Experimental design
We established a series of mesocosms for banana cultivation in a greenhouse located at the WanZhong Co., Ltd. in Jianfeng town, Ledong County, Hainan Province, China (108°45′E, 18°38′N). Mesocosms were constructed from polypropylene pots (25 × 30 × 30 cm) filled with 10 kg soil. The soil was loam sandy dry red soil collected from a field with a history of more than 10 years of banana monoculture cultivation and a high level of Fusarium wilt disease (approximately 60% at the time of soil collection). The soil had a pH of 5.75, a total C content of 4.42 g/kg, a total N content of 0.63 g/kg, and available P, K contents of 68.88, 360.33 mg/kg, respectively. Four different fertilizer treatments were applied as follows: OF, soil amended with organic fertilizer; OF+W19, soil amended with bio-organic fertilizer containing B. amyloliquefaciens W19; SOF, soil amended with sterilized organic fertilizer; and SOF+W19, soil amended with sterilized organic fertilizer supplemented with B. amyloliquefaciens W19. The mesocosm experiment was performed using a randomized complete block design with three replicates for each treatment, and each replicate contained ten pots. Each pot received one banana seedling (Musa AAA Cavendish cv. Brazil), which was provided by Hainan Wan Zhong Co., Ltd [39]. Bio-organic and organic fertilizers were produced as described by Wang B [39]. Fertilizer sterilization was performed by Co75 γ-ray irradiation at Nanjing Xiyue Technology Co., Ltd, Nanjing, China. The population density of strain W19 in the SOF+W19 treatment was confirmed to be at least 1.0×109 CFU g-1 dry weight of fertilizer at the start of the experiment. Each pot was supplemented with 180 g of the given amendment before banana seedlings were transplanted for each of three successive seasons, with each successive season using soil from the previous year after plant removal. Incidence of Fusarium wilt disease was monitored as described by Jeger MJ [40] and calculated as the percentage of infected plants relative to the total number of plants.
Soil sampling and DNA extraction
Bulk and rhizosphere soil samples were collected 4 months after seedling transplantation for each season of the greenhouse experiment. Bulk soil samples were collected by first removing banana plants and then taking soil cores to a depth of 10 cm. Representative bulk soil samples were obtained by combining the samples from three pots in a given replicate and subsequent passage through a 2 mm sieve [36]. Sampling of rhizosphere soil was performed as described by Fu L [37]. Briefly, soil tightly bound to the roots was recovered by rinsing with sterile saline solution, and this soil suspension was centrifuged at 10 000 x g for 10 min, with the resulting pellet defined as rhizosphere soil. All bulk and rhizosphere soil samples were stored at -80oC prior to DNA extraction, and for each soil sample (24 in total: 4 treatments × 3 replicates × 2 positions (bulk and rhizosphere)), total soil genomic DNA was extracted from 0.5 g soil using the PowerSoil DNA Isolation Kit (Mobio Laboratories, Carlsbad, CA, USA) following the manufacturer's instructions. The concentration and quality of the DNA was determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA).
Tag sequencing for bacterial and fungal communities analysis
Bacterial and fungal sequencing libraries were constructed according to previously described protocols [41, 42]. Investigation of bacterial and fungal communities was based on paired-end amplicon sequencing of the 16S rRNA gene and the ITS region of fungal ribosomal DNA on an Illumina MiSeq PE 250 platform at Personal Biotechnology Co., Ltd (Shanghai, China). Amplification of bacterial 16S rRNA gene fragments was performed using the general bacterial primers 520F (5’-AYT GGG YDT AAA GNG-3’) and 802R (5’-TAC NVG GGT ATC TAA TCC-3’), which are specific to the V4 hypervariable region. The ITS region was targeted with the primers ITS1F (5’- CTT GGT CAT TTA GAG GAA GTA A -3’) and ITS2 (5’- GCT GCG TTC TTC ATC GAT GC -3’).
Bioinformatics analysis
Raw sequences were split according to their unique barcodes and trimmed of the adaptors and primer sequences using QIIME [43]. After removal of low-quality sequences, forward and reverse sequences for each sample were merged. The sequences retained for each sample were processed according to the UPARSE pipeline to generate an operational taxonomic unit (OTU) table [44]. Finally, a representative sequence for each OTU was selected [44] and classified using the RDP classifier [45] against the RDP Bacterial 16S database for bacteria [45] and the UNITE Fungal ITS database for fungi [46]. All raw sequence data have been made available in the NCBI Sequence Read Archive (SRA) database under the accession number SRP239482.
The relative abundance of a given taxonomic group per sample was calculated as the number of sequences affiliated to that group divided by the total number of sequences. Non-metric multidimensional scaling (NMDS) based on a Bray-Curtis dissimilarity matrix was performed and plotted using the R vegan package to explore the differences in microbial communities [47]. Permutational multivariate analysis of variance (PERMANOVA) was conducted to evaluate the effects of fertilizer type on the whole soil microbial community by using the R vegan package [47, 48]. While mantel test was implemented in the R vegan package to identify the correlation between soil microbial community and Fusarium wilt disease incidence [47].
Quantitative real-time PCR analysis
Quantitative real-time PCR amplifications (qPCR) were used to determine the abundances of total bacteria, fungi, Fusarium oxysporum, Bacillus and Pseudomonas in the bulk soil and banana rhizosphere, according to previously described protocols [49]. Abundances of bacteria and fungi were quantified with primers Eub338F / Eub518R and ITS1f / 5.8s, respectively (Table S1), according to Fierer N [50]. Standard curves were generated using 10-fold serial dilutions of a plasmid containing a full-length copy of the 16S rRNA gene from Escherichia coli and the 18S rRNA gene from Saccharomyces cerevisiae. The abundance of Fusarium oxysporum was determined using a SYBR Green assay with the primers FOF1 and FOR1 [51] (Table S1), targeting the rRNA internal transcribed spacer (ITS). A serial dilution from 108 to 102 gene copies ml-1 of the ITS gene from the Foc-TR4 strain was used as a standard. The abundance of Pseudomonas and Bacillus were determined using SYBR Green assays with the primers Ps-for / Ps-rev [52] and Bs16S1 / Bs16SR [53], respectively (Table S1). A serial dilution from 108 to 102 gene copies ml-1 of the 16S rRNA gene from Pseudomonas fluorescens and Bacillus subtilis strains were used as standards. Each assay was performed in triplicate, and the results were expressed as log10 values (target copy number g-1 soil) prior to further statistical analysis.
Assay of culturable Fusarium and Bacillus
To complement the results of the molecular methods described above, we also determined the population densities of culturable Fusarium and Bacillus in bulk soil and banana rhizosphere samples. This was carried out used using a standard 10-fold dilution plating assay as described by Wang B [39]. For enumeration of Fusarium, three aliquots (100 ml) per dilution were spread on Komada’s medium [54], and colonies were counted after incubation at 28°C for 5 days. For quantification of Bacillus density, three aliquots (100 ml) per dilution were spread on salt V8 agar Bacillus-semi-selective medium [55], and plates were incubated at 30°C for 2 days prior to colony counting.
Pseudomonas CFU quantification, strain isolation and identification, and assays of Fusarium inhibition, biofilm formation and Bacillus attraction
Given the demonstrated role of members of the genus Pseudomonas in disease suppression [17, 33], and the results from bacterial community analyses (see below), we tracked the density and functional potential of this genus by cultivation-dependent methods. Pseudomonas counts for all samples were determined by 10-fold dilution plating as described by Wang B [39]. Three aliquots (100 ml) per dilution were spread on CFC agar Pseudomonas-selective medium, and the resulting plates were incubated at 30°C for 3 days prior to colony enumeration. We also isolated Pseudomonas strains from the bioorganic fertilizer-treated and organic fertilizer-treated soils after two years of plant growth to compare their potential to inhibit F. oxysporum and their ability to produce biofilms. Strains were isolated from the same dilution series described above, using plates with one order of magnitude greater dilution than those used for cell enumeration. A total of 88 Pseudomonas isolates (50 and 38 from the OF+W19 and OF treatments, respectively) were purified and identified according to Su L [56]. The ability of Pseudomonas isolates to inhibit the growth of F. oxysporum was tested using a dual culture assay as previously described [57].
We examined biofilm formation of each of the 88 Pseudomonas isolates both independently and in co-culture with B. amyloliquefaciens W19. Biofilm formation was assayed and quantified as previously described by Ren D [58]. Briefly, exponential phase cultures of Pseudomonas isolates and B. amyloliquefaciens W19 were adjusted to an optical density at 600 nm (OD600) of 0.15 in tryptic soy broth medium and then inoculated into Nunc-TSP plate. The inoculum volumes were 160 ul for TSB and 40 ul of bacterial suspensions (40 µl of W19 or each Pseudomonas isolate for monoculture assays and 20 µl of each Pseudomonas isolate + 20 µl of W19 for co-culture assays). After 72 h incubation at 30oC, biofilm formation was quantified by a modified crystal violet (CV) assay [59, 60]. Interactive effects on biofilm formation were calculated by comparing two-species biofilm results (Abs570 TB) to those of each individual Pseudomonas isolate (Abs570 PB), as well as B. amyloliquefaciens W19 (Abs570 BB) in monoculture. Results were subsequently presented as follows: Abs570 TB > Abs570 BB and Abs570 TB > Abs570 PB (t-test, P < 0.05) = biofilm enhancement [58].
Attraction between the B. amyloliquefaciens W19 and Pseudomonas isolates was quantified using petri-dish confrontation assays as described by Berendsen RL [61]. Briefly, each Pseudomonas isolate and B. amyloliquefaciens W19 was inoculated in 5 mL TSB medium and incubated overnight at 30oC at 180 rpm. The optical density of the bacterial cultures was adjusted to 0.1 at 600 nm. Five times 1 µl of these dilutions were inoculated in a diagonal row on both sides of a petri-dish with TSB agar with a multichannel pipet, creating a V-shape of inoculation sites with increasing proximity. Plates were sealed with Parafilm and incubated for 7 days at 25oC. Colony diameters were measured on an orthogonal to the line dividing the V-shape for calculation of antagonistic effects.
Effects of selected Pseudomonas strains on plant disease levels
We carried out plant-based disease inhibition assays using strain PSE78, which belonged to the most responsive OTU in the OF+W19 and SOF+W19 treatments based upon community sequence analysis (OTU7; see below). This strain also exhibited the strongest Fusarium inhibition and strongest stimulation of biofilm formation in co-culture with B. amyloliquefaciens W19 (see below). We also selected an additional strain, PSE82, which lacked these exceptional qualities to allow comparison. Pot experiments with banana were performed using the following four fertilizer treatments: PSE78, sterile organic fertilizer + strain Pseudomonas sp. PSE78; PSE82, sterile organic fertilizer + strain Pseudomonas sp. PSE82; SBF, sterile organic fertilizer; and CK, chemical fertilizer to the same nutrient levels as achieved by organic fertilizer amendment. The density of each Pseudomonas strain was confirmed to be at least 1.0×109 CFU g-1 dry weight of fertilizer at the start of the experiment. Experimental design and conditions were identical to those used in the main mesocosm experiment described above.
Effects of Bacillus-Pseudomonas co-culture on FOC density
Banana tissue culture seedlings were cultivated in Erlenmeyer flasks and watered with modified strength sterile Hoagland solution. After two weeks, seedlings were transferred to 400-mL pots filled with a sterile substrate pre-inoculated with B. amyloliquefaciens W19, Pseudomonas sp. PSE78 or Pseudomonas sp. PSE82, or a combination of W19 mixed with either PSE78 or PSE82. In all cases, the final inoculation density was 1×108 CFU/g of substrate. The pots (10 replicates per treatment with 3 times experimental repeated) were placed on small saucers, watered with modified strength Hoagland solution, randomly placed in trays and transferred to a growth chamber (28oC average temperature, 80% relative humidity, 16 h light/8 h dark). After thirty days, all plants were transplanted into a new sterile substrate. Banana plants were then inoculated with a Fusarium oxysporum f. sp. cubense (FOC) spore suspension (final density of 1×104 spores/g of substrate as describe above) or a mock suspension. Disease severity was quantified by counting the density of FOC colonizing banana plant roots three weeks after FOC inoculation. FOC, Bacillus and Pseudomonas densities in the banana roots were determined by suspending approximately 0.1 g of root of eight replicate pots per treatment and plating a dilution series on Komada’s medium, V8 agar Bacillus-semi-selective medium, and CFC agar Pseudomonas-selective medium as described above, respectively. Fig. S11 provides a schematic representation of this experiment.
Statistical analyses
All statistical analyses were performed by using the IBM SPSS 20.0 software program (IBM Corporation, New York, USA) and R software programs (Version 3.5.0). All statistical tests performed in this study were considered significant at P < 0.05. To determine significant differences, unpaired t-tests and one-way ANOVA were performed. Testing of linear discriminant analysis effect size (LEfSe) was performed to identify significant differences in bacterial and fungal taxa between fertilization regimes[62]. The Kruskal-Wallis (KW) sum-rank test was used in LEfSe analysis to detect the features with significantly different abundances between assigned classes, and linear discriminant analysis (LDA) was then performed to estimate the effect size of each differentially abundant taxon [62]. Spearman's rank correlation coefficients between the relative abundance of OTUs and Fusarium wilt disease incidence were calculated in R software. P-value adjustments for multiple comparisons were performed using the false discovery rate (FDR) correction [63]. Fold change of each OTU in treatments with the biocontrol agent (OF+W19 and SOF+W19) relative those without the biocontrol agent (OF and SOF) was calculated using the following formula: (B-N)/N, B is the relative abundance of a given OTU in Bacillus positive samples (OF+W19 and SOF+W19) samples and N represents the relative abundance of that OTU in Bacillus negative (OF and SOF) samples [64]. Structural equation modelling (SEM) was applied to evaluate the direct and indirect contributions of soil microbial community (bulk and rhizosphere soil) and F. oxysporum pathogen density to disease incidence [65]. The SEM fitness was examined on the basis of a non-significant chi-square test (P > 0.05), the goodness-of-fit index (GFI), and the root mean square error of approximation (RMSEA) [66, 67]. Model was fit using the lavaan package in R software [68]. The linear regression analyses relating disease incidence to the selected microbial taxa were conducted using the basicTrendline package in R software.