Introduction

Continuous cropping is the practice of cultivating the same crop in the same soil year after year and can potentially induce different crop-specific microbial communities [1]. This usually results in yield decline and soil-borne plant pathogen enrichment in annual crops, such as cucumber, cotton, and potato [2–4]. Perennial plants usually grow slowly and erratically and suffer from successive cropping obstacles when they are replanted [5–7].

Vanilla (Vanilla planifolia), a herbaceous perennial vine with high economic value, has been widely cropped in tropical and subtropical regions [8]. However, due to long-term monoculture, various types of wilts caused by soil-borne pathogenic fungi have been reported in all of the vanilla-growing countries [9], particularly stem and root rot caused by Fusarium oxysporum f. sp. vanillae, a devastating and widely distributed disease in many vanilla plantation areas [10, 11]. In China, this disease has attacked most long-term monoculture vanilla fields, resulting in vast economic losses [12].

Soil microorganisms are extremely important for maintaining soil health, which is one of the most important requirements for plant production in agricultural systems [13–15]. Many previous studies revealed that soil microbial communities were affected by various factors, including plant species, soil types, organic amendment, and agricultural management [16–19].

Recently, increasing numbers of studies have speculated that continuous cropping resulted in the disruption of soil microbial community membership and structure [4, 20]. Additionally, fungal pathogen populations easily accumulated in continuous cropping agricultural regimes [21, 22]. However, the detailed effects of continuous cropping on soil microbial communities and the link between these effects and soil sickness remain unclear [22]. Particularly, in tropical areas, few studies have been documented on the variations of soil microbial communities after monoculture of different tropical crops, including vanilla. To our knowledge, Zhao et al. [12] reported that the population of culturable bacteria in vanilla rhizosphere soil decreased with the years of vanilla monoculture, whereas fungal numbers increased over time.

16S ribosomal RNA (rRNA) and 18S rRNA gene library construction and denaturing gradient gel electrophoresis (DGGE) methods have been predominantly used in previous researches on soil microbial communities under continuous cropping system; however, only certain dominant microbial groups can be detected using these methods [21, 23]. The development of next-generation sequencing (NGS) technologies, such as 454 pyrosequencing, offers a powerful strategy for studying soil microbial diversity and community structure with high throughput, low price, and high accuracy and in a short time [24, 25]. Recently, NGS has been widely applied to investigate agricultural soil microbial communities both for bacteria [26–28] and fungi [29, 30].

Many investigations have focused on the soil microbial communities of continuous cropping systems using NGS, but only either soil fungal or soil bacterial communities have been targeted in a single experiment [4, 20, 31]. In this study, both the bacterial and fungal communities in vanilla-grown soil from different time-series fields were investigated by the 454 FLX+ sequencing technology using the V4-V5 hypervariable regions of 16S rRNA gene and internal transcribed spacer (ITS5-ITS4) as bacterial and fungal barcode markers, respectively. The objectives of the present study were (1) to compare the compositions and structures of the bacterial and fungal communities in the field soil samples taken from four time-series vanilla fields with entirely different production abilities and (2) to explore the potential correlations between the underlying microbial communities or prevalent taxa with vanilla Fusarium wilt disease.

Materials and Methods

Site Description and Sampling

The experimental site was located at the Tropical Spice and Beverage Research Institute of the Chinese Academy of Tropical Agricultural Sciences, Xinglong, Hainan Province (110°20′ E, 18°73′ N), which has a tropical monsoon climate. The mean annual temperature and precipitation in the area are 24.5 °C and 2201 mm, respectively. The cultivar of the planted vanilla (V. planifolia Andrews), agronomic management, and fertilization regime were similar between time-series fields. The soil samples were collected from four time-series vanilla fields with 1, 6, 11, or 21 years of continuous cropping history and marked as “a,” “b,” “c,” and “d,” respectively. For each time-series field, 45 cores (0–20 cm in depth and 2.5 cm in diameter) were randomly collected after removing vanilla plants and surface coverings, and 15 random cores were mixed for one sample (each time-series field had three replicates), and all 12 soil samples were put into separate sterile plastic bags and transported to the laboratory on ice. After sifting through a 2-mm sieve and thoroughly homogenizing, one portion of each sample was air-dried for chemical analysis and the other portions were stored at −70 °C for subsequent DNA extraction.

Determination of Soil Physicochemical Properties

Soil pH and electrical conductivity (EC) were measured using a glass electrode meter and a conductivity meter, respectively, in a soil water suspension (1:5 w/v) after shaking for 30 min. Organic matter (OM) and available N (hydrolyzable N) were determined using the potassium dichromate external heating method [32] and the alkaline-hydrolyzable diffusion method [33], respectively. Available P was extracted with sodium bicarbonate and then determined using the molybdenum blue method. Available K was extracted with ammonium acetate and determined by flame photometry [34]. Soil exchangeable Ca and Mg were detected using acetic acid extraction and atomic absorption spectrophotometer [35].

Evaluation of Soil-Borne Disease Using a Pot Experiment

Based on visual observation of the vanilla in four time-series fields, we found that the growths of vanilla were seriously hindered with the years of monoculture. However, the accurate field Fusarium wilt disease index (DI) could not be obtained because vanilla was replanted after the removal of diseased plants from the fields. To confirm the observations from the vanilla four time-series fields, a pot experiment was performed to evaluate disease development for the respective field soils. Stems of vanilla with five internodes (65 ± 5 cm) were selected for cutting seedlings, and three of the vanilla cutting seedlings were planted in one pot (12 kg soil). Each time-series treatment had three replicates, and each replicate contained nine vanilla cutting seedlings in three pots. All pots were randomly placed in a greenhouse located at the Tropical Spice and Beverage Research Institute in the Chinese Academy of Tropical Agricultural Sciences, Xinglong, Hainan Province, China, which has an average temperature of 30 °C and an average humidity of 72 % from April to June, 2013. Over 2 months, no fertilizer was used, and irrigation was performed every 3–4 days to keep the soil moisture constant. DI was calculated using the following formula according to Wei et al. [36] as DI = [∑(number of diseased plants in this index × dsi)/(total number of plants investigated × highest dsi)] × 100 %. Disease severity index (dsi; 0–5 scale) was obtained according to Xu et al. [30] with some modifications, where 0 = healthy vanilla stem without any visible symptoms, 1 = 1 to 25 % stem discoloration, 2 = 26 to 50 % stem discoloration, 3 = 51 to 75 % stem discoloration, 4 = 76 to 99 % stem discoloration, and 5 = plant completely dead.

DNA Extraction and PCR Amplification

Total DNA from field soil was extracted using Power Soil DNA Isolation Kit (MoBio Laboratories Inc., Carlsbad, USA), following the manufacturer’s protocol. The concentration and quality (ratio of A260/A280) of the DNA were determined using a spectrophotometer (NanoDrop 2000, Thermo Scientific, USA). DNA extracted from each soil sample served as a template in 16S rRNA gene and ITS region amplification. Bacterium-biased primers 515F (5′-GTGCCAGCMGCCGCGG-3′) and 907R (5′-CCGTCAATTCMTTTRAGTTT-3′) were used to amplify ∼400-bp fragments spanning the V4 to V5 hypervariable regions of the bacterial 16S rRNA gene [37]. Fungi-specific primers ITS5 (5′-GGAAGTAAAAGTCGTAACAAGG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) were employed to amplify ∼600-bp fragments of the fungal internal transcribed spacer (ITS) region [38]. These primer pairs were tailored for pyrosequencing by adding the Roche adapter B, Key and a unique 7-bp barcode at the 5′ end of the forward primer and the Roche adapter A and Key at the 5′ end of the reverse primer. PCR amplification was performed in a 25-μl reaction: 2.5 μl of ×10 reaction buffer, 10 μM of each primer, 2.5 mM dNTPs, 40 ng of template, and 0.625 units of Takara Pyrobest (Takara Biotechnology Co., Ltd., Japan). Amplifications were performed with the following temperature program: 4 min of initial denaturation at 94 °C, followed by 25 cycles of denaturation (94 °C for 30 s), annealing (55 °C for 45 s), extension (72 °C for 1 min), and a final extension at 72 °C for 7 min for the 16S V4-V5 rRNA gene; and 4 min of initial denaturation at 94 °C, followed by 35 cycles of denaturation (94 °C for 30 s), annealing (50 °C for 45 s), extension (72 °C for 1 min), and a final extension at 72 °C for 7 min for ITS genes. For each sample, three independent PCR amplifications were performed, and the products were pooled. Then, the products for all samples were purified using a PCR Purification Kit (Axygen Bio, USA), and the amplicons from each sample were pooled in equimolar concentrations into a single tube, and emulsion PCR was performed to make the single strands on beads as required for 454 barcode pyrosequencing. Pyrosequencing was performed on a 454 GS-FLX+ sequencer (454 Life Sciences) at Personal Biotechnology Co., Ltd (Shanghai, China).

Pyrosequencing Data Processing

Bacterial sequences were analyzed using Mothur v. 1.25.1 following the Schloss SOP [39]. Briefly, reads with a minimum flow length of 360 flows were de-noised using the Mothur-based re-implementation of the PyroNoise algorithm with the default parameters. Sequences with any ambiguous base, a homopolymer length equal or greater than 8 bp and shorter than 200 bp, were removed from the dataset. Then, the filtered sequences were assigned to soil samples based on their unique 7-bp barcodes. After removal of the barcode and primer sequences, the unique sequences were aligned against the Silva 106 database. Through screening, filtering, pre-clustering, and chimera removal processes, the remaining sequences were used to build a distance matrix with a distance threshold of 0.2. Using an average neighbor algorithm with a cutoff of 97 % similarity, these sequences were clustered into operational taxonomic units (OTUs). Representative sequences from each OTU were taxonomically classified with a confidence level of 60 % using an RDP Bayesian approach.

Fungal sequences were analyzed according to Lu et al. [4] with some modifications. Briefly, quality control was first performed using Mothur v. 1.25.1, allowing no more than two mismatches to the forward primer sequence, no more than one mismatch in the barcode sequence, and no homopolymeric runs of more than eight nucleotides. The remaining sequences were assigned to each soil sample based on their barcodes. Then, the full-length ITS1 subregion was extracted from the fungal ITS dataset using ITS Extractor v. 1.1 [40] because the presence of fragments might distort sequence clustering and dissimilarity searches [41]. After this filtering, only ITS1 region sequences remained and were further aligned using CAP3 [42]. The OTUs were assigned at a 97 % identity level using a minimum overlap of 100 bp. Ultimately, the OTU representative sequences were identified against the UNITE [43] and INSD [44] databases. OTUs at 97 % sequence similarity that could not be identified using the UNITE and INSD databases were recovered by a BLAST search against the NCBI database.

It is worth noting that the singletons (OTUs that contain only one sequence in all 12 samples) were removed for the downstream analyses.

Statistical Analyses

For all parameters, one-way analyses of variance (ANOVA) with Turkey’s HSD multiple range tests were performed for multiple comparisons. Pearson correlation coefficients between vanilla stem rot DI and the abundances of microbial taxa, soil properties, and phyla were all calculated using SPSS v20.0 (SPSS Inc., USA). For α-diversity, all analyses were based on the OTU clusters with a cutoff of 3 % dissimilarity. The Chao1 and abundance-based coverage estimator (ACE) were calculated to estimate the richness of each sample. The diversity within each sample was estimated using the non-parametric Shannon diversity index. Good’s coverage estimator was used to calculate the percentage of the total species that were sequenced in each sample. Rarefaction curves with average number of observed OTUs were generated using Mothur to compare relative levels of bacterial and fungal OTU diversity across the four time-series vanilla field soils. For β-diversity, hierarchical cluster dendrograms (with Bray-Curtis distance dissimilarities) were performed using Mothur based on the OTU composition for comparing the bacterial and fungal community structures across all soil samples. The weighted and unweighted UniFrac distance metrics (based on phylogenetic structure) [45] were used to generate PCoA plots to further assess the similarities between the samples’ community memberships. Lastly, Venn diagrams were constructed to visualize shared and unique OTUs between samples.

Sequence Accession Numbers

Sequence data have been deposited in the NCBI Sequence Read Archive (SRA) database with the accession number SRA149260.

Results

The Physicochemical Properties of Soils from Four Time-Series Fields and Pot Experiments

With the increased years of vanilla monoculture, soil organic matter (OM), available P and exchangeable Ca contents, and pH significantly increased, whereas soil electrical conductivity (EC) and exchangeable Mg content significantly decreased (Table S1).

In the vanilla pot experiment, stem rot disease index was significantly increased with the number of continuous cropping years. Vanilla growth was almost devastated in the soils collected from the field with 21 years of vanilla continuous cropping history, registering a DI value of 27.78 (Fig. 1), whereas the DI value was only 2.22 in the soils collected from only 1 year of vanilla cropping history. In addition, the average dry weight of the new shoots per vanilla significantly declined after 11 years of monoculture. This result suggests that vanilla should be mono-cultured for less than 10 years, to avoid significant losses.

Fig. 1
figure 1

Change in vanilla disease index and average dry weight per new vanilla shoot across vanilla continuous cropping time-series fields

Full size image

Composition of Bacterial and Fungal Communities

As shown in Table 1, the coverages of all samples, regardless of bacteria or fungi, were above 93 %, indicating that the depth of sequencing could meet the needs of our experiments. After filtering the reads based on basal quality control and removing singleton OTUs, a total of 121,026 sequences comprising a total of 4923 bacterial OTUs were obtained from 12 samples, and the numbers of high-quality sequences per sample varied from 7366 to 16,692. The classified sequences across all samples were affiliated in 18 bacterial phyla, and the remaining sequences were unclassified. The dominant phyla across all samples were Proteobacteria (35.98 %), Acidobacteria (26.28 %), Bacteroidetes (6.99 %), Firmicutes (3.06 %), Planctomycetes (2.04 %), Actinobacteria (1.45 %), and Nitrospira e (1.43 %) (relative abundance (RA) >1 %) (Fig. S1). In addition, Chloroflexi, Gemmatimonadetes, Verrucomicrobia, Armatimonadetes, WS3, and TM7 were present in most soils but at relatively low abundances (RA 0.1–1 %), and 5 other rarer phyla belonged to others (RA <0.1 %).

Table 1 Bacterial and fungal α-diversity indexes of soil from four time-series vanilla fields
Full size table

After filtering reads by basal quality control and removing singleton OTUs, the pyrosequencing based analysis of the ITS recovered 58,330 high-quality sequences across all samples with 3556–6003 high-quality sequences per sample. Based on 97 % species similarity, 1559 fungal OTUs were observed predominantly from three phyla (Ascomycota, Basidiomycota, and Zygomycota) and accounted for 69.11 % of the total fungal sequences obtained (Fig. S1). The 10 most abundant fungal OTUs accounting for 45.26 % of the total sequences (Table S2) were Gibberella moniliformis (17.06 %), Vascellum curtisii (8.41 %), Ceratocysti sparadoxa (4.96 %), F. oxysporum (3.93 %), Hypocrea virens (2.45 %), Aspergillus fumigatus (2.26 %), Uncultured fungus clone CT134 (2.14 %), Savoryella lignicola (1.53 %), Acremonium furcatum (1.30 %), and Trichoderma asperellum (1.23 %).

Bacterial and Fungal α-Diversity

Bacterial and fungal microbial richness and diversity are shown in Table 1. For bacteria, Chao1, ACE, and Shannon index showed no significant difference in soil samples of the four time-series vanilla fields. In contrast, for fungi, the values of Chao1, ACE, and Shannon index significantly increased after monoculture of vanilla for 6 years. In addition, rarefaction curves analysis of the mean sequences of the three replicates of each time-series field confirmed that the number of observed OTUs in fungal communities significantly increased with an increase years of vanilla momoculture, whereas the number of observed OTUs for bacteria was relatively stable (Fig. 2).

Fig. 2
figure 2

Rarefaction curves of soil bacterial and fungal communities at 97 % sequence similarity level in four time-series soil from vanilla fields. Error bars indicate standard deviation (n = 3)

Full size image

Bacterial and Fungal β-Diversity

Hierarchical clustering analysis based on the Bray-Curtis distance revealed that the bacterial and fungal communities in the soil samples collected from the same time-series field were more similar as the four highly supported clusters were grouped together (Fig. 3). Bacterial and fungal community structures from soil samples that were mono-cultured for 6 (b), 11 (c), and 21 (d) years clustered together but they were separated from the 1-year soil sample (a). In addition, for both the bacterial and fungal community composition, the 1-year soil sample (a) displayed the largest differences from the 21-year sample (d).

Fig. 3
figure 3

Bray-Curtis based hierarchical clustering of bacterial and fungal communities in soil from 4 time-series vanilla fields. a, b, c, and d represent field with 1, 6, 11, or 21 years of continuous cropping history, respectively

Full size image

UniFrac-weighted principal coordinate analysis (PCoA) based on the OTU composition also clearly demonstrated variations among these different soil samples, with the first two axes explaining 54.54 and 21.72 % of the total variation for the bacterial data and 55.73 and 21.80 % for fungi. In addition, both the bacterial and fungal communities in the 1-year soil sample (a) were obviously separated from the other nine samples by principle component 1, and furthermore, the 11- and 21-year soil samples had the most similar fungal community memberships (Fig. 4). The unweighted UniFrac algorithm displayed similar results, but for clarity, only the weighted UniFrac PCoA plot is shown here.

Fig. 4
figure 4

UniFrac-weighted PCoA plots of bacterial and fungal communities in four time-series vanilla fields. a, b, c, and d represent field with 1, 6, 11, or 21 years of continuous cropping history, respectively

Full size image

Venn diagrams were generated using Mothur based on the shared OTU tables from the four time-series field samples (Fig. S2). For bacteria, 624, 409, 267, and 304 unique OTUs were present in the 1-year (a), 6-year (b), 11-year (c), and 21-year (d) fields, respectively, whereas 127, 132, 179, and 186 unique OTUs were present for fungi, respectively. Moreover, 258 bacterial and 74 fungal OTUs accounting for 63.76 % of total bacterial and 69.69 % of total fungal sequences were shared across all the soil samples.

Pearson Correlation Coefficients Between the Abundances of Microbial Taxa and the Duration of Vanilla Monoculture (in Years) and Vanilla Stem Rot DI

As shown in Table 2, at phylum level, the relative abundances of the Bacteroidetes, Firmicutes, and Actinobacteria for bacteria and the Basidiomycota for fungi significantly (P < 0.05) decreased with the increase of continuous-cropping years, whereas the reverse was true for Nitrospirae variation. In addition, the abundances of Bacteroidetes, Firmicutes, and Actinobacteria revealed a significantly (P < 0.01) negative correlation with vanilla stem rot DI (Table 2). For fungi, a significantly (P < 0.05) negative correlation between DI and Basidiomycota abundance was observed.

Table 2 Pearson correlation coefficients between the abundance of microbial taxa and the duration of vanilla monoculture and vanilla stem rot disease index
Full size table

A large number of OTUs in bacterial and fungal communities are significantly correlated to the duration of vanilla monoculture or vanilla stem rot DI, and the vanilla wilt disease is caused by fungal pathogen. The 10 most abundant and typical fungal OTUs are shown in Table 2, and bacterial OTUs widely reported to possess plant growth-promoting or plant pathogen-suppressing function were also shown (the taxa with significance >0.05 are not shown). Results showed that for bacteria, Bacillus and Bradyrhizobium decreased significantly over the vanilla monoculture period, and for fungi, the abundance of F. oxysporum accumulated significantly over the monoculture span, and the T. asperellum abundance displayed a decreasing trend over time (Table 2). Moreover, the abundance of F. oxysporum, Aspergillus fumigates, and S. lignicola showed significantly positive correlation with DI. However, Bacillus and Bradyrhizobium abundances were negatively correlated with DI.

Pearson Correlation Coefficients Between Soil Properties and the Bacterial or Fungal Phyla

Soil pH showed significantly (P < 0.05) negative correlations with the relative abundances of Firmicutes and Actinobacteria but not with Acidobacteria. Bacteroidetes and Firmicutes abundances were positively correlated with soil EC. Soil OM content was negatively correlated with Proteobacteria abundance and positively correlated with Acidobacteria. For fungal phyla, only Basidiomycota was observed to have a significantly (P < 0.05) negative correlation with soil Ca. Ascomycota. However, Zygomycota had no significant correlation with all soil properties (Table S3).

Discussion

In the present study, pot experiments were performed to evaluate disease symptoms of vanilla in four time-series fields soil (Fig. 1), and the results were consistent with Zhao et al. [12], who found that vanilla stem rot disease index significantly increased with the increase of vanilla monoculture. This phenomenon has also been observed for many other plants, including potato [4], cucumber [22], and tomato [31], the growths of which were significantly hindered with an increasing incidence of soil-borne disease after monoculture.

The relative abundances of bacterial and fungal phyla reported in this study generally agree with previous pyrosequencing surveys of soil microbial communities. For bacteria, we observed similar results to Liu et al. [46], in which Acidobacteria and Proteobacteria were the top two abundant bacterial phyla in succession cropping soils (Fig. S1). For fungi, Ascomycota (according for 49.46 % RA) and Basidiomycota (according for 12.26 % RA) were the two main phyla (Fig. S1) and this was consistent with those of previous researches in which Ascomycota and Basidiomycota were the top two abundant phyla in continuous cropping peanut and soybean soils [20, 47]. Similar results have also been observed in forest soils and other soil types in nature [48, 49].

Drastic changes in bacterial and fungal community compositions in soil samples collected from four time-series vanilla fields were observed, and the potential correlations between their variations and vanilla Fusarium wilt disease were analyzed. For bacteria, at the phylum level, the abundances of Firmicutes and Actinobacteria significantly decreased with increasing years of vanilla monoculture (Table 2). This may be the main factor causing soil weakness after vanilla continuous cropping, as previous researches also reported that Firmicutes and Actinobacteria were less abundant in disease conducive soils than suppressive soils [50]. In addition, a positive correlation between Firmicutes and Actinobacteria levels and plant health was also observed in wheat-cropping region, in Canada’s oldest organic-conventional study field [51]. After amalgamation of fungal OTUs at phylum level, the relative abundance of Basidiomycota showed negative correlation with DI of vanilla stem rot (Table 2). In contrast, Xu et al. [30] found that no correlations were observed between the health status of the pea fields and the relative abundance of each fungal phylum. The difference was probably due to different plant species and soil types.

At the OTU level, the abundances of Bradyrhizobium and Bacillus decreased along continuous cropping years and significantly (P < 0.05) negatively correlated with vanilla stem rot DI (Table 2). Many previous studies demonstrate that Bacillus play important roles in vanilla health [52] and in the suppression of other soil-borne diseases [53] and Bradyrhizobium is important for the promotion of plant growth [54], indicating that the decrease of beneficial bacterial species may be the cause for soil weakness after vanilla long-term monoculture. For fungi, the abundance of T. asperellum displayed a decreasing trend over the vanilla monoculture span (Table 2), and this species had been demonstrated to be an efficient biocontrol agent against vanilla Fusarium wilt disease [52]. In contrast, the amount of the pathogen F. oxysporum, the most likely vanilla stem rot disease-causing pathogen [10], increased with increasing years of vanilla continuous cropping, similar to the results of Li et al. [20] in which accumulations of fungal pathogens were observed at the expense of plant beneficial fungi in consecutively cropped peanut soil. And, a significantly positive correlation of this pathogen with vanilla stem rot DI (Table 2) was observed. This result was in agreement with several previous studies in which soils subjected to the monoculture of crops such as banana and cucumber [55, 56], accumulated F. oxysporum, resulting in serious Fusarium wilt disease and yield losses. Another destructive vanilla pathogen, Phytophthora (a type of oomycete), affecting stems of V. planifolia in all vanilla growing regions of the world [57, 58], was found in very low abundance (data not shown). Similarly, oomycetes (fungus-like pathogens) were detected in low abundances in bulk soil of peanut and pea [20, 30]. A possible reason may be that Phytophthora is obligate parasite and thus predominates in roots, but not necessarily in bulk soil.

In this study, soil fungal diversity increased along the vanilla continuous cropping years (Fig. 2), and a similar result was also reported by Chen et al. [21] in which soil eukaryotic diversity increased with the years of peanut continuous cropping. Similarly, Liu et al. [46] observed that fungal diversity increased over a 7-year gradient of potato monoculture. All these facts verify that increased fungal diversity may be a common phenomenon in agro-ecosystems of monoculture.

Both hierarchical clustering and PCoA analyses indicated that vanilla monoculture span had the greatest effect on variation in fungal community structure (Figs. 3 and 4). This finding was consistent with those of previous researches that soil fungal community structure significantly changed along the years of succession cropping of peanut and cucumber [21, 22], based on 18S rRNA gene clone library analysis and PCR-denaturing gradient gel electrophoresis, respectively. Similarly, through 454 pyrosequencing, PCA analysis revealed distinctive differences in the composition and structure among the fungal communities of three different time-series peanut fields [20]. These results indicate that fungal communities affected by successive cropping could contribute to soil sickness associated with crops cultivation. In contrast, PCoA analyses showed that soil fungal community structure in 6-year fields were separated from the other three time-series fields (1-, 11-, and 21-year) by principle component 2 (Fig. 4). This observation can be attributed to the soil types, which was another important determinant of the composition of the microbial communities in arable soil [59]. Venn diagrams showed that the number of bacterial unique OTUs declined along the years of vanilla monoculture (Fig. S2). Interestingly, more bacterial unique OTUs numbers were found in healthy soil than in diseased soil in potato fields in Michigan, USA [60]. Thus, further research is needed to determine if the unique bacterial OTUs in vanilla 1-year fields have the potential functions for maintaining soil health.

In conclusion, long-term continuous cropping of vanilla significantly altered soil microbial community membership and structure with different responses in soil fungal and bacterial communities. Soil fungal diversity increased with vanilla monoculture time span, whereas no significant difference was observed for bacteria. Depletions of the Firmicutes, Actinobacteria, Bacteroidetes, and Basidiomycota phyla were observed along the years of vanilla monoculture. The abundance of F. oxysporum increased over the monoculture span and significantly positively correlated with vanilla Fusarium wilt disease. In general, soil weakness and vanilla stem wilt disease after long-term continuous cropping can be attributed to the alteration of the soil fungal community, the reduction of the beneficial microbes, and the accumulation of the soil borne pathogen Fusarium. spp.