Association of soil fungal community composition with incidence of Fusarium wilt of banana in Malaysia

ABSTRACT Banana (Musa spp.), an important food crop in many parts of the world, is threatened by a deadly wilt disease caused by Fusarium oxysporum f. sp. cubense Tropical Race 4 (TR4). Increasing evidence indicates that plant actively recruits beneficial microbes in the rhizosphere to suppress soil-borne pathogens. Hence, studies on the composition and diversity of the root-associated microbial communities are important for banana health. Research on beneficial microbial communities has focused on bacteria, although fungi can also influence soil-borne disease. Here, high-throughput sequencing targeting the fungal internal transcribed spacer (ITS) was employed to systematically characterize the difference in the soil fungal community associated with Fusarium wilt (FW) of banana. The community structure of fungi in the healthy and TR4-infected rhizospheres was significantly different compared with that of bulk soil within the same farm. The rhizosphere soils of infected plants exhibited higher richness and diversity compared with healthy plants, with significant abundance of Fusarium genus at 14%. In the healthy rhizosphere soil, Penicillium spp. were more abundant at 7% and positively correlated with magnesium. This study produced a detailed description of fungal community structure in healthy and TR4-infected banana soils in Malaysia and identified candidate biomarker taxa that may be associated with FW disease promotion and suppression. The findings also expand the global inventory of fungal communities associated with the components of asymptomatic and symptomatic banana plants infected by TR4.


INTRODUCTION
Fusarium wilt (FW) is the most destructive fungal disease affecting banana plantations across the globe (Dita et al. 2018). The disease is caused Fusarium oxysporum sp. cubense (Foc), a soil-borne fungus that penetrates through banana roots and dominates the vascular tissues, disrupting the dissemination of necessary nutrients from roots to the upper parts of the plant . Foc Tropical Race 4 (TR4) is considered the most virulent strain of Foc, causing epidemics in countries including China, Taiwan, Malaysia, Indonesia, and the Philippines (Jamil et al. 2019).
One of the important factors that influence the occurrence of soil-borne diseases is soil health, which is determined through multiple interactions of the chemical, physical, and biological processes (Larkin 2015). These processes are predominantly governed by the soil microbiome. The richness and diversity of soil microbiomes play a vital role in the development and maintenance of healthy soil, which protect the soil from microbial pathogens and, at the same time, enhance the soil and plant health conditions for better production (Berg et al. 2017;Pascale et al. 2020;Ray et al. 2020).
Among the microbiota present in the microbiome, soil fungi are recognized as the key functional component for soil and ecological systems. Their strong engagement with energy flow, nutrient cycle, and organic soil transformation reshape the microbial communities in the soil . High fungal richness in the rhizosphere soils has been associated with a higher incidence of FW disease in bananas and lower yield production (Kaushal et al. 2020;Shen et al. 2018;Zhou et al. 2019). In the soil suppressive to FW disease, Penicillium, Aspergillus, and Trichoderma were prevalent as beneficial fungi (Bidellaoui et al. 2019;Liu et al. 2019;Miao et al. 2019;Zhou et al. 2019). Through enzymatic antifungal compounds, these fungi have the potential to prevent pathogen invasion, indirectly acting as antifungal biological control agents (Wang et al. 2022;Win et al. 2021). Furthermore, unique fungal taxa can be used as markers to assess soil health.
In addition to intimate relationship of plant with microbes, soil physicochemical properties also affect plant health and influence soil microbial communities (Jamil et al. 2022;Sui et al. 2021;Wang et al. 2022;Zhao et al. 2018). For example, the microbial composition of the rhizosphere soil of bananas is affected by soil particle, soil pH, and potassium and phosphorus availability in the soil, where a high level of the soil properties have been shown to improve the suppression ability against banana FW disease (Shi et al. 2017;Wang et al. 2022;Wu et al. 2020;Zhou et al. 2019). Consequently, the composition of a soil microbiome is influenced by soil physicochemical properties and biotic parameters, and changes in these factors may result in disparities in plant health condition and agricultural yield (Santoyo et al. 2017;Tkacz and Poole 2015;Zheng et al. 2018). In the recent report by Hu et al. (2021), magnesium was proposed as the key nutrient mechanism related to FW resistance, hinting at the importance of Mg reservoir in the soil for healthy bananas.
Profiling fungal community structure and analyzing their richness and diversity in a microhabitat is important, since microbial populations can be preferentially localized in microhabitats and exhibit a high spatial heterogeneity (Hermans et al. 2019). Hence, the present study was conducted to analyze the composition of fungal community in healthy and FW-infected soils from a banana farm using the Illumina sequencing platform. We hypothesized that the diversity and composition of soil fungal communities between healthy and FW-infected soils are different. In addition to common taxa, we also hypothesized that rare operational taxonomic units (OTUs) are abundant in the banana farm due to their low local abundance and habitat specificity. Albeit rare, these OTUs can have a disproportionately large influence on ecosystem processes and can serve as potential biomarkers for the healthy soil against TR4 in the future.

MATERIALS AND METHODS
Soil sampling and DNA extraction.-Soil samples were obtained from a banana plantation in Selangor, Malaysia (3° 48ʹ10.1ʹʹN, 100°50ʹ42.5ʹʹE), planted with Musa acuminata cv. Berangan, as described previously (Jamil et al. 2022). Briefly, banana plants showing FW symptoms (splitting of pseudostem, skirting of wilting leaves, and leaf streaking) and nonsymptomatic plants from the same farm were investigated. The presence of TR4 as the causal agent of the FW disease in the symptomatic bananas was confirmed by molecular techniques (Jamil et al. 2022). Samples of rhizosphere and bulk soil were obtained using a completely randomized design from five symptomatic and nonsymptomatic plants. Soils that were strongly attached to the banana roots were collected as rhizosphere soils, whereas bulk soils were obtained from the loose soils surrounding the root region. Bulk soil was ground in a sterile mortar and pestle and sieved through a 2-mm sieve before being transferred to a microcentrifuge tube. Total soil DNA was extracted using DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol. The quantity and quality of the extracted DNA were verified using a Qubit 2.0 fluorometer (Thermo Scientific, Massachusetts, USA), and the DNA integrity was determined by running 1% (w/v) Tris base, acetic acid and EDTA (TAE) agarose gel electrophoresis. The DNA was stored at −80 C before being sent to NovogeneAIT Genomics Singapore (Biopolis, Singapore) for internal transcribed spacer (ITS) sequencing. Sequencing libraries were generated using NEBNext Ultra DNA Library Pre Kit for Illumina (California, USA) following the manufacturer's recommendations. The library quality was assessed on the Qubit 2.0 fluorometer (Thermo Scientific) and sequenced using the Illumina NovaSeq, generating 250-bp paired-end reads. The sequences were processed as described in Jamil et al. (2022). Representative sequence for each OTU was screened for further annotation. Sequences analysis were performed by BLAST with QIIME (1.7.0, http://qiime.org/scripts/assign_taxonomy. html) (Altschul et al. 1990) and UNITE database (https:// unite.ut.ee/) (Kõljalg et al. 2013) for species annotation at each taxonomic rank (kingdom, phylum, class, order, family, genus, species). To obtain the phylogenetic relationship of all OTU representative sequences, MUSCLE (Edgar and Edgar 2004) (3.8.31, http://www.drive5.com/ muscle/) can compare multiple sequences rapidly. OTU abundance information was normalized using a standard of sequence number corresponding to the sample with the least sequences. Subsequent analyses of alpha diversity and beta diversity were all performed basing on this output normalized data.

Bioinformatic analysis and statistical method.-
The bioinformatic analyses were conducted using MicrobiomeAnalyst as described in Jamil et al. (2022). Using a linear discriminant analysis effect size (LEfSe), a heatmap of biomarker taxa was created based on Euclidean distance and the Ward linkage method. Taxa were judged significant if their adjusted P-value cutoff was less than 0.05, and only taxa with a linear discriminant analysis (LDA) score greater than 2 were visualized. The R package PHYLOSEQ (McMurdie et al. 2013) was utilized to plot alpha and beta diversity values. Canonical correspondence analysis (CCA) was conducted using R package VEGAN 2.5-3 (Oksanen et al. 2018) to examine relationships among soil samples, significant soil physicochemical properties, and selected biomarker taxa. The top three taxa were selected for fungi for the CCA.

RESULTS
Analysis of microbiome data.-Fungal communities associated with the rhizosphere and bulk soil of infected and healthy banana plants were characterized based on the fungal internal transcribed spacer (ITS). A total of 2 991 397 reads were generated following quality filtering and chimeric sequence removal from 20 soil samples, ranging from 140 674 to 159 764 for each sample data set (SUPPLEMENTARY TABLE 1). The length distribution of trimmed sequences ranged from 154 to 438 bp. All of the 20 samples were rarefied to the minimum number of sequences, and they were clustered into 2410 distinct fungal operational taxonomic units (OTUs), representing a mean Good's coverage of 0.994. The rarefaction curves of all four groups of samples (infected rhizosphere, RI; infected bulk soil, BI; healthy rhizosphere, RH; and healthy bulk soil, BH) were near saturation, indicating sufficient sequencing depth to cover the fungal diversity within individual samples ( SUPPLEMENTARY FIG. 1A). The numbers of OTUs were similar between the rhizosphere and bulk soils (SUPPLEMENTARY FIG. 1B) but higher in infected soils (BI and RI) compared with healthy soils.

Composition of fungal richness and diversity.-
The fungal sequences of healthy and infected rhizosphere and bulk soils were assigned at genus level, with Fusarium, Penicillum, Tricholoma, Trichoderma, Talaromyces, Myrothecium, Cordana, Cortinarius, Tricothecium, and Acremonium identified as the major genera associated with all soils (FIG. 1a). The dominant genus in all soil samples was Fusarium (3-13%), followed by Penicillum (3-7%), whereas the relative abundance of other genera was ≤1%. The highest percentage belongs to unassigned sequences. Across the treatments, Fusarium was found significantly higher in RI at 13% (t-test, P < 0.05), compared with RH at 4% (FIG. 1b). In bulk soil, Fusarium was also significantly higher in BI at 5% (t-test, P < 0.05) than in BH at 3% (FIG. 1c).
Fungal communities were also evaluated using richness and diversity indices (FIG. 2). Unfortunately, there is no significant difference in ACE (abundance-based coverage estimators), Chao1, Observed, and Simpson indices between bulk soil and rhizosphere soil (data not shown), except for the Shannon index where the diversity of RI was significantly higher (P < 0.05).
To evaluate the differences in fungal composition of all infected and healthy soils in the banana farm, principal coordinate analysis (PCoA) was performed based on Bray-Curtis distances. Only the fungal communities from the rhizosphere soils showed significant clustering according to soil type (FIG. 3). Axis 1 and Axis 2 explained 56.8% of variances between RH and RI (FIG. 3). Assessment using analysis of similarity (ANOSIM) supported the PCoA analysis (R = 0.584; permutational multivariate analysis of variance [PERMANOVA], P < 0.05) (SUPPLEMENTARY TABLE 2). No significant clustering of fungal communities was observed between all soil samples (SUPPLEMENTARY FIG. 2A) and bulk soils (SUPPLEMENTARY FIG. 2B).

Identification of potential biomarkers for healthy
and FW-infected rhizosphere soils.-Fungal groups that were significantly different between RI and RH were distinguished by employing a linear discriminant analysis effect size (LEfSe) with linear discriminant analysis (LDA) effect size greater than 2. The analysis resulted in a total of 24 fungal taxa, and the majority of them (20 taxa) were more abundant in RI (FIG. 4A). The top three fungal taxa identified in RI were Trichothecium ovalisporum, Nectriaceae, and Sarocladium strictum.
Meanwhile, the top three fungal taxa in RH were Hydnodontaceae, Penicillium, and Sordariomycetes. Unfortunately, Fusarium genus was not detected as one of the discriminating taxa for the analyzed soils, even though it was highly abundant in RI (FIG. 1b). Heatmap and the hierarchical clustering in FIG. 4B showed a well-clustered community in RH and RI, which supported the result shown in FIG. 4A. Relationships between soil physicochemical properties and potential biomarkers.-In a previous report (Jamil et al. 2022), cation exchange capacity (CEC) and magnesium (Mg) were identified as significant properties of the healthy soils in the banana farm.
To determine the correlation of Mg and CEC with the selected abundant fungal species from the LEfSe analysis (Trichothecium spp., Sarocladium spp., Phoma spp., Cordana spp., and Penicillium spp.), canonical correspondence analysis (CCA) was performed. FIG. 5 shows that CCA components explained 38.15% of the total fungal genus variation. There was no correlation between CEC and the analyzed fungal taxa. However, Mg exhibited a positive correlation with Penicillium sp.

DISCUSSION
Plants are associated with a large variety of microbes in the endosphere and episphere of the plant tissues that contribute to their overall fitness (Dastogeer et al. 2020). The community composition and diversity of plant microbiomes are influenced by various biotic and abiotic factors related to plant, microbes, as well as environment. The important role of fungal biodiversity and Figure 1. Relative abundance of fungal genera in the rhizosphere (RH and RI) and bulk (BH and BI) soils. a. A color-coded bar plot shows the composition of fungal genera in different soil groups. b. Significant difference between the fungal genera in RH and RI analyzed using t-test. c. Significant difference between the fungal genera in BH and BI analyzed using t-test. The y-axis represents the classification at genus level. The x-axis represents the mean percentage values in groups. The blue and orange columns represent the average results in the healthy and infected soils, respectively. The color of the circle is in agreement with the group whose mean value is higher. The right-most value is the P-value of the significance test of between-group variations.
their functions in soil health has been well described (Frąc et al. 2018). In this study, we investigated the difference in soil fungal community related to FW infection caused by TR4 in a banana farm planted with the susceptible variety cv. Berangan.
We demonstrated that the infected soils have higher abundance of fungal communities, which is significantly dominated by Fusarium, compared with the healthy soils. The Shannon index, which is a measure of diversity, was much higher in RI than in RH, indicating that the species structures were dissimilar in the rhizospheres influenced by soil health status. Our finding of the relationships between fungal diversity, especially in the rhizosphere, and FW disease incidence agrees with previous reports (Kaushal et al. 2020;Shang et al. 2016;Shen et al. 2018;Shi et al. 2017;Zhou et al. 2019). We propose that the consecutive Figure 2. Alpha diversity index, Shannon index, at the OTU level represented as a boxplot. A boxplot represents the diversity distribution of a group present within a type class conducted using unrarefied data, and pairwise comparison was performed using t-test. Significant difference was accepted when P < 0.05 between the two groups. *Denotes P < 0.05.  monoculture of banana may drive the abundance of fungal diversity, especially Fusarium, thus leading to the increment of FW disease incidence in the farm (Xiong et al. 2016).
In addition, the high pathogen pressure observed in monoculture could be due to low species richness and maximum relative abundance of the host (Mommer et al. 2018). Another theory is that higher fungal OTUs in the diseased soils may be due to an increase of Foc colonization, which might provoke the changes of microbial species and distributions (Zhou et al. 2019). In a study that analyzed bacterial and fungal communities in the soil samples collected from FW-infected banana fields caused by Foc Tropical Race 4 in China, fungal communities were reported to respond more obviously and quickly to a continuous monocropping in the same field than bacterial communities (Shen et al. 2018). The study also showed that FW disease incidence is correlated with a higher fungal richness.
However, our findings on the fungal diversity contradict other studies that reported significant positive correlations between fungal richness and banana FW disease incidence (Shen et al. 2018;Wang et al. 2022). This incongruence could be explained by the study locations that did not share the same climatic conditions and agronomic management. Therefore, the soils are expected to harbor different microbial communities that can be reflected by the fungal communities in a specific microhabitat. In addition, there could also be the effects of genotype. In this study, the farm was planted with a local banana cultivar (cv. Berangan), as opposed to the Cavendish subgroups that are mostly reported in the microbiome studies involving Fusarium wilt of bananas.
Consistent with the analysis of alpha diversity via the Shannon index, the analysis of beta diversity also showed a separation between rhizosphere soils (RH and RI) (ANOSIM, R = 0.584, P < 0.05), suggesting that the compositions of fungal communities in bulk soils and among all soil groups are similar. The rhizosphere zones are nutrient-rich and a hotspot for microbial populations that can change dramatically when pathogens invade (Praeg et al. 2019;Zhang et al. 2017). Extensive microbe-microbe and plant-microbe communications that occur in the rhizosphere niche combined with changes in climate, root exudates, soil types, plant genotype, and developmental phases of a plant all have an impact on the microbial communities (Enebe and Babalola 2019;Pascale et al. 2020). It also explains the apparent variation in the composition of the community of fungi in the healthy and infected rhizospheres compared with bulk soil (Qu et al. 2020;Sasse et al. 2018;Tao et al. 2020).
We also attempted to decipher the keystone or biomarker taxa in the rhizospheric fungal communities with strong correlation with the FW disease. Identification of biomarker taxa is critical for harnessing the plant microbiome to enhance plant growth and health. The LEfSe analysis identified substantially distinct fungal taxa between RI and RH. Hydnodontaceae and Penicillium were enriched in healthy soils, suggesting that their presence could antagonize the pathogenic Fusarium sp. and suppress FW in bananas. This is the first report that identifies Hydnodontaceae as one of the biomarker taxa in association with the healthy soil of bananas. Previously, there are several reports that described the beneficial role of Hydnodontaceae against Panax notoginseng root-rot disease that is caused by Fusarium oxysporum and Phoma sp. Wei et al. 2020). Hydnodontaceae is a prolific producer of enzymes that break down lignin and wood (Sharma-Poudyal et al. 2017). The breakdown of organic matter in the soil regulates the balance of carbon and nutrients, making more nutrients available for plant growth (Frąc et al. 2018). Meanwhile, the roles of Penicillium as beneficial microbes and pathogen defense have been described in recent studies where the Penicillium genus acted as a biocontrol agent to reduce or suppress the development of pathogens through the production of antifungal proteins (AFPs) (Garrigues et al. 2017) and enzymes (Puig and Cumagun 2019). Penicillium sp. isolated from FW-free banana soil was shown to produce enzymatic antifungal compounds (chitinase and β-1,3-glucanase) that could inhibit pathogenic Foc in greenhouse conditions (Win et al. 2021). Many other soil-borne plant diseases caused by pathogenic fungi have been successfully controlled by Penicillium (Garrigues et al. 2018(Garrigues et al. , 2017Puig and Cumagun 2019;Raza et al. 2008;Win et al. 2021). Thus, it can be considered a promising biomarker candidate for healthy banana soil.
In the infected soils, Trichothecium ovalisporum, Nectriaceae, and Sarocladium strictum were enriched. Fungal species belonging to Trichotecium sp., Sarocladium sp., Phoma sp., and Cordana sp. had been discovered in all types of banana soil in northern Taiwan together with various Fusarium sp. (Chen 2017). All the identified species are mostly pathogenic, as they were highly abundant in infected soil of various types of disease, including FW of bananas. (Dal Bello 2008;Ji et al. 2021;Lan et al. 2017;Molina and Williams 2010;Yang et al. 2013;Zhao et al. 2016). Nectriaceae, on the other hand, were found to be preferential colonizers of rhizosphere, root, and corm in samples collected from Tanzania, East Africa (Kaushal et al. 2020).
Despite the great abundance of Fusarium, the genus was not found as a biomarker in RI. The genus is known to contain a wide range of pathogenic and nonpathogenic fungi, which are likely difficult to distinguish in the soil studied in this study. As previously reported by Yongmei et al. (2020), the Fusarium genus was not found to be an abundant genus or species in both healthy and FW-infected rhizosphere soils. Therefore, we cannot depend on Fusarium alone to detect whether the soil is likely to cause FW infection, considering the pathogenic and nonpathogenic fungi of the Fusarium genus.
Changes in soil characteristics can be linked to variations in the fungal community composition of healthy and diseased soils. Previously, we have identified Mg and CEC as the soil properties associated with the healthy soils in the farm (Jamil et al. 2022). In this study, Mg was found positively correlated with the abundance of Penicillium sp. in the healthy soil. A study conducted decades ago revealed that high Mg availability in the soil encouraged the growth of Penicillium sp. and, at the same time, Penicillium sp. depended on Mg to produce penicillin (Jarvis and Johnson 1950). Penicillium sp. was also found in the magnesium-rich soil in the Oregon forest (Crawford et al. 2000). Fouda et al. (2021) reported that the production of Mg was associated with metabolites secreted by Penicillium sp., which could control the disease caused by microbes. However, we could not find any study that explained the relationship between Mg and Penicillium sp. concerning FW disease in bananas.

CONCLUSIONS
Our study revealed that the composition and diversity of fungal communities differ between healthy and FWinfected rhizosphere soils of banana farm in Malaysia, supporting the role of the rhizosphere as the key influencer of plant health. We detected more fungal communities in the infected soil, which is dominated by Fusarium.
Hydnodontaceae, Penicillium, and Sordariomycetes were significantly enriched in healthy soil, which could be promising biomarker candidates. Among the highly abundant fungal taxa, Penicillium was found to be positively correlated with Mg, one of the important soil properties associated with FW disease incidence. These findings add to the pool of knowledge on banana microbiome. However, there are still plenty of gaps in the underlying mechanisms of microbiome assemblages and how they influence the host plants. Hence, further work is needed to fill the knowledge gaps and to connect the microbiome composition and diversity to their function before we could use and manipulate the microbiomes for sustainable agricultural production.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author(s).