Microbial community overlap between the phyllosphere and rhizosphere of three plants from Yongxing Island, South China Sea

Abstract Phyllosphere and rhizosphere are unique and wide‐ranging habitats that harbor various microbial communities, which influence plant growth and health, and the productivity of the ecosystems. In this study, we characterized the shared microbiome of the phyllosphere and rhizosphere among three plants (Ipomoea pes‐caprae, Wedelia chinensis, and Cocos nucifera), to obtain an insight into the relationships between bacteria (including diazotrophic bacteria) and fungi, present on these host plants. Quantitative PCR showed that the abundances of the microbiome in the soil samples were significantly higher than those in the phyllosphere samples, though there was an extremely low abundance of fungi in bulk soil. High‐throughput sequencing showed that the alpha‐diversity of bacteria and fungi was higher in the rhizosphere than the phyllosphere samples associated with the same plant, while there was no obvious shift in the alpha‐diversity of diazotrophic communities between all the tested phyllosphere and soil samples. Results of the microbial composition showed that sample‐specific bacteria and fungi were found among the phyllosphere and rhizosphere of the different host plants. About 10%–27% of bacteria, including diazotrophs, and fungi overlapped between the phyllosphere and the rhizosphere of these host plants. No significant difference in microbial community structure was found among the tested rhizosphere samples, and soil properties had a higher influence on the soil microbial community structures than the host plant species.


| INTRODUC TI ON
The leaf surface, also known as the phyllosphere, is a large and extremely diverse habitat for terrestrial microorganisms. The estimated total leaf surface area in the word is approximately twice as large as the land surface area (Zimmerman & Vitousek, 2012). Many species of bacteria and fungi colonize leaf surfaces (where they mostly form aggregates) and the spaces inside the leaves (Vorholt, 2012).
The analysis of microbial community composition in the rice revealed clear differences, both in terms of composition and complexity, between the rhizosphere and the phyllosphere, although some genera (like the Methylobacterium) were shared between the two (Knief et al., 2012). A recent study of the common grapevine (Vitis vinifera) microbiota showed that the root-associated bacterial communities differed significantly from the aboveground communities, yet the microbiota of the leaves, flowers, and grapes shared a greater proportion of taxa with the soil communities than with each other. This suggests that soil may serve as a common bacterial reservoir for belowground and aboveground plant microbiota (Zarraonaindia et al., 2015). However, little is known about the overlapping microbial communities (both bacteria and fungi) between the phyllosphere and rhizosphere that play an important role in the ecology and evolution, and the promotion of plant growth and health.
Yongxing Island (Sansha City, Hainan Province, China) is a tropical island and the biggest coral island in the South China Sea. In our previous study, we described the distinct microbial communities of the phyllosphere associated with five plants on the island (Bao et al., 2019). The study showed that bacterial communities of the tropical forest soils are exceedingly distinct from those found in other ecosystem types (Delgado-Baquerizo et al., 2018). From a functional viewpoint, there is a priori evidence that bacteria in tropical ecosystems may be more important than those in cooler areas, as the rates of nitrogen fixation estimated in a tropical forest, in all system components (soil and vegetation), are commonly thought to be among the highest of any natural ecosystem (Cleveland et al., 1999). There is also evidence to suggest that N 2 fixation in the phyllosphere is the main mechanism for the addition of N in humid tropical ecosystems (Abril, Torres, & Bucher, 2005). The establishment of phyllospheric populations of diazotrophs has mostly been reported in several tropical plants (Fuernkranz et al., 2008;Goosem & Lamb, 1986). The abundance and composition of shared taxa between the phyllosphere and rhizosphere may differ from those in temperate ecosystems, but that is just an analogy. There is, however, mounting evidence that targeted manipulation of microorganisms can lead to more environmentally and economically sustainable production systems. To provide a certain theoretical basis for the above hypothesis and the construction of sustainable production systems, we collected the phyllosphere, rhizosphere, and bulk soil from three different host plant species (Ipomoea pes-caprae, Wedelia chinensis, and Cocos nucifera) based on our previous study. We used high-throughput amplicon sequencing to compare the microbial community composition, structure, and interaction, especially the shared microbes between the phyllosphere and rhizosphere, among the different host plant species in this tropical island.

| Site description and sample preparation
In December 2017, we chose three common and abundant plant species, Ipomoea pes-caprae (IP), Wedelia chinensis (WC), and Cocos nucifera (CN), growing on the Yongxing Island, in the South China Sea (Hainan Ocean Administration, 1999). This island has a true tropical maritime monsoon climate, with an average annual precipitation of more than 1,300 mm. The rainy season begins in late May and ends in early November, and the dry season lasts from late November to early May of the following year. The mean annual temperature of the island was 26-27°C, with the highest temperatures occurring in May and June. The sampling area for the three collected plant species was within 500 m 2 , to avoid a possible effect of the environmental factors, like spatial distance. Twelve individual plants from each species were randomly chosen in this area. The sampled IP and WC were not covered by any plants, including the CN.
Rhizosphere soils collected by shaking off the soil attached to the roots, and mature stage leaves were collected for the phyllosphere samples. Bulk soil, or the surface soil that is not penetrated by roots, was also collected. The detailed sampling process of the phyllosphere samples was introduced in our previous study (Bao et al., 2019).
Considering the heterogeneity of the tested leaves and soils, leaves and rhizosphere soil samples were collected from at least four plants, to form a composite sample from a 1.5 m × 1.5 m area. Three composite samples (more than 10 m apart) of the phyllosphere and rhizosphere were collected from each plant. Due to the wide space between CN trees (>5 m), the collecting area of CN was much bigger. At least four bulk soil samples were randomly selected from a 1.5 m × 1.5 m area and mixed to form one composite sample. Then, three composite samples were taken from each sampling site with more than 10 m distance.
We collected all the phyllosphere, rhizosphere, and bulk soil samples as quickly as possible. And there was no rain in the week before the sampling. We assumed that micrometeorological conditions were similar across sites. To standardize conditions as much as possible, we only chose green, healthy-looking, and intact leaves. Leaves from each plant were cut with a pair of sterilized scissors, and each leaf or soil sample was put in a Labplas bag, placed on ice, and quickly transported to the laboratory. Subsequently, each phyllosphere sample was immediately processed for DNA extraction, and each soil sample was divided into two parts: (a) One part was sieved through a 2.0 mm mesh and stored at 4°C for soil properties analysis, (b) and the second part was stored at −80°C for DNA extraction and molecular analysis.

| DNA extraction and determination of soil properties
Thirty grams (more than five pieces) of leaves were placed in a 1,000-ml sterile Erlenmeyer flask and filled with 500 ml sterile PBS buffer (pH 7.4, 1 × phosphate-buffered saline buffer). Sonication, at 40 kHz frequency for 6 min was then performed in an ultrasonic cleaning bath to wash the microbial cells off the leaves, followed by shaking at 200 rpm for 20 min at 30°C. After shaking, sonication was continued for a further 3 min. To separate the microbial cells from the leaves, the cell suspensions were then filtered through a 0.22 µm × 50 mm sterile nylon membrane. Phyllospheric DNA was directly extracted from each of the collected membranes. Soil DNA was then extracted from 0.5 g of the fresh soil using the Fast®DNA SPIN Kit (MP Biomedicals) and was stored at −80°C.
Soil pH was measured using a pH meter (PB-10, Sartorious) with a water-to-soil ratio of 2. Ammonium (NH 4 + -N) and nitrate (NO 3 − -N) were extracted from the soil samples with 2 mol/L KCl and determined using a Continuous Flow Analyzer (AA3, SEAL Analytical).
Total carbon (TC) and total nitrogen (TN) were determined using the elemental analyzer (rapid cube, Elementar). Dissolved organic carbon (DOC) was determined using a Total Organic Carbon Analyzer (Vario TOC, Elementar). Soil moisture represented the quantity of water in the soil and was measured by a drying method. Soil properties were described in Table 1.

| Quantitative PCR analysis
The quantitative PCR (qPCR) thermal profiling of the fungal ITS region and bacterial 16S rRNA genes was performed using primers ITS1/ITS2 and 799F/1115R, respectively. Primer sets PolF/PolR were used to amplify a region of the nifH genes that is the DNA barcode marker for the molecular identification of diazotrophic bacteria (Poly, Monrozier, & Bally, 2001;

| High-throughput sequencing and bioinformatics analysis
The 16S rRNA genes, ITS region, and nifH genes of the total DNA were sequenced using the Illumina paired-end approach. PCR was Sequencing data analyses were performed using the free online platform of the Majorbio I-Sanger Cloud Platform (www.i-sanger. com). Raw sequence data with fastq files were quality-filtered using Trimmomatic (Bolger, Lohse, & Usadel, 2014) and merged using FLASH (Magoc & Salzberg, 2011) with the following criteria. (a) Reads were truncated at any site receiving an average quality value below 20 over a 50-bp sliding window. (b) Sequences whose overlap being longer than 10bp were merged according to their overlap with mismatch no more than 2bp. (c) Sequences for each sample were separated according to barcodes (exactly matching) and primers (allowing 2 nucleotide mismatching), and reads containing ambiguous bases were removed. Thereafter, operational taxonomic units All six correlation networks were constructed using the Maslov-Sneppen procedure (Maslov & Sneppen, 2002;Wang et al., 2019) and visualized using Cytoscape 3.5.1. The statistical significance level for all the analyses was set at 0.05.

| Abundances and diversities of bacteria, fungi, and diazotrophs between leaf and soil samples
The abundances of bacteria, fungi, and diazotrophs among the phyllosphere, rhizosphere, and bulk soils of Ipomoea pes-caprae (IP), Wedelia chinensis (WC), and Cocos nucifera (CN), detected by performing qPCR assays, were found to be different ( Figure 1).
The qPCR data showed that for three genes the copy numbers were mostly higher among the soil samples, than the phyllosphere samples (except for an extremely low abundance of copy numbers for the ITS region in the bulk soil). The abundance of diazotrophs, measured with the nifH genes, was lower than fungi and bacteria measured in the same habitat. Among the phyllosphere samples, CN

F I G U R E 1
The copy numbers of the 16S rRNA genes, ITS region, and nifH genes on the phyllosphere, rhizosphere, and bulk soils. Different lowercase letters above the columns indicate significant differences among all the samples at p < .05; IP indicates Ipomoea pescaprae, WC indicates Wedelia chinensis, CN indicates Cocos nucifera, R indicates rhizosphere soil, and Bulk indicates bulk soil harbored the highest ITS region copies, and the smallest 16S rRNA and nifH gene copies, especially when comparing it with the phyllosphere IP and WC, described in detail in our previous study (Bao et al., 2019). Among the rhizosphere samples, WC-R and CN-R had a higher abundance of diazotrophs, bacteria, and fungi compared to IP-R (p < .05). Due to the very low abundances of fungi determined in all bulk soil replicates through qPCR assays and high-throughput sequencing analysis, no results for fungi in the bulk soil samples are shown below.
To estimate sampling completeness, Good's coverage was determined (Rea et al., 2011). The high index of Good's coverage at a 97% similarity level of bacteria, fungi, and diazotrophs indicated that the sequencing depth contained almost all bacterial, fungal, and diazotrophic communities in all samples. The alpha-diversities (Shannon index, Chao1, and Heip's evenness index) of bacteria, fungi, and diazotrophs were calculated with 22,264, 38,451, and 8,394 rarefied sequences per sample, respectively (Gotelli & Colwell, 2011;Heip, 1974; Table 2). Results showed that the rhizosphere and phyllosphere bacterial Shannon index, Chao1, and Heip's evenness index of IP were lower than for the other two plant species, with some of these indices having significant differences.
Among the phyllosphere fungal communities, the Shannon index and Heip's evenness index of IP were significantly higher than that of WC and CN (Bao et al., 2019), while the fungal Shannon index and Heip's evenness index of the CN-R rhizosphere were significantly higher than that of IP-R and WC-R. These three bacterial and fungal indices were mostly lower in the phyllosphere samples, compared to soil samples, although the Shannon index and Chao1 of fungi in rhizosphere IP-R were slightly lower than that of phyllosphere IP.
These results revealed that the sampled soils harbored a higher diversity, richness, and evenness of bacteria and fungi than the tested phyllosphere samples. Also, correlation analysis indicated that the rhizosphere Shannon index and Chao1 of the bacterial and fungal communities were negatively correlated with moisture and TC content, and positively correlated with soil NH 4 + -N content (Spearman's correlation test, Table A2). The rhizosphere's Heip's evenness index of bacterial communities was positively correlated with the soil's NO 3 − -N, DOC, and TN content, and negatively with the soil pH (Spearman's correlation test, Table A2).
The phyllosphere and rhizosphere diazotrophic Shannon index and Chao1 of IP were significantly lower than that of the other two plant species. However, there was no obvious shift in these three indices of diazotrophic communities between all the tested phyllosphere and soil samples (

| Differences in the community structure between phyllosphere, rhizosphere, and bulk soil samples
To compare the similarities and differences among the test sam- Therefore, the overlapping microbial community composition between the phyllosphere and rhizosphere of the same host plant became the focus of the next analysis in this study.

| Ubiquity and bacterial dominance between the phyllosphere and the rhizosphere
A total of 39 bacterial phyla and 961 bacterial genera were detected in the sampled phyllosphere, rhizosphere, and bulk soil samples.

Bacteroidetes, Firmicutes, Deltaproteobacteria, and
Betaproteobacteria were the dominant taxa in the phyllosphere, rhizosphere, and bulk soil samples, the sum of which was more than 90% of the total bacteria. The relative abundances of the major bacterial phyla in the test samples varied greatly between the phyllosphere, rhizosphere, and bulk soil samples ( Figure A1). in the phyllosphere was lower than that in the rhizosphere; however, the number of unique bacteria in the phyllosphere was higher than that in the rhizosphere.

| Shared bacterial and fungal communities between the phyllosphere and rhizosphere
Comparing the bacterial community composition of the phyllo-

F I G U R E 5 Fungal ITS region generabased correlation network for IP (a), WC (b), and CN (c). A node represents a genus.
A connection stands for a strong (Spearman's q > 0.97 or q < −0.97) and significant (p < .01) correlation. Edge widths were scaled according to their weights, and edge colors indicated a positive (red) or negative (gray) correlation for the nodes they connect while Gaiella, Rubrobacter, Bacillus, and Nitrospira were the dominant genera in rhizosphere CN-R (Figure 4b).
Comparing the fungal community composition of the phyllo-

| Correlations between the environmental factors and the bacterial, fungal, and diazotrophic community structures
The Mantel test, based on the Bray-Curtis method, was used to examine the effect of environmental factors on the microbial community. Soil pH, moisture, NH 4 + -N, DOC, TC, and TN content were positively correlated with the bacterial community in the rhizosphere samples. Among these correlations, the most significant was with the DOC content (Table 3, R = .9952, p < .05). The correlation between soil pH, NH 4 + -N, DOC, and TN content and the rhizosphere fungal community structure was also found to be significant, and NH 4 + -N had a higher correlation with the fungal community in the rhizosphere samples (

| D ISCUSS I ON
Previous studies have already demonstrated the composition of microbial communities among the phyllosphere and rhizosphere in different plants, but few studies investigated the shared microbiome between these two habitats. This question remains important, as the shared microbiome seems to be responsible for the connection between soil and plants and could play an important role in plant growth and health. Knief et al. (2012) observed the presence of the one-carbon conversion processes in the rhizosphere, as well as in the phyllosphere. Chen et al. (2017) found 10 antibiotic resistance genes (ARGs) in the soil that also found their way onto the phyllosphere, giving reasons for possible concern. In this study, we systematically characterized the phyllosphere and rhizosphere microbiome of three different tropical plant species, growing on Yongxing Island in the South China Sea. We also revealed an interaction between plants and microorganisms, including bacteria, fungi, and nitrogenfixing bacteria.
Overall, the abundances and diversities of the microbiome in the soil samples (including rhizosphere and bulk soil samples) were higher than those in the phyllosphere, which was consistent with a previous study (Kim et al., 2012). The partial reasons behind this phenomenon may be explained by the following hypothesis. The phyllosphere is a relatively harsh habitat, characterized by rapid changes in water and nutrient availability, UV radiation intensity, and other environmental stresses (Beattie & Lindow, 1999). Roots are known to have a certain influence on the community structure (Dennis, Miller, & Hirsch, 2010;Shu, Pablo, Jun, & Danfeng, 2012), evidenced by a significant shift of microbial community structures between the rhizosphere and bulk soil samples, while no significant differences in the community structure were found in all rhizosphere samples of three different plant species. These results indicated that the soil properties had a greater influence than host plant species on soil community structure (including rhizosphere and bulk soil samples). Many studies found that soil dissolved carbon contents and N-related soil properties were important factors in shaping the soil microbial community structure (Sun et al., 2016;Wang et al., 2017;Zeng et al., 2016). In this study, we also found that the microbial community structure in the rhizosphere and bulk soil samples was significantly affected by soil properties.
"Everything is everywhere, but the environment selects" is famously formulated in the Baas Becking hypothesis (de Wit & Bouvier, 2006), which is a good explanation for the different soil environments (different physical and chemical properties) affecting the distribution of microorganisms in our study. Our results also showed that the phyllosphere and rhizosphere samples of three plants had their unique microorganisms, which might be due to the different environmental factors, such as different oxygen concentration, temperature, between the leaf surfaces and rhizosphere soils, and strong UV radiation on the leaf surface (Beattie & Lindow, 1999). Moreover, precipitation events are considered to be one of the main abiotic factors that promoted the vertical migration of microorganisms throughout different habitats (Van Stan II et al., 2020). We assumed that micrometeorological conditions were similar across sites; however, we have no micrometeorological data to confirm this assumption. Therefore, there may be some unexplainable variability related to the unknown micrometeorological variability across sites.
In our research, we also found that among the shared microorganisms between all the phyllosphere and rhizosphere samples from the different host species, several shared genera were ubiquitous.
This might be because these shared microorganisms are generalists, which could transfer horizontally between distantly related plants and survive well (Frank, Saldierna Guzmán, & Shay, 2017). It was also possible that these shared microorganisms might be transmitted in plants through a vertical transfer of seeds and pollen, or horizontal transfer of soil, atmosphere, and insects (Frank et al., 2017).
In our study, the proportion of shared microorganisms between the phyllosphere and rhizosphere among three different plant species was a little bit smaller than what was generally observed in previous studies (Knief et al., 2012;Martins et al., 2013). One reason may be that one of the primer sets used in this study was selected to screen  (Seo, Keum, & Li, 2009;White, Sutton, & Ringelberg, 1996). Sphingomonas species are often found in association with plants. Among the members of this genus, Sphingomonas paucimobilis has been shown to exhibit antagonism against the phytopathogenic fungus Verticillium dahlia (Berg & Ballin, 1994). Many strains have been isolated from the rhizosphere (Takeuchi et al., 1995). In our study, the abundance of the genus Sphingomonas was significantly higher in the phyllosphere than that in the rhizosphere, especially in association with the plant Wedelia chinensis. Due to the catabolic capacity and widespread distribution of Sphingomonas, the tropical plants in this study may play a significant role in environmental protection. The genus Actinomycetospora has been predominantly isolated from subtropical/tropical regions (Jiang et al., 2008). It is believed that the abundance or diversity of Actinomycetospora correlates with the climate (Yamamura et al., 2011a). Some strains belonging to the Actinomycetes genus have also been isolated from lichen samples (Yamamura et al., 2011b).
Nocardioides is a common endophytic actinobacteria genus isolated from a diverse range of plant species, including those found in estuarine/mangrove ecosystems and algae and/or seaweeds of the marine ecosystems (Govindasamy, Franco, & Gupta, 2014). Interestingly, the as GA producers in previous studies (Kawaide, 2006;MacMillan, 2001 (Crous & Wingfield, 1996), which would need to be prevented.  (Soo, Wood, Grzymski, Mcdonald, & Cary, 2009). Therefore, the chemical and physical characteristics of IP-R could be the reason behind the strong relationship between IP-R and Mastigocladus, but this remains speculative.

ACK N OWLED G M ENTS
This study was supported by the National Natural Science QYZDB-SSW-DQC026).

E TH I C S S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are included in the main manuscript apart from the raw sequencing data files of bacteria, fungi, and diazotrophs which are available at the NCBI Sequence Read Archive under accession numbers SRP148402, SRP158738, and SRP144329, respectively.