Contrasting Patterns in Diversity and Community Assembly of Phragmites australis Root-Associated Bacterial Communities from Different Seasons

Understanding the composition and assembly mechanisms of root-associated microbial communities of plants is crucial for understanding the interactions between plants and soil. Most previous studies of the plant root-associated microbiome focused on model and economic plants, with fewer temporal or seasonal investigations. The assembly mechanisms of root-associated bacterial communities in different seasons remain poorly known, especially for the aquatic macrophytes. In this study, we compared the diversity, composition, and relative importance of two different assembly processes (stochastic and deterministic processes) of bacterial communities associated with bulk sediment and the rhizosphere and endosphere of Phragmites australis in summer and winter. While we found apparent differences in composition, diversity, and assembly processes of bacterial communities among different compartments, season played important roles in determining BCCs and their diversity patterns and assemblages. We also found that endosphere bacteria mainly originated from the rhizosphere. The results add new knowledge regarding the plant-microbe interactions in aquatic ecosystems.

ABSTRACT

The common reed (Phragmites australis), a cosmopolitan aquatic macrophyte, plays an important role in the structure and function of aquatic ecosystems. We compared bacterial community compositions (BCCs) and their assembly processes in the root-associated compartments (i.e., rhizosphere and endosphere) of reed and bulk sediment between summer and winter. The BCCs were analyzed using high-throughput sequencing of the bacterial 16S rRNA gene; meanwhile, null-model analysis was employed to characterize their assembly mechanisms. The sources of the endosphere BCCs were quantitatively examined using SourceTracker from bulk sediment, rhizosphere, and seed. We observed the highest α-diversity and the lowest β-diversity of BCCs in the rhizosphere in both seasons. We also found a significant increase in α- and β-diversity in summer compared to that in winter among the three compartments. It was demonstrated that rhizosphere sediments were the main source (∼70%) of root endosphere bacteria during both seasons. Null-model tests indicated that stochastic processes primarily affected endosphere BCCs, whereas both deterministic and stochastic processes dictated bacterial assemblages of the rhizosphere, with the relative importance of stochastic versus deterministic processes depending on the season. This study suggests that multiple mechanisms of bacterial selection and community assembly exist both inside and outside P. australis roots in different seasons.

IMPORTANCE Understanding the composition and assembly mechanisms of root-associated microbial communities of plants is crucial for understanding the interactions between plants and soil. Most previous studies of the plant root-associated microbiome focused on model and economic plants, with fewer temporal or seasonal investigations. The assembly mechanisms of root-associated bacterial communities in different seasons remain poorly known, especially for the aquatic macrophytes. In this study, we compared the diversity, composition, and relative importance of two different assembly processes (stochastic and deterministic processes) of bacterial communities associated with bulk sediment and the rhizosphere and endosphere of Phragmites australis in summer and winter. While we found apparent differences in composition, diversity, and assembly processes of bacterial communities among different compartments, season played important roles in determining BCCs and their diversity patterns and assemblages. We also found that endosphere bacteria mainly originated from the rhizosphere. The results add new knowledge regarding the plant-microbe interactions in aquatic ecosystems.

INTRODUCTION

The rhizosphere—the interface between plants and soil—represents one of the richest microbial ecosystems on Earth. The physicochemical properties and bacterial community compositions (BCCs) of the rhizosphere are strongly influenced by root growth, respiration, and nutrient exchange (1, 2). Bacteria within the endosphere—the microbial habitat inside a plant root (2)—help plants resist phytopathogens (3) and improve plant growth through promoting the production of phytohormones (2). Therefore, the rhizosphere and endosphere are very important microbial habitats for plant-microbe as well as plant-soil interactions.

Unlike the BCCs in bulk soil, root-associated BCCs are influenced by plants’ activities, and they may vary between a plant’s different developmental stages (4–7). Current evidence illustrates that the initial BCCs of the rhizosphere are similar to those of the surrounding soil, and the BCCs become more plant specific as plants grow and develop (4–8). For instance, Dombrowski et al. (4) observed marked differences in BCCs from the Arabidopsis rhizosphere throughout a 28-week greenhouse experiment as the plant passed through three growth periods. In a field study, Zhang et al. (7) observed that the rhizosphere BCCs of rice varied considerably during the initial period of plant growth. Nonetheless, most previous studies focused mainly on annual plants under controlled conditions or during a short interval; thus, these plants and their associated bacterial communities experienced had minimal community variation given the negligible environmental change during the experiment. Perennial plants growing over multiple seasons undergo various growth periods usually associated with marked changes in environmental factors, such as temperature and nutrient availability. In addition, perennial plants usually go into dormancy or a quiescent state during the winter, accompanied by aboveground leaf fading or browning. As a result, the photosynthetic capacity of the aboveground part is also reduced, which may affect the secretory activity of the underground part (i.e., the root). The variability of these environmental factors may lead to strong variations of microbial communities associated with roots of plants between different seasons.

Understanding the mechanisms of microbial community assembly is a central topic in microbial community ecology (9–12). Despite the increased recognition of the taxonomic and phylogenetic diversity of plant microbiomes, the mechanisms underlying assembly of microbial communities remain poorly understood (13). It is generally accepted that both neutral (stochastic) and deterministic processes determine the assembly of microbial communities (9–11, 14). Based on the null-model theory, several studies have reported the recruitment mechanisms of the root-associated microbiome of plants (15–18). For example, Fan et al. (15) found that the importance of deterministic processes in determining diazotrophic communities decreased with distance away from wheat roots; therefore, the rhizosphere can selectively filter the specific functional microbial groups. In contrast, stochastic processes drove the random phylogenetic turnover of root-associated microbial communities of speargrasses subjected to the extreme conditions in deserts (16). These studies suggested that the relative importance of ecological processes (i.e., stochastic and deterministic processes) that drive the assembly of microbial communities may vary depending on the edaphic environment and plant species.

Most previous studies about plant root-associated microbiomes focused on terrestrial plants without regard to temporal and developmental parameters. Aquatic macrophytes, as major components of aquatic ecosystems, serve important functions in purifying water and protecting shores (19, 20). However, few studies have investigated the BCCs in their rhizospheres and endospheres (21–23). Phragmites australis (common reed) is a cosmopolitan aquatic macrophyte and plays an important role in the structure and function of aquatic and wetland ecosystems. We investigated the diversity, composition, and assembly processes of bacterial communities associated with different compartments (i.e., bulk sediment, rhizosphere, and endosphere) of reed in summer and winter. Furthermore, we quantitatively examined the sources (bulk sediment, rhizosphere, and seed) of the endosphere BCCs using SourceTracker. We aimed to determine the (i) variations of BCCs between bulk sediment, the rhizosphere, and the endosphere of reed in summer and winter, (ii) mechanisms underlining the assembly of BCCs among the three compartments during different seasons, and (iii) sources of endosphere BCCs in both seasons.

RESULTS

Physicochemical properties of the bulk sediments.

The measured physicochemical properties of the collected bulk sediments are presented in Table S1 in the supplemental material. Higher temperature was observed in summer (32°C) than in winter (8°C). Pairwise t tests indicated that the winter possessed higher contents of total nitrogen (TN), total phosphorus (TP), and loss on ignition (LOI) than those in summer (P < 0.05) (Table S1).

Richness and diversity of the bacterial community between compartments and seasons.

We obtained a total of 1,537,082 high-quality sequences from 60 samples representing three different compartments (bulk sediment, rhizosphere, and endosphere) across two seasons (summer and winter) after trimming, denoising, and removing chimeras, nonbacterial sequences, and rare sequences. The minimum sequence number (12,753 sequences) was found in the endosphere compartment in summer; therefore, to compare the diversity of each group, the sequences obtained from each sample were subsampled randomly into 12,753 sequences. These high-quality sequences were then clustered into 9,165 operational taxonomic units (OTUs) at a 97% cutoff for all rarefied samples. Rarefaction curves indicated a thorough sampling across the entire bacterial communities (Fig. S1).

The bacterial diversity of bacterial communities was estimated by OTU richness, Faith’s phylogenetic diversities (PD) (24), and the Shannon index, respectively (Fig. 1). Two nonparametric tests (Kruskal-Wallis H test and Mann-Whitney U test) were applied to compare differences in diversities of bacterial communities among the six sample groups. We observed no significant differences in bacterial diversities between the bulk sediment and rhizosphere compartments irrespective of the seasons; however, the bacterial diversity (OTU richness and Faith’s PD) of rhizosphere was slightly higher than that of the bulk sediment in summer (Fig. 1; Fig. S1). Moreover, the bacterial diversities in the endosphere compartment were significantly lower than those in the bulk sediment and rhizosphere compartments, regardless of season (Kruskal-Wallis H test, P < 0.05) (Fig. 1; Fig. S1). The bacterial diversities of the bulk sediment and the rhizosphere in summer were significantly higher than those in winter (Mann-Whitney U test, P < 0.05) (Fig. 1; Fig. S1). In summary, the diversity of bacterial communities outside the root tissue of P. australis was much higher than that inside the tissue, and, except for the endosphere, bacterial assemblages were more diverse in summer than in winter (Fig. 1; Fig. S1).

FIG 1.

Richness and diversity of the bacterial community of the various compartments and seasons. (A) OTU richness; (B) Faith’s PD; (C) Shannon index. Boxes represent the upper and lower quartiles, horizontal lines indicate the median, whiskers show 95% range, and points are outliers. Uppercase and lowercase letters above the bars indicate significant differences (P < 0.05) between the different compartments in winter and summer, respectively (Kruskal-Wallis H test; multiple comparisons were corrected by Bonferroni correction; n = 10 for each group). Asterisks indicate significant differences between two seasons in the same compartment (Mann-Whitney U test; n = 10 for each group). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Variations of BCCs between compartments and seasons.

We used nonmetric multidimensional scaling (NMDS) based on Bray-Curtis and unweighted UniFrac distance (phylogeny-based) matrices to examine the differences between the BCCs of the six sample groups. We found significant differences in BCCs between the three compartments and two seasons (Fig. 2A and B; Fig. S2). Results of both the analysis of similarity (ANOSIM) (Bray-Curtis dissimilarity, R = 0.783 to 1 and P < 0.001; unweighted UniFrac dissimilarity, R = 0.623 to 1 and P < 0.001) and the permutational multivariate analysis of variance (PERMANOVA) (Fig. S2) confirmed that the clusters were statistically significant among the different groups. Within the same season, the bacterial communities in the endosphere possessed the highest beta dissimilarity, or the highest variation in community composition, across the three compartments (Kruskal-Wallis H test, P < 0.05). The rhizosphere BCCs exhibited the lowest beta dissimilarity in both seasons (Fig. 2C and D). As well, the BCCs of three compartments in summer exhibited higher beta diversities than those in winter (Mann-Whitney U test, P < 0.05) (Fig. 2C and D).

FIG 2.

Comparisons of β-diversity of bacterial community among the various compartments during different seasons. Shown are nonmetric multidimensional scaling (NMDS) plots of the bacterial community composition of the different sample groups based on Bray-Curtis distance (taxonomy dissimilarity) (A) and unweighted UniFrac distance (phylogeny dissimilarity) (B). Distances in each group are compared using boxplots (C and D). Boxes represent the upper and lower quartiles, horizontal lines indicate the median, whiskers show 95% range, and points are outliers. Different letters above the bars indicate significant differences (P < 0.05) among the various sample groups (Kruskal-Wallis H test; multiple comparisons were corrected by Bonferroni correction; n = 10 for each group). Asterisks indicate significant differences between the two seasons for the same compartment (Mann-Whitney U test; n = 10 for each group). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

PERMANOVA (Table 1) showed that season (including its associated environmental factors) explained the greatest amount of BCC variation, accounting for more than 30% (Bray-Curtis dissimilarity, R² = 0.309 and P < 0.001; unweighted UniFrac dissimilarity, R² = 0.336 and P < 0.001). This was followed by compartment (Bray-Curtis dissimilarity, R² = 0.222 and P < 0.001; unweighted UniFrac dissimilarity, R² = 0.124 and P < 0.001) and interaction between the two factors (Bray-Curtis dissimilarity, R² = 0.120 and P < 0.001; unweighted UniFrac dissimilarity, R² = 0.073 and P < 0.001) (Table 1).

TABLE 1.

PERMANOVA results based on Bray-Curtis and unweighted UniFrac distance to partition sources of variation (compartment, season, interaction between compartment and season) for bacterial communities^b

Distance	Source of variation	SS	MS	F. model	R2	Pa
Bray-Curtis	Compartment	3.628	1.814	17.210	0.222	0.001
	Season	5.042	5.042	47.838	0.309	0.001
	Interaction	1.958	0.979	9.289	0.120	0.001
	Residuals	5.692	0.105		0.349
	Total	16.320			1
Unweighted UniFrac	Compartment	1.297	0.648	7.182	0.124	0.001
	Season	3.513	3.513	38.917	0.336	0.001
	Interaction	0.760	0.380	4.206	0.073	0.001
	Residuals	4.874	0.090		0.467
	Total	10.443			1

^aStatistical significance (P) was calculated based on sequential sums of squares from 999 permutations.

^bSS, sum of squares; MS, mean squares; F. model, F statistic.

Taxonomic differences across compartments and seasons.

We identified 16 bacterial phyla (and classes for Proteobacteria) (Fig. 3; Fig. S3; Table S2). All phyla having a relative abundance of <1% were defined as minor. The relative abundances of the dominant bacterial phyla varied between compartments and seasons (Fig. 3). Unclassified bacteria—defined as those sequences having a <80% confidence value as assigned to any phylum in SILVA with RDP—accounted for a large proportion of each sample group (16.17% ± 1.72% to 30.38% ± 15.62%). With the exception of the unclassified bacterial group, Actinobacteria (12.85% ± 2.64% to 17.27% ± 4.34%) dominated the bacterial communities of both the bulk sediment and rhizosphere, regardless of the season. Alphaproteobacteria were more abundant in winter (two-way ANOVA, F = 39.399 and P < 0.001), whereas Deltaproteobacteria (two-way ANOVA, F = 16.440 and P < 0.001) and Firmicutes (two-way ANOVA, F = 9.974 and P < 0.001) exhibited a significantly higher abundance in summer than in winter, regardless of the compartment (Fig. 3). The bulk sediment samples were more enriched with Acidobacteria than the other two compartments (two-way ANOVA, F = 36.638 and P < 0.001) (Fig. 3). Within the same season, Bacteroidetes had the greatest relative abundance in the rhizosphere compartment as opposed to the bulk sediment and endosphere compartments (two-way ANOVA, F = 44.027 and P < 0.001) (Fig. 3). Gammaproteobacteria had the greatest relative abundance in the endosphere in winter (two-way ANOVA, F = 71.248 and P < 0.001) (Fig. 3).

FIG 3.

Comparisons of dominant bacterial communities among the sample groups at the level of phyla or subphyla. All phyla having a relative abundance of >1% are shown, and phyla having a relative abundance of < 1% are defined as minor. Each error bar represents the standard deviation of each sample group. Asterisks indicate significant differences between the sample groups (two-way ANOVA; *, P < 0.05; **, P < 0.01; ***, P < 0.001) based on the different factors (blue represents the effects of compartment, green represents the effects of season, and red represents the effects of interaction).

LEfSe (linear discriminant analysis effect size) analysis indicated strong variation in BCCs at the genus level. We listed the top 20 genera with the highest relative abundances on a heat map (Fig. S4). The relative abundances of two genera, Bacillus and Paenibacillus (both affiliated with Firmicutes), significantly decreased from bulk to rhizosphere and to endosphere (Fig. S4, Wilcoxon rank sum test, P < 0.05, and linear discriminant analysis [LDA] scores > 2). Only a few genera, including Dechloromonas, Denitratisoma, Methylibium, and Leptothrix (all affiliated with Betaproteobacteria), were found to be significantly abundant in both the rhizosphere and endosphere compartments (Fig. S4, Wilcoxon rank sum test, P < 0.05, and LDA scores > 2). In contrast, relatively more genera (Agromyces, Arthrobacter, Mycobacterium, Streptomyces, Devosia, Kaistobacter, Phenylobacterium, and Steroidobacter) were abundant exclusively in the rhizosphere or endosphere compartment in different seasons (Fig. S4, Wilcoxon rank sum test, P < 0.05, and LDA scores > 2). A Venn diagram illustrates the distribution of different OTUs for the three compartments in both seasons (Fig. S5). The number of shared OTUs between two compartments decreased from the bulk sediment to the rhizosphere compartment and then to the endosphere compartment in both seasons, as did the number of unique OTUs within each compartment (Fig. S5).

Assembly processes of the bacterial communities between different seasons.

We calculated the beta nearest taxon index (βNTI) values to compare the relative importance of two kinds of assembly processes (deterministic and stochastic) of the BCCs in the bulk sediment, rhizosphere, and endosphere compartments of each season (Fig. 4). Both deterministic and stochastic processes drove the assembly of BCCs (Fig. 4). In winter, deterministic processes were the main force that determined the bacterial community assembly of the bulk and rhizosphere compartments (62.2% for the bulk sediment and 84.4% for the rhizosphere) (Fig. 4). Stochastic processes, however, played more important roles in driving the assembly of bacterial communities in both the bulk sediment (66.7%) and the rhizosphere compartment (73.3%) in summer (Fig. 4). Surprisingly, the majority of the βNTI values for the endosphere BCCs during both seasons fell between −2 and 2 (62.2% in winter and 86.7% in summer) (Fig. 4), suggesting that stochastic processes dominated the assembly of the endosphere BCCs in both seasons.

FIG 4.

Distribution of beta nearest taxon index (βNTI) based on the different compartments and seasons. Positive (or negative) βNTI values indicate greater (or less) than expected turnover in phylogenetic composition. The horizontal dotted blue line (above +2 or below −2 are statistically significant) shows the 95% confidence intervals around the expectation under neutral community assembly.

Sources of the bacterial community in the endosphere.

We quantitatively determined the sources of the bacterial community found within the endosphere compartment in both seasons. Rhizosphere sediments were more likely to be the primary source of the bacterial community for the root invasion, contributing, on average, between 71.06% and 67.65% of the BCCs in winter and summer (Fig. 5). In addition, a small proportion of the endosphere bacteria originated from seed endophytic bacteria (14.66% in winter and 7.86% in summer) (Fig. 5). Only 2% to 4% of the endosphere bacteria originated from bulk sediment (3.35% in the winter and 2.03% in summer) (Fig. 5). However, these three sources did not differ significantly between the two seasons (Mann-Whitney U test, P > 0.05) (Fig. 5).

FIG 5.

Sources of bacterial community in the root endosphere compartment in summer and winter as predicted by SourceTracker. Each error bar represents the standard deviation of each source. Uppercase and lowercase letters above the bars indicate significant differences (P < 0.05) among different compartments in winter and summer, respectively (Kruskal-Wallis H test; multiple comparisons were corrected by Bonferroni correction; n = 10 for each group).

DISCUSSION

Phragmites australis (common reed) is a cosmopolitan perennial emergent plant in aquatic ecosystems (20, 25) and has been widely used in constructed wetlands for pollution control (19, 20, 25–27). It was found that the rate of removal of nitrogen, potassium, and chemical oxygen demand (COD) by reed differed significantly between different seasons (28, 29). These differences may be partly attributable to the different BCCs associated with its root and plant-bacterial interactions. In this study, while we observed apparent differences in composition, diversity, and assembly processes of bacterial communities among the bulk sediment, rhizosphere, and endosphere compartments, season accompanied by changes in environmental factors explained the largest amount of variation (accounting for more than 30%) of the BCCs.

Higher α- and lower β-diversity of BCCs in the rhizosphere microhabitat.

Although the rhizosphere occupies the same medium as the bulk sediment, bacterial diversity in the rhizosphere was slightly higher than that in the bulk sediment in summer (Fig. 1; Fig. S1). In a hypersaline pond, Borruso et al. (30) also found that bacterial diversity in the rhizosphere of P. australis was higher than that in bulk sediment. Similarly, several macrophyte species, including Acorus calamus (31), Spartina alterniflora (32), and Avicennia marina (33), also exhibited a greater bacterial diversity in the rhizosphere than in bulk sediment. Therefore, the results showing higher diversity in the reed rhizosphere than in the bulk sediment are consistent with previous studies of different aquatic macrophytes (30–33). Previous studies have focused on several terrestrial plants that illustrate how root exudates of maize (34) and barley (35) enrich and filter for microbes maintaining specific functions (36); therefore, they observed a reduced diversity of the bacterial community in the rhizosphere compartment (36). Unlike the terrestrial soil, lake sediments are usually anaerobic a few millimeters below the surface. An emergent plant not only provides nutrients through root exudates (37) but also releases oxygen into the surrounding sediments to enhance microbial colonization (37, 38). Armstrong (39) found that emergent plant roots release oxygen through the aerenchyma to the rhizosphere, a process known as root radial oxygen loss (ROL). Other research suggested that wetland plants transfer a portion of the oxygen—produced through photosynthesis within the aerenchyma to the roots—to the wetland substrate (40, 41). Many previous studies have found direct and indirect evidence of aeration in the rhizosphere of Phragmites australis, including increased oxygen partial pressure and oxidation reduction potential (ORP) (42–45). Therefore, we speculated that increased oxygen content might be a major reason for the increased bacterial diversity in the rhizosphere in summer. In addition, the lower bacterial β-diversity (i.e., variation in community composition) in the rhizosphere (Fig. 2C and D) may suggest filtration effects in the root-sediment interface, which have been observed for many terrestrial plants (8, 46).

Enrichment of specific bacterial taxa in the rhizosphere and endosphere.

For both seasons, Proteobacteria, Actinobacteria, Acidobacteria, Bacteroidetes, and Firmicutes were the most abundant phyla (Fig. 3; Fig. S3; Table S2), consistent with previous studies (22, 30, 47–49). We selected the significantly different abundant genera influenced by seasons and compartments and found that some of these groups were related to specific functions. In our study, the relative abundances of Arthrobacter (affiliated with Actinobacteria) in the rhizosphere and Streptomyces (affiliated with Actinobacteria) and Devosia (affiliated with Alphaproteobacteria) in the endosphere were significantly higher than those in the bulk sediment (Fig. S4, Wilcoxon rank sum test, P < 0.05, and LDA scores > 2). These genera have been reported as important plant growth-promoting rhizobacteria (PGPR) (50–52). PGPR could greatly expand host adaptations and performance by various promoting activities, including indole acetic acid (IAA) production, siderophore production, phosphate solubilization, and induced systemic resistance (53, 54). In addition, Agromyces (affiliated with Actinobacteria) and Mycobacterium (affiliated with Actinobacteria) were enriched in the rhizosphere, while Methylibium (affiliated with Betaproteobacteria) was enriched in both the rhizosphere and endosphere; these taxa all play important roles in the degradation of a wide spectrum of substrates, such as sugars, acids, cellulose, and xylan, as well as many aromatic compounds that can be excreted by the root (55, 56). Dechloromonas (affiliated with Betaproteobacteria) organisms are autotrophic, denitrifying, and phosphate-accumulating bacteria (57, 58) which have been found in the seeds of P. australis and Typha angustifolia L. (48). This genus was more abundant in the rhizosphere and endosphere compartments in summer than in winter (Fig. S4, Wilcoxon rank sum test, P < 0.05, and LDA scores > 2). Therefore, the growth of P. australis may promote the enrichment of Dechloromonas around the root and enhance denitrification processes and phosphorus absorption by the root. The respective abundances of these genera suggest that the rhizosphere and endosphere of P. australis tend to resist adverse external factors by enriching bacteria that are beneficial for the plant’s growth and health (59, 60).

Season accompanied by the changes in environmental factors influences the composition and assembly of root-associated bacterial communities.

Season explained the majority of the variation in the BCCs of root-associated compartments in P. australis (Table 1). In this study, the season encompasses multiple factors, including the growth condition of P. australis and changes in environmental variables (e.g., temperature) (Table S1). In winter, reeds may go into the dormant stage due to relative lower temperatures and photosynthetic capacity. While in summer, reeds may recover into the vegetative stage with new grown leaves. Significant differences in the diversity and community composition of the bulk sediments between the two seasons (Fig. 1B; Fig. 2C and D) were found in the present study; therefore, we cannot ignore the effects of the external environment on the composition of the bacterial community. However, differences in diversity (α-diversity and β-diversity) (Fig. 1; Fig. 2C and D) of the bacterial community between the rhizosphere and bulk compartments were higher at the vegetative period (summer) than those of the dormant period (winter); this pattern relates to the plant itself, as the rhizosphere and bulk compartments shared the same environment during the same season.

For the plant, the amount and composition of root exudate varied between the seasons, which could affect the respective responses of root-associated bacterial communities (61). In a laboratory experiment, the root exudates of reed increased with plant growth during the experimental period (50 days) (ranging from 0.03 to 1.53 μmol g⁻¹ day⁻¹) (62). Therefore, increased amounts of root exudate (e.g., organic acids) may be produced during the vegetative stage (summer) in our study. Nutrient enrichment accelerates the growth of some rare bacterial species; therefore, the diversity of the bacterial community is increased, and the relative importance of stochastic processes is also enhanced in the assembly of the bacterial community (63).

Except for the specific effects of the plant itself, temperature is likely the most important factor contributing to differences in the bacterial communities given that it changed markedly between the two seasons (8°C in winter and 32°C in summer) (Table S1). Temperature is one of the major determinants of bacterial assemblage composition, as the temperature optima of bacterial taxa differ greatly (64, 65). In this study, the wide temperature difference between the two sampling times (24°C) could have resulted in the differences of bacterial community between the seasons. Moreover, almost all biological processes are affected by temperature increases, as the metabolic rates increase exponentially as temperatures rise (26, 65, 66). Increased temperature promotes the kinetic metabolic rate of microorganisms, thereby increasing the diversity and the stochasticity of the colonization and extinction of members of the bacterial community (65). Therefore, we can explain why the bacterial communities within the same compartments were more affected by stochastic processes in summer marked by higher temperatures.

Composition and source of the bacterial community in the endosphere.

Previous studies have indicated that endosphere bacteria originated mainly from the soil environment because root exudates and rhizodeposits attract microorganisms (3, 59, 67). Our results were in line with this idea, as 69.68% to 74.41% of the endosphere bacteria originated from the rhizosphere and bulk compartments (Fig. 5). Most of the observed endosphere bacterial community originated from the rhizosphere sediment (71.06% in winter and 67.65% in summer) (Fig. 5), consistent with the idea that the invasion of root endophytic microbiome follows a two-step selection model in which microorganisms first colonize the rhizosphere and then invade the roots (2, 8, 46). In addition, the endophytic bacterial community of the seed also contributed to the composition of endosphere bacteria (Fig. 5). This is consistent with recent discoveries (68–70) that endophytes could be vertically transmitted through seeds (59). Despite the fact that significant differences in the bacterial communities were observed in the bulk sediment and rhizosphere compartments between the two seasons (Fig. 1 and 2), the contributions of these sources (bulk sediment, rhizosphere, and seed) to the endosphere bacterial community were consistent between the two seasons (Mann-Whitney U test, P > 0.05) (Fig. 5); thus, the assembly of the endosphere bacterial community remained stable.

It is generally assumed that roots are effective habitat filters, restricting membership to progressively more narrowly defined lineages as environments deviate from soil to roots (1, 71). Therefore, the lowest diversity of the bacterial community was observed within the root tissue (Fig. 1), in line with previous studies (8, 71–73). More surprising, however, the community composition of the endosphere bacteria showed a high heterogeneity, including a higher β-diversity (Fig. 2C and D), regardless of the season. Gottel et al. (73) also observed more variable microbial communities (including bacterial and fungal communities) in the endosphere than those of the rhizosphere of Populus deltoides. They attributed these differences to the sporadic and nonuniform colonization of roots by the bacterial community of the rhizosphere. In our study, stochastic processes dominated the assembly of the endosphere bacterial community (Fig. 4), seemingly supporting this hypothesis. A passive mode of transmission has bacteria survive after they enter the plant—often via natural breaks or wounds in roots (59, 67, 74); these breaks or wounds may be caused by stochastic processes during the growth of the plant. These random processes provide opportunities for the colonization of specific bacterial species within individual plants.

In conclusion, this study revealed the differences in diversity, composition, and assembly processes of root-associated bacterial communities of P. australis in two specific seasons. We showed that season played more important roles in the diversity, composition, and assembly processes of the root-associated bacterial communities. In addition, our results suggested that the surrounding sediments provide the primary sources of bacteria that invade the roots. While we could explain the possible assembly mechanisms of bacterial communities in the rhizosphere and endosphere, there remain uncertainties regarding the considerable effect of environment variables within in situ field conditions. Therefore, further larger-scale investigations or experiments under controlled conditions are required.

MATERIALS AND METHODS

Study sites and sample collection.

The sampling site (118.870°E, 32.052°N) is located near a lake on the Zijin Mountain in Nanjing, Jiangsu Province, China, where Phragmites australis is the dominant aquatic macrophyte and relatively few anthropogenic activities affect the lake. A more detailed description of the sampling sites can be found in our previous study (75) and Fig. S6. Samples in winter representing the dormant stage were collected in December of 2017 at 8°C (air temperature), and samples in summer representing the vegetative stage were collected in July of 2018 at 32°C (air temperature). To reduce bias related to random environmental variation and ensure a sufficient number of samples for subsequent analyses, we selected two 10-m by 10-m plots for sampling. Within each plot, samples were collected using a five-point sampling method (center point and four vertex positions) (Fig. S7). We collected 10 sediment cores that were each 30 cm in diameter and 5 to 15 cm deep for each P. australis station. And at least one taproot of the same length as the core was present in the center of each sediment core. The collected sediment core containing root was transported to the laboratory intact for further processing. Bulk sediments were collected at the same sediment depth using a core sampler (model no. DM60; China Mingyu Holdings Group, Shanghai, China) at about 30 cm away from each sampled P. australis. Moreover, 10 seed samples were also collected from the ears of 10 individual P. australis plants. All samples were kept at 4°C and transported to the laboratory within 4 h of collection. Samples were stored at –80°C until further processing.

All samples, except the seed samples, were freeze-dried using a vacuum freeze dryer (FD-1A-50; Beijing Boyikang Lab Instrument Co. Ltd., Beijing, China). Then the roots and rhizosphere samples were separated as described by Bulgarelli et al. (1). Briefly, for each sediment core, the loosely attached sediments were shaken off. After that, the fibrous roots scattering over the taproot, which was covered by ca. 1-mm freeze-dried sediments, were collected of all the depths by sterilized scissors and put into a 50-ml sterile Falcon tube with 35 ml of sterile phosphate-buffered saline (PBS) (130 mM NaCl, 7 mM Na₂HPO₄, 3 mM NaH₂PO₄ [pH 7.0], and 0.02% Silwet L-77). Then the whole Falcon tube containing roots and PBS was put onto a shaking platform for 20 min at 180 rpm. The washing buffer was then centrifuged (1,500 × g, 20 min); the collected particles were homogenized and defined as the rhizosphere sample (i.e., all the freeze-dried sediment adhered tightly to the whole root, a layer about 1 mm thick) and stored at –80°C. We repeated the washing steps (20 min at 180 rpm) twice more. If the supernatant was still cloudy or colored, we continued washing the roots until the supernatant was clear. The roots were then transferred to another 50-ml Falcon tube (containing 35 ml of PBS) and sonicated for 10 cycles consisting of 30-s pulses at 160 W and breaks of 30 s using an ultrasonic cleaner (model no. KH-500DB; Hechuang Ultrasonic, China). After all washing and sonicating steps, the remaining roots were frozen and defined as the endosphere compartment. We confirmed that the surfaces of all root samples were sterile by rubbing a subsampled root from each collection onto LB plates and incubating the plates overnight at 30°C (73). Note that each rhizosphere and endosphere sample that we obtained was specific to the same plant. Seed samples were surface sterilized following the protocol of Gao and Shi (48).

Sediment properties.

Sediment properties of the bulk samples, including pH, total nitrogen (TN), total phosphorus (TP), nitrate nitrogen (NO₃⁻-N), nitrite nitrogen (NO₂⁻-N), and loss on ignition (LOI), were measured following the methods of Zeng et al. (76).

DNA extraction, PCR amplification, and high-throughput sequencing.

Seeds and roots were pulverized with a mortar and pestle in liquid nitrogen prior to DNA extraction. Freeze-dried sediment samples (including rhizosphere and bulk compartments) were ground and homogenized before DNA extraction. DNA was extracted from all samples using the FastDNA SPIN for soil kit (MP Biomedicals, Solon, OH) following the manufacturer’s instructions. After checking that DNA was extracted successfully from each sample via electrophoresis on a 0.8% (wt/vol) agarose gel, the concentration and quality (purity) of the extracted DNA were evaluated by measuring the absorbance at 260 nm and 280 nm using a BioPhotometer (Eppendorf, Hamburg, Germany). Each sample was extracted twice, and duplicate DNAs from each sample were pooled and kept at –80°C until further analysis.

We selected the primers 799F (5′-AACMGGATTAGATACCCKG-3′) and 1115R (5′-AGGGTTGCGCTCGTTG-3′), which target the V5-V6 hypervariable regions of the bacterial 16S rRNA gene, for PCR amplification to avoid contamination of chloroplast DNA (77, 78). PCR amplification was performed in a 20-μl mixture that including 10 ng of template DNA, 4 μl of 5× FastPfu buffer, 2 μl of 2.5 mM deoxynucleoside triphosphates (dNTPs), 0.8 μl of 5 μM (each) concentrations of the forward and reverse primers, and 0.4 μl of FastPfu polymerase. Thermal cycling conditions were set at a 2-min initial denaturation at 95°C, 25 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s, and a final extension at 72°C for 5 min. Following amplification, the PCR products were verified by 2% (wt/vol) agarose gel electrophoresis. After amplifying each sample in triplicate, we purified the combined PCR products with the AxyPrepDNA gel purification kit (Axygen Biotechnology Hangzhou Ltd., Hangzhou, China). High-throughput sequencing was performed using an Illumina MiSeq platform at Shanghai Biozeron Co., Ltd.

Sequence analysis.

The following standard operating procedure of QIIME (v.1.9.1) (79) was applied to filter the sequencing quality for each forward and reverse fastq file. First, libraries (paired-end data) were demultiplexed into each sample based on the index sequences downloaded from the Illumina MiSeq platform. Next, paired-end sequences of each sample were trimmed for their quality and length using Trimmomatic (v.0.33) (80). The 250-bp reads were truncated at any site having an average quality score of <20 over a 10-bp moving window, discarding the truncated reads that were <50 bp. Reads with any barcode mismatch or any nucleotide mismatch in primer matching or that contained ambiguous characters were removed. Trimmed sequences overlapping only by >10 bp and having a mismatch of <0.25 were assembled based on the overlap sequences with FLASH (v.1.2.11) (81). We discarded pair-end reads that could not be assembled.

We then detected and removed chimera reads from downstream analyses using USEARCH software v.4.2.52 based on the UCHIME algorithm with a de novo strategy (82). Finally, operational taxonomic units (OTUs) were clustered using the script pick_otus.py of QIIME via the uclust method at a similarity cutoff of 97%. Rare OTUs (sequence number < 17) were removed from downstream analyses (83). In addition, representative reads of each OTU were aligned against the SILVA v.132 databases (16S rRNA genes) and alignments were applied to calculate the phylogenetic tree in FastTree v.2.1 (84). Taxonomic information of each OTU was retrieved using the online RDP (Ribosomal Database Project) classifier with a bootstrap cutoff of 80% (85). Chloroplast, archaeal, mitochondrial, and unclassified reads were discarded. All samples were randomly rarefied to the lowest number of sequences (12,753 sequences) for further analysis.

Statistical analysis.

We calculated α-diversity (including OTU richness, Faith’s phylogenetic diversities [PD] [24], and Shannon index) using the command alpha_diversity.py in QIIME (79). β-Diversity of the bacterial community was calculated based on Bray-Curtis (86) and unweighted UniFrac (87) distances. We assessed significant differences in the diversity indices between the two seasons and three compartments via two nonparametric tests (Kruskal-Wallis H test and Mann-Whitney U test) (88). Nonmetric multidimensional scaling (NMDS) provided a visualization of the bacterial community composition of the sample groups. The bacterial community differences across the compartments and seasons were identified using analysis of similarities (ANOSIM) and permutational multivariate analysis of variance (PERMANOVA). Moreover, PERMANOVA tested whether compartment or season had a greater effect on the composition of the bacterial community (89). The relative abundances of dominant phyla and subphyla were compared between the compartments using SPSS v22.0 software (IBM-SPSS, Chicago, IL). We applied linear discriminant analysis effect size (LEfSe) (90), and only significant genera (Wilcoxon rank sum test, P < 0.05, and LDA scores > 2) were accepted if enriched in the three compartments. The relative abundances of significantly different genera were compared using a heat map. Similarity and clustering relationships of distinctive OTUs for each compartment were calculated and presented in a Venn diagram. All these analyses were performed using the functions in the vegan package (91) of R v.3.3.1 (92). SourceTracker (v.0.9.5), based on a combination of Bayes’ theorem and Gibbs sampling (93), was used to predict the sources of bacterial communities in the endosphere compartments. Bulk sediments, the rhizosphere compartments, and seed compartments were set as sources; the endosphere samples were set as sinks. Source predictions were run with 100 burn-ins, 10 random restarts, and a rarefaction depth of 20,000 (94).

Analysis of bacterial community assembly.

We applied a null-model theory to evaluate the assembly processes of bacterial communities associated with the three compartments (bulk sediment, rhizosphere, and endosphere). The mean nearest taxon distance metric (βMNTD) for each compartment of both seasons was calculated based on the phylogenetic turnover between communities in pairs (14) using the package picante (95) in R. In addition, we calculated the beta nearest taxon index (βNTI)—the difference between observed βMNTD and mean of the null distribution of βMNTD normalized by its standard deviation—to evaluate the dominant assembly processes (deterministic or stochastic) of the bacterial communities (96). βNTI values between −2 and +2 indicate the dominance of stochastic processes, while βNTI values above +2 or below −2 indicate the dominance of deterministic processes (14, 97).

Accession number(s).

The raw sequences obtained in this study have been submitted to the National Center for Biotechnology Information (NCBI) Sequence Read Archive database under BioProject accession number PRJNA528336. Please note that 10 sequences of rhizosphere community of July 2018 were involved in our previous study (75) to compare the differences of bacterial communities in the rhizosphere and phyllosphere compartments of Phragmites australis. These two batches of the data share the same BioProject number.