayesian network model (Supplementary Fig. 2A and B) shows a robust core bacterial and fungal microbiota and biomarkers linked to ginger plant health, including identifying Pectobacterium associated with rhizome rot in ginger plants. A greater network complexity has been associated with microbial communities exhibiting more intense activity and higher resilience to perturbation [16]. Our analysis of intrakingdom networks revealed a higher network complexity associated with increased nodes and edges in the bacterial networks (Supplementary Fig. 2B and C) than in the fungal networks (Supplementary Fig. 2D and E). The node average degree (206.48) and positive edges (48,524) were higher in the bacterial co-occurrence network in diseased plants than the node average degree (106.67) and positive edges (24,449) in healthy plants. However, the bacterial community in healthy plants had much higher opposing edges (1458) than in diseased ones (206). For fungal networks, the node average degree (35.97), positive edges (6,487), and opposing edges (23) were all higher in the healthy plants related to the node average degree (29.64), positive edges (4,816), and opposing edges (cero) in diseased plants.

Network statistics can determine the importance of microorganisms in co-occurrence networks [46]; in a co-occurrence network, hub or keystone species can be inferred by identifying species with the highest network centrality indices. The network analysis revealed that all bacterial biomarkers were highly prevalent in the system. Bacillus and Sphingomonas were identified as the most crucial nodes in the genus-level network within the healthy ginger ecosystem (Supplementary Table 6). The co-occurrence network for diseased plants revealed significant bacterial biomarkers. However, the ginger pathogen Pectobacterium [26] was identified as the top-ranking bacterium (Supplementary Table 7). Moreover, the fungal biomarker exhibited a significant correlation within the microbial co-occurrence network of healthy ginger plants. Notably, Pseudaleuria and Mortierella emerged as prominent nodes with high degrees within the top 10 hub nodes (Supplementary Table 8). Conversely, the fungal networks of diseased plants featured Plectosphaerella and Gibellulopsis (Supplementary Table 9). These results emphasize the potential significance of these microbial strains in preserving the health of ginger plants.

Metabolites driving ginger microbial community assembly and plant health

Overview of metabolite information

We used untargeted metabolomics to simultaneously detect and analyze small-molecule metabolites that impact microbiome assembly and the health of ginger plants. The metabolomes of the niches corresponding to the vegetative organs of healthy and diseased plants were analyzed via LC–MS/MS, which revealed a total of 10,415 chromatographic peaks with 735 metabolites, 500 of which were in the library (annotated to public databases like HMDB and Lipidmaps), and 199 of which were annotated to the KEGG database (Table 1, Supplementary Table 10).

Table 1 Total ion numbers and identification statistics

The metabolites identified across all the samples included 170 lipids and lipid-like molecules, 79 organic acids and derivatives, 63 organic oxygen compounds, and other compounds (Fig. 6A). The highest numbers of differentially accumulated metabolites (total; specific to each niche) were found in the rhizomes (164; 86), followed by the leaves (135; 63), roots (89; 35), and stems (76; 25). Interestingly, a metabolite associated with the health of the whole ginger plant (6-({[3,4-dihidroxi-4-(hidroximetil)oxolan-2-il]oxi}metil)oxano-2,3,4,5-tetrol) was identified (Fig. 6B).

Fig. 6

Overview of metabolite information related to plant health. A Superclass classification of the main types of metabolites found in diverse organs of healthy and diseased plants. B Number of differentially accumulated metabolites in the microbial niches of healthy plant organs. The histogram in the lower left corner refers to the number of overexpressed compounds in each microbial niche. The bar graph on the right shows the number of compounds after the intersection of various metabolic sets belonging to microbial niches. The single point below represents specific metabolite within the metabolite set, and the connection between some points shows the number of common metabolites

The plant health-associated microbiome is driven by the metabolome

The relationship between plant health-associated microbes and metabolites was examined. Procrustes analyses were performed using distance plots (PCA) as input based on the matrix of endophytic microbial communities (Bray–Curtis). Significant associations were found between certain bacterial (M2 = 0.58, P = 0.00) and fungal (M2 = 0.84, P = 0.04) genera and metabolite synthesis. The associations varied based on plant health and microbial niches (Fig. 7A). In diseased plants, only fungi in the rhizomes and roots were closely linked to metabolite synthesis (Fig. 7B).

Fig. 7

Associations between the endophyte microbiota and the plant metabolome. Procrustes correlation between the metabolome and the endophyte bacterial (A) or fungal (B) microbiota. M2 represents the sum of squared distances between matched sample pairs; the lower the value is, the greater the correlation between the two sets of data. The Monte Carlo P value was determined from 999 labeled permutations and provides a general measure of consistency between the two datasets (P < 0.01 indicates that the composition of the microbial community and the expression of metabolites are very consistent; P < 0.05 indicates consistency between these two datasets; and P > 0.05 indicates that the trend of association between the datasets is not significant). The connection represents the Procrustes residue of the two ordered configurations, which can be used to assess the variation between the two. The longer the connection is, the lower the consistency between the two datasets. A heatmap based on Pearson’s correlation indicated the associations between metabolites and bacterial (C) or fungal (D) communities at the genus level. Metabolites are shown on the right; microbial genera are shown at the bottom. Asterisks indicate Pearson’s correlation coefficient (*P < 0.05; **P < 0.01; ***P < 0.001). The “blue to red” color gradient indicates the Pearson’s correlation index value. A darker red color indicates a greater positive Pearson correlation coefficient, while a darker blue color indicates a greater negative correlation coefficient

Furthermore, the metabolites that drive the assembly of the potentially plant health-determining microbiota according to the LEfSe analysis are detailed below. Similarly, trans-EKODE-(E)-Ib (P = 0.0137) and 2,3-dinor prostaglandin E1 (P = 0.0359) were positively related to Sphingomonas, while piperidine (P = 0.0004), cyclohexane (P = 0.0006), tripropylamine (P = 0.0005), palmitoleamide (P = 0.0012), farnesyl acetone (P = 0.0031), and C16 sphinganine (P = 0.0055) were negatively correlated with this bacterial genus. trans-EKODE-(E)-Ib (P = 0.0341) was positively related, while ethyl hydrogen sulfate (P = 0.0008), 2-dodecylbenzenesulfonic acid (P = 0.0009), piperidine (P = 0.0034), farnesyl acetone (P = 0.0040), 9,12,15-octadecatrien-1-ol (P = 0.0042), 13Z-docosenamide (P = 0.0090), palmitoleamide (P = 0.0108), cyclohexane (P = 0.0227), palmitic amide (0.0358), p-chlorophenylalanine (P = 0.0424), and tripropylamine (P = 0.0452) were negatively correlated with Methylobacterium–Methylorubrum. Moreover, PA (16:0/18:2(9Z,12Z)) (P = 0.0120) and isocitrate (P = 0.0482) were positively related to Quadrisphaera. 2-Amino-4-methylpentanoic acid (P = 0.0421) and p-chlorophenylalanine (p = 0.0449) were positively related to Pectobacterium, while 6-{4-[3-(3,7-dimethylocta-2,6-dien-1-yl)-7-hydroxy-8-(4-hydroxy-3-methylbut-2-en-1-yl)-4-oxo-4H-chromen-2-yl]-3-hydroxyphenoxy}-3,4,5-trihydroxyoxane-2-carboxylic acid (P = 0.0364) and quercetin tetramethyl (5,7,3',4') ether (P = 0.0379) were negatively correlated with these bacteria. There was a significant positive correlation between linoleamide (P = 0.0185) and Alcaligenes, and a negative correlation between this bacterial genus and DG (18:4 (6Z,9Z,12Z,15Z)/18:2 (9Z,12Z)/0:0) (P = 0.0086) was detected. DG (18:4(6Z,9Z,12Z,15Z)/18:2(9Z,12Z)/0:0) was negatively correlated with Klebsiella (P = 0.005). DG (18:4(6Z,9Z,12Z,15Z)/18:2(9Z,12Z)/0:0) (P = 0.005) and 6-{4-[3-(3,7-dimethylocta-2,6-dien-1-yl)-7-hydroxy-8-(4-hydroxy-3-methylbut-2-en-1-yl)-4-oxo-4H-chromen-2-yl]-3-hydroxyphenoxy}-3,4,5-trihydroxyoxane-2-carboxylic acid (P = 0.0025) were negatively correlated with Enterobacter (Fig. 7C).

12,13-Epoxy-9,15-octadecadienoic acid (P = 0.0235), L-glutamate (P = 0.0353), and (E)-9,12,13-trihydroxyoctadec-10-enoic acid (P = 0.0481) were positively associated with Gibellulopsis, while PA(16:0/18:2(9Z,12Z)) (P = 0.0079) and LysoPA(0:0/18:2(9Z,12Z)) (P = 0.0207) were negatively related to this genus. 12,13-Epoxy-9,15-octadecadienoic acid (P = 0.0468) was positively correlated with Plectosphaerella. ( ±)9-HpODE (P = 0.0428) was positively correlated with Lophotrichus. Isocitrate (P = 0.0087) and 3'-methoxy-[6]-gingerdiol 3,5-diacetate (P = 0.0167) significantly affected the presence of Mortierella (Fig. 7D).

The metabolome can directly impact the health of ginger plants

To identify the compounds that play roles in plant health, the VIP combined with univariate statistical analysis was used. Of the total metabolites (anion plus cation), 470 (4.51%) were enriched and 711 (6.83%) were depleted in healthy ginger plants compared to diseased ginger plants (Fig. 8A). One hundred five annotated metabolites exhibited significant differences in abundance (Welch’s two-sided t test, P < 0.05) between the diseased and healthy ginger plants. In contrast, the abundances of 469 annotated metabolites were unchanged in either plant group (Fig. 8B). The abundance of 74 named metabolites was reduced, and 31 annotated metabolites were enriched in healthy plants compared to diseased plants. Particularly notorious, niacinamide, a heterocyclic aromatic amide (P < 0.001), the metabolic intermediates involved in de novo lipid synthesis 1-oleoyl lysophosphatidic acid (P < 0.001), and the phospholipid PG (16:0/16:0) (P < 0.05) were enriched in healthy ginger plants, while the nonproteinogenic L-alpha-amino acid 4-methylene-L-glutamine, the alkaloid xanthine, and the purine derivative hypoxanthine, among others, were significantly more abundant (P < 0.001) in diseased plants (Fig. 8C).

Fig. 8

Differences in the expression of all metabolites (A) and annotated metabolites (B) are associated with plant health. The first 12 metabolites in B are labeled according to the P value (Welch’s two-sided t test). The values of the abscissa and ordinate are logarithmic. C The top 30 differentially expressed metabolites with variable importance in projection (VIP) values in the PLS-DA model for each group comparison. The columns represent samples from healthy (HP) and diseased (DP) plants, and each row represents a metabolite. The bar chart on the right shows the VIP scores of the metabolites. The length of each bar indicates the contribution of the metabolite to the difference between the two plant groups. The color of the bar indicates the significance of the difference in metabolite levels between the two groups, that is, the P value. One asterisk (*) represents P < 0.05, two (**) represent P < 0.01, and three (***) represent P < 0.001

Metabolite profiles of plant niches revealed that niacinamide and PG (16:0/16:0) were upregulated in rhizomes (VIP value = 2.56, P = 0.0002, and VIP value = 2.37, P = 0.0012) and leaves (VIP value = 2.54, P = 0.0002, and VIP value = 2.36, P = 0.0012), while 1-oleoyl lysophosphatidic acid was upregulated in rhizomes (VIP value = 2.19, P = 0.0003) and roots (VIP value = 2.19, P = 0.0003) of healthy plants. In diseased plants, 4-methylene-L-glutamine was upregulated in leaves (VIP = 3.49, P = 0.0000). Hypoxanthine and xanthine also were upregulated in leaves (VIP = 3.16, P = 0.0000, and VIP = 3.29, P = 0.0000), and the latter was also in rhizomes (VIP = 3.16, P = 0.0000) (Supplementary Table 11).

The analyzed data support that the identified metabolites drive the assembly of the healthy endophytic microbiota and directly influence plant health. However, further research is required to define whether the metabolites come from the plants or their microbiota.

Discussion

We performed untargeted metabolomic and metataxonomic analyses based on 16S and internal transcribed spacer (ITS) rRNA gene amplicons to identify metabolome-driven microbiome changes associated with ginger plant health and rhizome rot disease. The key findings of our study present a comprehensive overview of the biodiversity of soilborne and endophytic microbiota in both healthy and diseased ginger plant environments. This highlights the bacterial and fungal microbes that may contribute to plant health, as well as the specific metabolites that play a role in healthy microbial assembly and overall plant health.

Members of Proteobacteria, such as Burkholderiales, Rhizobiales, and Enterobacteriales, were the predominant members of the global bacterial community in ginger plants. Actinobacteria, Bacteroidetes, and Firmicutes followed in abundance. This differs from the top four reported for natural ecosystems [46]. However, it has been reported that host species and soil type [47], crop rotation [48], and environmental conditions like temperature, relative humidity, and pH [49] cooperatively modulated microbiome assembly.

The global fungal assemblage of ginger plants was dominated by members of Ascomycota, with Hypocreale, Glomerellales, Pezizales, and Sordariales being the most abundant. The kingdom of fungi, including true fungi (Fungi) and fungus-like organisms (e.g., Oomycota), is the second largest group of organisms, with an estimated 2.2 to 3.8 million species worldwide [50]. Surprisingly, approximately 60% of the fungal taxa are classified as unclassified_k_Fungi, indicating a need for further analysis. More comprehensive information on the complete ITS sequence of these microbes in databases is required to address this issue.

Several agents can cause soft rot (rhizome rot) in ginger, but generally, the “bad guys” are fungi from the Fusarium [25] and Pythium [28] genera. Interestingly, despite sequencing, these soilborne fungus was rarely detected. Previous studies also failed to identify Pythium, possibly due to the limitations of the ITS region [51]. The ITS3/ITS4 primer set effectively analyzed soil fungal biodiversity in various soil types [52]. DNA metabarcoding targeting the ITS region revealed the widespread presence of potentially plant-pathogenic Phytophthora and Pythium species in rhizospheric soil associated with internationally transported plants [53]. However, the ITS region lacks sufficient resolution for distinguishing closely related species of indoor and foodborne molds, plant pathogens, or other fungi, for which secondary barcode markers have been suggested [50]. We identified these species using ITS3/ITS4 barcoding, except for oomycotes in the ginger ecosystem.

Further research is required to understand better the absence of such globally widespread fungal species in ginger ecosystems. However, manure application promotes saprotrophic fungi while suppressing potential soilborne pathogenic fungi [54]. Pectobacterium spp. use synchronized production of plant cell wall-degrading enzymes (PCWDEs) as their primary virulence attribute. These bacteria enter the host through stomatal openings and wounds, colonizing xylem vessels, parenchyma, and protoxylem cells [55].

At the genus level, 16S rRNA gene sequencing revealed Flavobacterium, Acidovorax, Sphingomonas, Methylobacterium-Methylorubrum, and Bacillus as the most abundant genera. These genera were shared across all the ginger microbial niches. Research on the assembly of the bacterial microbiota in the endosphere and rhizosphere of rice plants has identified Acidovorax, Sphingomonas, Bacillus, and Pseudomonas as members of the core generalist microbiota [56].

The diversity and species richness of the ginger microbiota narrowed from the soil as a “seed bank” to the plant organs, which suggest that the plants actively filtered the microbiota composition. Rhizome rot disease causes a significant change in the microbial community of ginger plants, especially in terms of microbial diversity. This change may be due to the plant’s reduced ability to filter organisms as the disease progresses.

The microbial structure detected in the rhizomes of both healthy and diseased plants revealed that specialist microbes did not cause rhizome rot. Instead, an imbalance caused by satellite [57] microbes like Pectobacterium was primarily detected in the stems and rhizomes of diseased plants. Saprotrophic fungi often take advantage of weakened diseased plants by colonizing their roots and rhizomes, while healthy plants maintain them in the soil. The presence of these fungi suggests that the cause of rhizome root disease is a necrotrophic pathogen that kills plant cells to feed on dead tissues and encourages the presence of other saprotrophs [58]. Most plant pathogens are mainly found in the rhizomes of diseased plants, although they have been discovered in all plant organs. This aligns with disease symptoms that spread to the entire plant.

Healthy plants harbor a more significant number and variety of bacterial microbes compared to diseased plants, while rhizome rot increases the diversity of fungal microbes. Changes in microbiota composition have been associated with immune suppression during pathogen infections. In the leaves of Arabidopsis immune-compromised mutants, the Shannon diversity index and the relative abundance of Firmicutes were significantly decreased, while Proteobacteria were more prevalent [59]. These findings are similar to some aspects of dysbiosis in human inflammatory bowel disease [60].

The higher diversity of endophytic bacteria in healthy plants is likely due to the abundance of beneficial bacteria. Conversely, diseased plants have a more diverse range of bacteria in the rhizosphere, possibly due to decaying roots providing nutrients for soil organisms. In a study involving tobacco plants infected with Ralstonia solanacearum wilt, researchers found that healthy plants had a greater diversity of microorganisms than diseased plants. They observed increased levels of certain bacteria that promote plant growth and suppress diseases [61]. Similarly, healthy mulberry plant samples exhibited greater diversity of beneficial bacteria compared to those infected with bacterial wilt [62].

Among the bacterial species important in keeping plants healthy, Q. granulorum is capable of nitrification, denitrification, and polyphosphate accumulation [63]. M. komagatae was reported to be a potential biostimulator against fungal pathogens of ginger [64]. Sphingomonas species have variable functions, ranging from the remediation of environmental pollution to the production of highly beneficial plant growth regulators [65], and some strains are also involved in nitrogen fixation [59]. Bacillus spp. serves multiple ecological functions, from soil nutrient cycling to inducing plant growth and stress tolerance [66].

In contrast, among the bacteria that were associated with the disease, only a P. brasiliense strain TS20HJ1 was isolated from ginger rhizome and shown to cause soft rot symptoms [25]. A. faecalis is a heterotrophic nitrifying bacterium that oxidizes ammonia and generates nitrite and nitrate [67], and K. aerogenes significantly enhances the production of plant biomass and plant secondary metabolites [68].

In relation to the fungi that were enriched in the disease-suppressive soil, Pseudaleuria had been negatively correlated with the disease severity index of Pisum sativum L. [69] and its abundance was favored by the application of manure rather than mineral fertilization [70]. A high abundance of Pseudogymnoascus in the rhizosphere contributes to the nutrient cycling and helps crops better adapt to the environment [57]; these fungi are antagonistic to potato scab pathogens [71]. Gymnoascus spp. can also antagonistically affect pathogens and promote plant growth [72]. Mortierella species promote plant growth and have beneficial effects by modifying the soil microbiological composition [73].

Interestingly, P. cucumerina served as a biomarker for the endophyte microbiota of diseased plants and soil. The Plectosphaerellaceae species G. piscis and P. cucumerina have been previously described as pathogens of essential crop plants [74, 75]. However, to our knowledge, these fungi have not been previously reported as pathogens of ginger.

Analysis of the correlation between microbial communities and metabolomes remains scarce. Specific metabolites can attract beneficial microbes that defend against pathogens, while others exclude specific species from the microbial community [14]. The results revealed a metabolome-associated deterministic assembly process in the microbiota of the various microbial niches of ginger plants. The highest number of differentially accumulated metabolites between healthy and diseased plants was found in the plant compartments that hosted a greater diversity of fungal microbes, i.e., rhizomes and roots.

Recent research on ginger has revealed detailed information about its over 60 bioactive compounds, including phenolic compounds, terpenes, polysaccharides, lipids, and dietary fibers [76,77,78,79]. Some compounds can attract beneficial microbes that protect the plant from pathogens, while others may harm the microbial community [13, 80]. Remarkably, our research has shown that lipids and lipid-like molecules are the most prevalent metabolites, among the more than 700 identified using untargeted metabolomics, that contribute to the health of ginger plants, particularly in preventing rhizome rot disease. Lipids, a principal constituent of cell membranes, act as the interface and mediate cell signaling pathways after microbe recognition, allowing advantageous resource exchange or inhibiting interaction through downstream signaling cascades [81, 82]. Furthermore, when plants are exposed to necrotrophic pathogens such as Pectobacterium species, their immune responses often involve oxylipins, signaling molecules derived from oxygenated fatty acids and related metabolites [83].

We hypothesized that the metabolites exhibiting more variability in abundance in healthy or diseased ginger plants may be closely associated with the plants' responses to disease onset. Interestingly, the organoxygen compound 6-({[3,4-dihydroxy-4-(hydroxymethyl)oxolan-2-yl]oxy}methyl)oxane-2,3,4,5-tetrol was the only overexpressed metabolite in all the vegetative organs of healthy plants related to those of diseased plants, but its role in plant protection needs to be elucidated.

The levels of numerous rice amino acids increased in response to high saline–alkali stress, with threoninyl-proline showing the most significant increase [84]. Glu-Val is a dipeptide composed of L-valine and L-glutamic acid residues. Amino acids and their metabolites have also been observed to stimulate the immune system in plants. Treating rice roots with Glu, and to a lesser extent Val, led to systemic disease resistance against rice blast (Magnaporthe oryzae) in leaves.

Niacinamide derivatives have been synthesized, and their fungicidal activity has been demonstrated [85]. Arachidonic acid (AA), a microbe-associated molecular pattern (MAMP) not commonly found in plants, is a potent elicitor of plant defense. Treating roots with AA-protected pepper and tomato seedlings from root and crown rot caused by Phytophthora capsici, leading to lignification at sites of attempted infection [86]. A relative of the ginger health biomarker M. polycephala, M. alpina has also been identified as an attractive AA producer [87]. In transgenic A. thaliana plants producing arachidonic acid, levels of jasmonic acid were increased, while levels of salicylic acid were decreased [88].

4-Methylene-L-glutamine is a nonproteinogenic L-alpha-amino acid that has been implicated in the transport of nitrogen [89]; coincidentally, the most prominent features of bacterial dysbiosis related to rhizome rot are related to the nitrogen cycle. Asparagine accumulation as part of nitrogen remobilization has been recorded in response to diverse abiotic and biotic stressors, such as disease and mineral limitation, as an adaptative process [90]. These changes in amino acids may be the result of disease in niches of ginger plants, although members of the Rhizobium complex of nitrogen-fixing bacteria were also enriched in the rhizome, stem, and leaves of diseased ginger plants.

Palmitoleamide is a primary fatty amide. A crude extract from the endophytic fungus Botryodiplodia theobromae containing fatty acid amides was observed to be broadly antimicrobial [91]. This metabolite was accumulated in stems of diseased ginger plants and showed a negative effect on microbes of the plant growth-promoting bacterial genus Methylobacterium–Methylorubrum. 4-Hydroxy nonenal alkyne, primarily detected in leaves of diseased ginger plants, is a significant aldehyde produced during the lipid peroxidation of ω-6 polyunsaturated fatty acids [92, 93].

Despite the limitations of this study, particularly concerning abundance thresholds for microbe inclusion, which need to be proven by culturomics methods, these limitations do not negatively impact the conclusions. Our findings provide a foundation for achieving disease suppression via modification of the metabolome-associated microbiome and have implications for further exploring pathogens, biocontrol agents, and plant growth promoters associated with economically important crop. Most microbial species and metabolites have not been previously identified in ginger plants. The assembly of the microbiota rather than the occurrence of a particular microbe drove plant health.