Community Dynamics of Arbuscular Mycorrhizal Fungi in High-Input and Intensively Irrigated Rice Cultivation Systems

Abstract

Application of a mycorrhizal inoculum could be one way to increase the yield of rice plants and reduce the application of fertilizer. We therefore studied arbuscular mycorrhizal fungi (AMF) in the roots of wetland rice (Oryza sativa L.) collected at the seedling, tillering, heading, and ripening stages in four paddy wetlands that had been under a high-input and intensively irrigated rice cultivation system for more than 20 years. It was found that AMF colonization was mainly established in the heading and ripening stages. The AMF community structure was characterized in rhizosphere soils and roots from two of the studied paddy wetlands. A fragment covering the partial small subunit (SSU), the whole internal transcribed spacer (ITS), and the partial large subunit (LSU) rRNA operon regions of AMF was amplified, cloned, and sequenced from roots and soils. A total of 639 AMF sequences were obtained, and these were finally assigned to 16 phylotypes based on a phylogenetic analysis, including 12 phylotypes from Glomeraceae, one phylotype from Claroideoglomeraceae, two phylotypes from Paraglomeraceae, and one unidentified phylotype. The AMF phylotype compositions in the soils were similar between the two surveyed sites, but there was a clear discrepancy between the communities obtained from root and soil. The relatively high number of AMF phylotypes at the surveyed sites suggests that the conditions are suitable for some species of AMF and that they may have an important function in conventional rice cultivation systems. The species richness of root-colonizing AMF increased with the growth of rice, and future studies should consider the developmental stages of this crop in the exploration of AMF function in paddy wetlands.

INTRODUCTION

Rice (Oryza sativa L.) is one of the economically most important plants worldwide and the main food crop cultivated in wetland environments. China is the largest rice producer and consumer in the world, and wetland rice production is of vital importance for food security. In the past 5 decades, rice production in China has more than tripled, mainly due to increased grain yield. Meanwhile, the overuse of nitrogen and phosphorus fertilizers as well as pesticides in rice production has induced serious challenges regarding water and soil pollution (1), calling for an urgent need to develop a “green” rice production system that can replace the conventional intensive cultivation system without a yield reduction.

Crop yield depends not only on the crop plant and its physical and chemical environment but also on biological interactions. It may be possible to increase crop yield and harvest quality and decrease the application of fertilizer by careful selection of the fungal partner in mycorrhizal associations. A prerequisite for making this possible is the study of existing associations. It is widely accepted that the symbiotic arbuscular mycorrhizal fungi (AMF; phylum Glomeromycota) play a key role in sustainable production systems and that rice readily forms a symbiotic relationship with these fungi when it is grown in drained soils (i.e., upland conditions) (2). AMF form symbiotic relationships with most terrestrial plants, including many crop plants. In exchange for the plant-assimilated carbon, the fungal partner benefits the host plant by facilitating plant access to mineral nutrients and water and by improving pathogen resistance (3). In addition to these direct effects, the AMF community composition and diversity may also influence plant productivity and community structure (4). In China, as well as in most other places, rice is grown under waterlogged conditions. Recently, the occurrence of AMF in wetland ecosystems has received increased attention (5, 6, 7). Evidence is accumulating that AMF are not only present but also ubiquitous in some wetland ecosystems (8, 9). The AMF communities of some wetland habitats have been investigated (7, 8, 9, 10, 11), but whether AMF occur in paddy rice fields under waterlogged conditions is still under debate (12). Some studies found that AMF are able to survive in waterlogged paddy wetlands (13), but there are also reports showing that AMF are absent or rare due to the anoxic conditions that often prevail in waterlogged systems (2, 14).

AMF communities are not static within an ecosystem as the fungi are affected by changing biotic and abiotic factors (3). In order to gain insights into the functional role of AMF, it is important to obtain more knowledge on their distribution dynamics in different ecosystems. The physiological stage of the host plant has been suggested to be an important factor shaping AMF communities as AMF are dependent on the quantity and quality of plant-assimilated carbon sources (3). However, studies of the dynamics of the AMF community inhabiting plant roots across different plant growth stages or seasons have revealed contradictory patterns. Some studies showed significant seasonal changes of the AMF community in several habitats (15, 16), while no temporal (seasonal) changes of the AMF community in roots were observed in other studies (17, 18). AMF are aerobic microbes, and the occurrence of AM symbiosis in wetlands could be associated with the development of aerenchyma in wetland plant roots (5, 12). This should allow AMF to obtain atmospheric oxygen when flooded or waterlogged and should result in a dynamic change in the AMF community through different plant physiological stages. The aim of the present study was therefore to determine the extent of AMF colonization and diversity in high-input and intensively waterlogged paddy wetlands, hypothesizing that different growth stages of rice plants vary in AMF colonization intensity, species diversity, and community structure. We measured the AMF colonization intensities at the seedling, tillering, heading, and ripening stages of rice from four paddy wetlands under continuous (>20 years) high-input and waterlogged cultivation systems and analyzed the AMF community structures in the rhizosphere soil and in rice roots collected at the different developmental stages. Knowledge of the presence of AMF in paddy wetlands as well as AMF species composition and distribution patterns could not only increase our understanding of AMF diversity in wetland habitats but also provide clues for how to enhance rice yield and to promote sustainable agriculture.

MATERIALS AND METHODS

Study site and sample collection.

Four typical paddy wetlands around the cities of Guangzhou (site HN in Huanong community in the Tianhe district, 23°17′N, 113°36′E; site BY in the Baiyun district, 23°39′N, 113°44′E) and Qingyuan (site LT in Liantang village in the town of Fogang, 23°87′N, 113°60′E; site TX in Tianxin village in the town of Fogang, 23°87′N, 113°61′E) in Guangdong Province, China, were chosen as the sampling sites. The climate in the selected sites belongs to the subtropical monsoon climate type, with approximately 1,700 to 1,800 mm of rainfall (mostly in April, May, and June) and an average temperature of 22 to 23°C over the year and of 24 to 25°C during the crop season (from early March to mid-July for early-season rice and from early August to mid-November for late rice). The investigated paddy fields (>1,000 m² each) had been under a continuous intensive rice cultivation system for more than 20 years. Briefly, rice was cultivated twice each year, with no cover crops or weeds present in winter. During the crop season, inorganic N, P, and K fertilizers were applied at a rate of approximately 400 kg ha⁻¹ annually (40% as the base fertilizer was applied before rice transplanting, and 60% as topdressing was applied at the tillering (30%) and heading (30%) stages). Pesticides (mainly organophosphorus and organochlorine pesticides) were applied when there were symptoms of pest and disease in the rice plants, with an average frequency of three to four times per crop season. The farmland soils belong to the typical acidic red soil type. To get more details on the basic properties of the farmland soils, three soil samples (0 to 15 cm in depth) from each paddy wetland were collected at the tillering stage of rice, before the topdressing application, and the soil properties including the organic matter content, total and available N and P contents, pH, and electrical conductivity were measured based on the methods described by Wang et al. (5). The basic soil properties in each paddy wetland are shown in Table S1 in the supplemental material. Each year, the fields were plowed in December and August. The rice cultivars planted during the sampling season at HN, BY, LT, and TX were Huajingxian 74, Meixiangzhan, Jingdaoyou 998, and Shengangliangyou, all of which were currently widely used Indica rice (Oryza sativa var. indica) cultivars in South China. During the crop season, the topsoil layers of the paddy fields were flooded most of the time, except that they were drained once or twice at the tillering and ripening stages (which lasted for approximately 10 days each time).

At each site, the juvenile root and soil surrounding the roots (called rhizosphere soil) of 10 individual rice plants were separately collected at the seedling (early August), tillering (early September), heading (early October), and ripening (later October) stages, yielding 40 root samples and 40 soil samples per site for a total of 160 root samples and 160 soil samples. The collected root and soil samples were kept in plastic bags and transferred to the laboratory within 24 h. Root samples were washed three times in tap water and twice in distilled water before AMF colonization was determined. Soil samples (silica gel dried) and washed root samples for DNA extraction were stored at −80°C until further analyses.

Assessment of AMF colonization intensity.

A subsample from each collected root sample was used for analysis of AMF colonization rates. Fine root samples were cleared in 10% (wt/vol) KOH at 90°C for 30 min and then stained with trypan blue. Percentage root colonization was quantified using the magnified intersection method (19), and 200 intersects of 40 root segments per root sample were scored using a compound microscope (Axiostar Plus; Carl Zeiss, Germany) at a ×400 magnification.

DNA extraction, PCR, cloning, and sequencing.

Root and soil samples from the HN and LT sites were used for analysis of AMF communities. For the root sample, 3 of 10 replicates from each rice growth stage were randomly selected, and genomic DNA was extracted from each of the selected root samples (0.1 g of fresh weight) using the cetyltrimethylammonium bromide (CTAB) method (20). For the soil samples, three random replicates collected at the vigorous growth period of rice (heading stage) were selected from HN and LT, and DNA was extracted from 0.4 g of silica gel-dried soil using a FastDNA spin kit for soil (MP Biomedicals, USA) according to the manufacturer's protocol. Both root and soil DNA extracts were used for PCR after a 1:10 dilution with distilled water.

A nested PCR was performed to amplify a fragment covering part of the small subunit (SSU), the whole internal transcribed spacer (ITS), and part of the large subunit (LSU) rRNA operon regions. This SSU-ITS-LSU region has previously proven to be suitable in field studies of AMF communities (9) and was recommended as a DNA bar-coding region for AMF (21); four primer mixtures (SSUmAf, SSUmCf, LSUmAr, and LSUmBr), each of which targets one binding site in the SSU or LSU rRNA operon and efficiently and specifically amplifies the target SSU-ITS-LSU region from all known AMF groups, were used in the nested PCR procedure (SSUmAf-LSUmAr in the first-round PCR and SSUmCf-LSUmBr in the second-round PCR) (22). The nested PCR conditions were the same as those we previously described (9). The target sequence fragment (ca. 1.5 kb) was successfully amplified from all selected root samples collected at the heading (three samples) and ripening (three samples) stages at HN and LT. At the tillering stages, AMF sequences were amplified in only 3 of the 20 root samples (2 of 10 at LT and 1 of 10 at HN), and no positive amplifications were obtained from the 20 root samples from plants at the seedling stage. The SSU-ITS-LSU fragments were successfully amplified from all six (three at HN and three at LT) selected soil samples.

PCR products with the expected length (approximately 1.5 kb) were first purified using a High Pure kit (Pearl, China) and cloned into the pMD-18T vector (TaKaRa, Japan) according to the manufacturer's protocol. They were then transformed into competent cells of Escherichia coli according to the manufacturer's instructions. PCR amplification using vector-specific primers was applied to screen for putative positive transformants, and approximately 35 positive clones were randomly selected from each sample to construct an SSU-ITS-LSU library. All selected clones from each library were then sequenced in both directions using an ABI Prism 3730XL automatic sequencer. In total, 555 positive clones from 15 SSU-ITS-LSU libraries were sequenced from the root samples, and 228 clones from 6 SSU-ITS-LSU libraries were sequenced from soil samples.

Sequence analyses and construction of phylogenetic trees.

All the obtained sequences were manually checked and edited, and the forward and reverse sequences from each clone were assembled into a consensus sequence. They were compared to the GenBank and the MaarjAM databases (23) to determine whether they could be attributed to Glomeromycota (the nomenclature of Glomeromycota applied in this study follows that developed by A. Schüßler and C. Walker [http://www.amf-phylogeny.com]), and then screened for possible chimeras according to Kohout et al. (24). Excluding the 58 and 86 nontarget or chimeric sequences from the root and soil samples, respectively, a total of 497 and 142 sequences of Glomeromycota fungi were separately obtained from the root and soil samples. Multiple alignments were performed with the online software Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/). The vector and primer sequences were excluded before further analyses.

All obtained AMF sequences were grouped into operational taxonomic units (OTUs) with sequence similarities of ≥97% using the mothur program (25). According to the intra- and interspecific variation data of AMF presented by Stockinger et al. (21), this should not result in single OTUs representing several species. A similar grouping method has also been reported in several previous studies (9, 26). A neighbor-joining (NJ) tree was constructed in MEGA, version 6 (27), using representative sequences from each OTU (derived using the mothur program) and the phylogenetic reference AMF sequences from Krüger et al. (28). The reliability of clades in the NJ analysis was assessed using nonparametric bootstrapping in MEGA (Kimura's two-parameter model [K2P]; 1,000 replicates). The AMF phylotypes were defined by a monophyletic clade approach (29), and the average pairwise distances (calculated based on the K2P model using MEGA) between different OTUs were considered in cases where it was difficult to decide whether two phylogenetically adjacent clades should be placed in the same phylotype. Two adjacent clades were separated into different phylotypes if their pairwise distances were greater than 0.055, as discussed in a previous study (9).

Statistical analysis.

Sampling efficacy was assessed with rarefaction analysis of data subsets using the mothur program (25). AMF phylotype richness (N) was calculated as the number of phylotypes recorded in each sample. The Shannon's index (H) (30) of AMF communities was calculated for each sample using the following equation: $H=\sum _{i=\hspace{0.17em}1}^n\hspace{0.17em}pi\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\ln \hspace{0.17em}\hspace{0.17em}pi$, where n is the number of phylotypes obtained in the respective sample and pi is the relative abundance of the i-th phylotype among all phylotypes in the respective sample. A parametric one-way analysis of variance (ANOVA), followed by a least significant difference test, was used to determine differences in the soil properties among the different sites and differences in the AMF colonization intensity, phylotype richness, and Shannon's index among the different rice growth stages and sampling sites. To justify the assumption of normality and homogeneity of variances, square root arcsine transformation was applied to AMF percentage colonization data before ANOVA. Two-way ANOVA was applied to analyze the effects of sampling site and growth stage and their interactions on colonization intensity, phylotype richness, and Shannon's index. As a different rice cultivar was separately planted in each site in this study, the detected effects of sampling site also include the potential effects of rice cultivar. A permutation-based nonparametric multivariate analysis of variance (PerMANOVA) (31) was carried out to investigate the effects of sampling site (in combination with rice cultivar), rice developmental stage, and sample type (soil versus root) on the community composition of AMF, using the Bray-Curtis distance as a measure of community dissimilarity; these results were graphically presented using a nonmetric multidimensional scaling (NMDS) ordination. The indicator species analysis were also performed to further test whether there were specific AMF phylotypes for rice growth stages, sampling site, and the types of samples.

Nucleotide sequence accession numbers.

All sequences obtained from this study were deposited in the GenBank under accession numbers KM207866 to KM208504.

RESULTS

Colonization intensity between AMF and rice at different growth stages.

The occurrence of AMF in rice roots was highly dependent on the growth stage of the host (Table 1). AMF structures were absent or very rare at the seedling and tillering stages at all investigated sites. At the heading and ripening stages, on the other hand, typical vesicular and arbuscular structures of AMF were commonly observed. At each site, AMF colonization intensities (including hyphal, vesicular, and arbuscular colonization rates) were significantly higher at the ripening stage than that at the heading stage (P < 0.01). At both heading and ripening stages, AMF colonization intensities at TX were higher (P < 0.01) than those at HN, BY, and LT, while the colonization rates among the last three sites were similar. The results of two-way ANOVA showed that both growth stage and sampling site (in combination with rice cultivar) had significant effects on the AMF colonization intensity (Table 2).

TABLE 1.

The colonization intensity of AMF in rice roots collected at the seedling, tillering, heading, and ripening stages from four paddy wetlands

Growth stage	Site	AMF colonization intensity by structure type (%)a
		Total	Hyphal	Vesicular	Arbuscular
Seedling	HN	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
	BY	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
	LT	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
	TX	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
Tillering	HN	0.6 ± 0.4 F	0.0 ± 0.0 C	0.5 ± 0.3 G	0.2 ± 0.2 F
	BY	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
	LT	0.4 ± 0.4 F	0.2 ± 0.2 C	0.2 ± 0.2 G	0.1 ± 0.1 F
	TX	0.0 ± 0.0 F	0.0 ± 0.0 C	0.0 ± 0.0 G	0.0 ± 0.0 F
Heading	HN	5.0 ± 0.5 E	0.9 ± 0.2 BC	3.7 ± 0.5 F	2.0 ± 0.3 D
	BY	6.2 ± 0.6 DE	1.8 ± 0.4 B	3.9 ± 0.5 EF	2.5 ± 0.4 CD
	LT	6.2 ± 0.5 DE	0.8 ± 0.3 C	3.8 ± 0.4 F	2.9 ± 0.5 CD
	TX	7.4 ± 0.6 D	0.6 ± 0.0 C	5.2 ± 0.5 DE	2.8 ± 0.4 CD
Ripening	HN	9.3 ± 0.6 C	3.3 ± 0.2 A	5.3 ± 0.5 D	3.8 ± 0.2 BC
	BY	10.1 ± 0.6 C	3.7 ± 0.6 A	7.0 ± 0.5 C	3.3 ± 0.4 BC
	LT	12.7 ± 1.1 B	3.4 ± 0.2 A	9.3 ± 0.9 B	4.5 ± 0.3 B
	TX	17.6 ± 0.8 A	4.1 ± 0.3 A	12.8 ± 0.8 A	6.8 ± 0.9 A

^aValues with different letters in the same column are significantly different at a P value of <0.05 (means ± standard errors; n = 10).

TABLE 2.

Analysis of parameter estimates from generalized linear models (univariate two-way ANOVA): effects of sampling site, growth stage, and their interactions on AMF colonization intensity, phylotype richness, and Shannon's index within each root sample

Variable(s)	Value for the parameter by AMF colonization intensity								Na		Hb
	Total		Hyphal		Vesicular		Arbuscular
	F	Pc	F	P	F	P	F	P	F	P	F	P
Site	8.6	***	1.4	NS	9.8	***	2.8	*	1.1	NS	3.8	NS
Stage	1,000	***	213	***	795	***	331	***	26.8	***	67.0	***
Site × stage	7.3	***	2.3	*	8.9	***	2.4	*	0.5	NS	4.3	*

^aN, phylotype richness.

^bH, Shannon's index.

^cNS, not significant at a P value of 0.05; *, **, ***, statistically significant at P values of 0.05, 0.01, and 0.001, respectively.

Phylogenetic analysis of AMF sequences.

The 639 AMF sequences from roots and soils were grouped into 101 OTUs based on 97 to 100% sequence similarities by the mothur program and finally assigned to 16 phylotypes based on phylogenetic analysis (Fig. 1). Among these 639 sequences, 638 sequences (15 phylotypes) could be identified at the family level, with 411 sequences (12 phylotypes) in Glomeraceae, 39 sequences (one phylotype) in Claroideoglomeraceae, and 188 sequences (two phylotypes) in Paraglomeraceae. One sequence showed no close relation to any known AMF group in the GenBank database and could not be identified at the family level. Based on the topology of the phylogenetic trees and the results of the pairwise distance calculation, the AMF phylotypes Rhi1 and Rhi6 could be identified as Rhizophagus clarus (previously referred to as Glomus clarus; GenBank accession number FM865536 to FM865544; >98.0% similarity) and Rhizophagus irregularis (JN417527 (99.9%) and HF968932 (99.9%), respectively. The AMF phylotypes Rhi5, Rhi8, and Fun1 were closely related to Rhizophagus fasciculatus (GenBank accession number FR750073; 97.4%), Rhizophagus cerebriformis (FR750095, 97.6%), and Funneliformis mosseae (FR750030; 98.3%), respectively. All of the remaining phylotypes were related to sequences not identified at the species level.

FIG 1.

Neighbor-joining phylogenetic tree showing the relationships of the representative sequences from each OTU (■) and the reference sequences from GenBank (□, sequences from the same species or genus). The DNA fragment of AMF, covering the partial SSU, ITS region, and partial LSU rRNA operon sequences (approximately 1.5 kb), were obtained from rice roots and rhizosphere soils from LT and HN paddy wetlands. The numbers of sequences obtained from the soil samples (with underlining) and root samples (without underlining) are shown to the right of the name of each defined AMF phylotype. The values above the branches are bootstrap values (1,000 replicates); only support greater than 50% is shown. ∥, sequences found only at LT; †, sequences found only at HN; &, sequences found at both LT and HN.

AMF diversity and distribution patterns in the roots and soils.

The rarefaction curves of the number of AMF sequences sampled from roots and soils at LT and HN based on OTU threshold levels of 97% sequence similarity are shown in Fig. 2. At both sites, the rarefaction curves of AMF sequences sampled from roots started, to some degree, to level off (Fig. 2A), with less than one OTU included following additional sampling of 10 sequences, indicating that the sequencing efforts detected a large proportion of AMF diversity in analyzed roots from both sites. However, the rarefaction curve from the soil samples showed that more OTUs could be expected from further sequencing (Fig. 2B), implying higher AMF richness in the paddy soils than detected in this study.

FIG 2.

Rarefaction curves showing the sequencing efforts in the rice root (A) and rhizosphere soil (B) at an operational taxonomic unit (OTU) threshold level of 97% sequence similarity.

One to four AMF phylotypes were detected within each tested root sample, and a total of 12 AMF phylotypes were finally detected from all the root samples (seven phylotypes in HN and seven phylotypes in LT, with two shared phylotypes). At both sites, the AMF diversity and phylotype richness were generally similar in the rice roots of the heading stages and the ripening stages (Fig. 3). In contrast, the Shannon's index at the heading stage was significantly higher than that at the ripening stage at the HN site (P < 0.05). The AMF diversity values (based on phylotype richness and Shannon's index) at the heading and ripening stages were both significantly higher than those at the seedling and tillering stages (P < 0.05). A two-way ANOVA showed that the growth stage had significant effects on AMF phylotype richness and Shannon's index (P < 0.01), while sampling site (in combination with rice cultivar) did not (P > 0.05). For the soil samples, which were collected at the heading stage, the AMF phylotypes obtained from each individual sample ranged from two to four, and a total of four phylotypes were separately detected at LT and HN, with three shared phylotypes (Rhi4, ClaI, and Par1) (Fig. 1). In both paddy wetlands, the Shannon's index of AMF community from individual soil samples was significantly higher than that of root samples (P < 0.05).

FIG 3.

Graphical representation of the AMF phylotype richness and Shannon's index within root and soil samples in LT and HN paddy wetlands. S, T, H, and R represent root samples collected at the seedling, tillering, heading, and ripening stages, respectively. Different letters above the columns indicate significant differences at the level of 0.05. Values are means ± standard errors.

The PerMANOVA showed that the sampling site (in combination with rice cultivar) had significant effects on the community composition of root-colonizing AMF (F = 7.6, P = 0.0014) but not on soil AMF community (F = 13.6, P = 0.10), and the developmental stage of rice showed no significant role in structuring AMF communities (F = 1.1, P = 0.36). The sample type (root or soil) also significantly affected the AMF community composition (F = 14.0, P = 0.0002). The NMDS ordination showing the effects of sampling site (in combination with rice cultivar), rice developmental stage, and sample type (soil versus root) on the AMF community composition is shown in the supplemental material (see Fig. S1). The indicator species analysis indicated that two AMF phylotypes were indicators of the presence of AMF propagules at the LT site and of their absence at HN (Rhi1 and Glome3; P < 0.05). Three other phylotypes, Rhi4, ClaI, and Par1, were indicators of the presence of soilborne AMF and the absence of plant-derived propagules (P < 0.01) in the paddy wetlands.

DISCUSSION

The presence of AMF in wetland ecosystems has received increased attention in recent years (7, 9, 32, 33), but only a few field studies documenting AMF colonization have been conducted in paddy wetlands (2). In agreement with Watanarojanaporn et al. (13), who reported that AMF are commonly present in rice roots from conventional paddy wetlands in Thailand, we found that AMF colonization in rice roots is commonly present at the heading and ripening stages though absent or rare at the early growth stages. This is in contrast to studies from Italy showing that AMF were absent in roots of rice grown under a conventional cultivation system (2, 12). Several studies have demonstrated that agricultural management practices, such as cultivation intensity, fertilizer application, and water management, can severely affect AMF communities (2, 34). Since the surveyed paddy fields mentioned above were all under conventional rice cultivation systems (2, 14), management differences alone cannot explain the different results. Differences in rice cultivars and climatic conditions may be among factors that contribute to the inconsistency of the results, but further research is needed to clarify this.

In this study, the investigated paddy fields were under waterlogged conditions most of the time, and as AMF are aerobic, the occurrence of AMF in paddy wetlands could be related to the development of aerenchyma in the rice plant (35). At the heading and ripening stages, oxygen is provided by the aerenchyma in rice roots; but at the early developmental stages the aerenchyma tissue has not been well developed (35), and the AMF may be oxygen limited. Although there is evidence showing that AMF are commonly present in some types of wetlands, such as mangrove (5, 9, 32), salt marsh (8), and lakes (6, 7, 11), the diversity and species composition of AMF in wetland habitats remain poorly understood. The relatively high number of AMF phylotypes in the surveyed conventional paddy wetlands from this study indicates that the AMF diversity in a wetland ecosystem is not necessarily low (9). This can be explained by the well-developed aerenchyma in rice roots as discussed above but could also be attributed to the high ecological adaptability of some AMF species or ecotypes. So far, the functional role of AMF in wetland ecosystems remains poorly understood (5, 36), but several studies have found that the AMF colonization of wetland plants produces a wide range of benefits to AMF plant partners (5, 36, 37), suggesting that AMF are an important component in the conventional rice cultivation systems. Considering the high N and P inputs and the frequent disturbance that occur in the surveyed fields, the observed high AMF diversity indicated that some AMF groups can be adaptive to a high-input and intensively managed environment.

So far, most studies on the AMF species composition have analyzed either roots or rhizosphere soil samples (38, 39, 40). Generally speaking, the soil contains not only the AMF groups actively colonizing roots but also those inactively present in the soil (39). Therefore, the total AMF community in soil should be expected to represent a species pool from which in planta AMF communities would be recruited (39). It is noticeable that in the surveyed habitats the species composition in the root and soil were distinctly different (P < 0.01) (Fig. 1; see also Fig. S1 in the supplemental material). The indicator species analysis identified three AMF phylotypes (Rhi4, ClaI, and Par1) that were frequently found in most soil samples at both sites but were not detected in any of the root samples. Five of the six dominant phylotypes (Rhi5, Glome1, Glome2, Glome3, and Par6) in roots of LT and/or HN were also not found from the soil samples. These phylotypes may represent “rare” AMF phylotypes that may have escaped detection due to methodological limitations, such as insufficient sampling and sequencing efforts, but a more plausible explanation could be that the AMF groups colonizing roots extend only a short distance into the soil. Different AMF groups may allocate biomass differently between roots and soil (41). Considering the hypoxia and frequently disturbed environment in the paddy wetlands, the AMF groups present in these habitats could represent types that produce little external mycelium or types that quickly extend their intra- or extraradical structures and recover fast after disturbances. Also, in the paddy wetlands it may not be necessary for AMF to extend far out in the soil as the diffusion distance of the nutrient may not be limiting for the plants.

The AMF phylotypes in the soils from the HN and LT sites were very similar, with three phylotypes (Rhi4, ClaI, and Par1, accounting for 84.5% of the all obtained AMF sequences from soil) shared between the two sites although the distance between HN and LT exceeded 100 km. Such consistency in phylotype distribution strongly suggests that the AMF community assembly is not random but, rather, a result of habitat filtering (42). Interestingly, the similarity in AMF communities between the two sites was not reflected in the root-based AMF community as seen from the PerMANOVA analysis. It implied that some indigenous AMF were adapted to the soil environments of the paddy wetlands. Johnson et al. (43) proposed that the edaphic origin of AMF should be considered when managing AMF for their benefits in agriculture. However, further studies are needed to demonstrate if AMF groups obtained from paddy soils are more beneficial to rice plants grown in paddy soils than AMF of other origins. Such knowledge is essential for application of AMF in a green rice production system.

The main biotic and abiotic factors affecting the diversity of AMF within roots of terrestrial plants have been extensively reported (40, 44), while the key influencing factors of AMF diversity in wetland habitats are rarely known (7, 9). In this study, the sampling site (in combination with rice cultivar) had no significant effects on the AMF diversity (based on phylotype richness and Shannon's index) in roots of wetland rice, which should be attributed to the uniformity in the agricultural practice of the last 20 years and the high similarity in field management programs between the surveyed sites. The PerMANOVA did not reveal any effect of growth state on the community composition. In the LT paddy wetland, two AMF phylotypes (Rhi1 and Par6) were found to be the dominant AMF groups in roots throughout the tillering, heading, and ripening stages (Fig. 4), indicating that they are well adapted to the environment of rice roots throughout rice development. Phylotype richness of root-colonizing AMF significantly increased with growth stage (P < 0.001), implying that the developmental stages of rice should be considered in the exploration of AMF function in paddy wetlands. This increase in AMF richness with growth stage could simply be due to the fact that the AMF have more time to colonize the rice roots, but it could also be related to the development of aerenchyma in rice in combination with a higher mineral nutrient requirement of rice plants at the heading and ripening stages (13).

FIG 4.

Distribution pattern of AMF phylotypes in the roots collected at the tillering, heading, and ripening stages of rice in LT and HN paddy wetlands. The numbers in the parentheses indicate the number of constructed SSU-ITS-LSU libraries at the corresponding growth stage.

Future studies should focus on the functional role of AMF during the development of a sustainable rice production system, especially on potential application of AMF for improving the nutrient use efficiency in rice cultivation systems.