FWD affected pepper microbiome assembly
In total, 8,672,206 bacterial 16S and 7,677,988 fungal ITS high-quality reads from 144 samples were obtained, and these reads were sorted into 14,976 bacterial and 4,277 fungal ZOTUs. To examine the dimensions in which the multiple factors shape the pepper microbiome, we assessed the relative contribution of multiple factors in terms of plant compartment, Fusarium wilt disease (FWD), and sampling site in shaping the microbial communities. NMDS ordinations and PERMANOVA analysis showed that the greatest effect on the total microbiome was exerted by the compartment (R2 = 0.47 for bacteria and R2 = 0.53 for fungi), followed by FWD (R2 = 0.06 for bacteria and R2= 0.03 for fungi), and lastly by sampling site (R2= 0.01 for bacteria and R2= 0.02 for fungi) (Fig.1b and Table S2). FWD explained higher variation of fungal community than of the bacterial in compartments of root endosphere, stem bottom epidermis, stem middle epidermis and xylem, stem upper epidermis and xylem, and fruit episphere (Fig. 1c, Table S3, and Table S4). Notably, fungal community was less affected by FWD in pepper fruit than that in stem and root (root/stem/fruit R2: 0.17/0.22/0.15, in average, respectively; Fig. 1c and Table S4). For stem, FWD effect on fungal community was stronger in epidermis than that in xylem (epidermis/xylem, R2: 0.24/0.13 in stem bottom, 0.23/0.14 in stem middle, 0.39/0.16 in stem upper, respectively; Fig. 1c, Fig S4, and Table S4). Through all compartments, the greatest effects of FWD on fungal community were found in stem upper epidermis and root endosphere (R2 = 0.39, P = 0.001 in stem upper epidermis, and R2 = 0.25, P = 0.001 in root endosphere) (Fig. 1c, and Table S4). In contrast, sampling site explained higher variation of bacterial communities than FWD in most compartments (Fig. 1c, Fig S3, and Table S3). In addition, both bacterial (P = 0.012) and fungal communities (P = 0.037) in diseased plant were more variable than the healthy based on beta-dispersion using Bray-Curtis dissimilarity (Fig. 1d). In the bacterial community, the diseased plant were more variable than the healthy in bulk soil, root endosphere, stem bottom epidermis and xylem, stem upper epidermis, and fruit episphere (Table S6). In the fungal community, the diseased plant were more variable than the healthy in bulk soil, rhizosphere soil, stem middle xylem, stem upper epidermis, and fruit episphere (Table S6).
We used linear mixed models (LMMs) to explore the most important driver of microbial alpha diversity, and the result showed that both bacterial and fungal Shannon diversity were mainly influenced by compartment (P < 0.0001, Table S5). FWD had stronger effect on fungal Shannon diversity (P = 0.00172) than on bacterial (P = 0.023, Table S5). Remarkably, fungal Shannon diversity significantly decreased in stem upper epidermis (30%), stem bottom epidermis (28%), root endosphere (28%), and rhizosphere soil (19%) under FWD (P < 0.05, Fig. 1f). In contrast, sampling site had stronger effects on bacterial Shannon diversity (P = 0.006) than on fungal community (P = 0.831, Table S5).
The LMMs of compositional variations showed that FWD had significant effects on the relative abundance of fungal class Tremellomycetes (P < 0.05) but not on any bacterial phyla (Table S7, Fig S5h, and Fig S6a). The differential abundance analysis showed a higher relative abundance of class Tremellomycetes in the diseased plant (Fig. 2a), and this class belonged to the same fungus functional guild Saprotroph (Yeast) (Fig S6b). The relative abundance of several pathogenic fungi in the genera Diaporthe, Fusarium, Gibberella, Phomopsis, Plectosphaerella, Stemphylium, and Cryptococcus were also significantly higher in diseased plant root and stem (P < 0.001), but not in fruit (Fig. 2a&c). However, several potential beneficial taxa in the genera Pseudomonas, Streptomyces, Klebsiella, Enterobacter, Microbacterium, Bacillus, Chitinophaga, and Citrobacter were significantly enriched in the diseased plant (P < 0.001, Fig. 2b&d and Fig S6c).
FWD affected pepper microbiome co-occurrence network
To further explore how FWD effects on pepper microbiome co-occurrence patterns, we conducted bacterial-bacterial, fungal-fungal intra-kingdom networks, and bacterial-fungal interkingdom networks. The intra-kingdom co-occurrence networks analyses showed that the bacterial networks had properties that suggest higher stability than fungal networks. A higher number of nodes and edges were recorded in bacterial networks than fungal networks (Fig. 3 and Table S8). Further, the edges of top 10 hub nodes that had higher degree and centrality values (closeness) in bacterial networks were primarily negative particular for the healthy network (Fig. 3b&c). In contrast, most edges of fungal networks were primarily positive (Fig. 3b&c). Moreover, the healthy bacterial network showed higher complexity (based on the number of nodes and edges) than the diseased network, with the contrasting pattern in the fungal network (Fig. 3b, d&e and Table S8).
The interkingdom co-occurrence networks indicated a more important role of fungal taxa played in diseased network. The number of nodes and connections of fungal taxa increased in diseased network compared with the healthy one, while an opposite pattern was observed in bacterial taxa (Fig. 4a-c and Table S8). These increased connections of fungal taxa were mainly related to bacterial-fungal (BF) and fungal-fungal (FF, especially the positive one) correlations, and the decreased edges of bacterial taxa were related to the bacterial-bacterial (BB) correlations (Fig. 4d). Additionally, the interkingdom correlations between bacterial and fungal taxa (BF) were primarily negative (92.1% in healthy and 78.3% in diseased), whereas positive correlations dominate intrakingdom correlations (60% BB and 98% FF in healthy, and 66% BB and 99% FF in diseased; Fig. 4d). The top 10 hub species were wholly belonged to bacterial taxa in the healthy network, while fungal taxa accounted for a half number in the diseased network (Fig. 4e&f, Table S10). The similar patterns were also revealed in most single compartment networks (Fig S7). Overall, the intra- and interkingdom networks analyses indicated higher stability properties in the bacterial networks than the fungal networks, and FWD decreased the complexity of bacterial taxa, whereas increased the complexity of fungal taxa.
FWD affected pepper microbiome functions
Metagenomic sequencing was conducted to explore the impact of FWD on the functional genes in pepper microbiome; two compartments (stem upper epidermis and root endosphere) having high variations of microbiome assembly between healthy and diseased pepper were selected for this analyses.
The results indicated significant effects of FWD on KO, CAZ, and ResFam functional profiles in stem upper epidermis microbiome (P < 0.05, Fig.5a), but FWD had no significant effects in root endosphere microbiome in any functional profiles (P > 0.05, Fig S8a). FWD significantly decreased the functional diversity of KO (P = 0.0314), COG (P = 0.0074), and Resfam (P = 0.0065) profiles in stem upper epidermis microbiome, but showed no significant effects in root endosphere microbiome (P > 0.05, Fig. 5b).
We performed differential abundance analysis to identify how FWD affected the functional properties. The results demonstrated that the microbiome in stem upper epidermis possessed higher number of specific functional genes than in root endosphere (Table S11). The phoD Alkaline Phosphatase Gene (K01113) and mprF:peptide antibiotic resistance gene was significantly enriched in healthy root endosphere, and vancomycin resistance gene clusters were significantly enriched in healthy stem upper epidermis (P < 0.05, Fig. 5c, Fig S8c&f, and Table S12). Notably, in diseased plant, functions of csgD LuxR family transcriptional regulator (K04333) was significantly enriched in root endosphere, as well as UDP-glucuronosyltransferase (GT1) and replication, recombination and repair (COG_L) were significantly enriched stem upper epidermis (P < 0.05, Fig. 5c, Fig S8c-e, and Table S12).