Molecular diversity of arbuscular mycorrhizal fungi and their distribution patterns related to host-plants and habitats in a hot and arid ecosystem, southwest China


  • Editor: Yongguan Zhu

Correspondence: Zhi-Wei Zhao, Laboratory of Conservation and Utilization for Bio-resources, Yunnan University, Kunming 650091, China. Tel./fax: +86 871 503 4799; e-mail:


The communities of arbuscular mycorrhizal fungi (AMF) colonizing the roots of Bothriochloa pertusa, Cajanus cajan and Heteropogon contortus in a fallow land (FL) and an undisturbed land (UL) were characterized. The large subunit rDNA genes of AMF from roots were amplified and cloned. A total of 2353 clones were screened by restriction fragment length polymorphism, and 428 clones were subsequently sequenced. A total of 393 AMF sequences, which were grouped into 100 operational taxonomic units, were obtained. Phylogenetic analysis revealed that the AMF sequences belonged to Glomus, Acaulospora and Scutellospora, and that Glomus was the dominant genus. Of the 393 AMF sequences, 81% were novel. The diversity of AMF colonizing the same plant species was higher in the UL than in the FL, which confirmed strongly from the molecular evidence that soil disturbance reduced AMF population and species richness. The results revealed that AMF communities were significantly different among host-plant species and between the two habitats. The similarity of AMF communities colonizing different plant species within a habitat was higher than that of the same plant species from different habitats. The molecular evidence supported our previous hypothesis based on morphological analyses that AMF communities were more influenced by habitats compared with host preference.


Arbuscular mycorrhizal fungi (AMF) are ubiquitous in natural ecosystems and form intimate symbiotic associations with the majority of terrestrial plant roots. It is well documented that AMF can promote the uptake of plant nutrition (especially phosphorus), alleviate drought stress, improve soil structure and protect plants against root pathogens and nematodes (Smith & Read, 1997; Koide & Mosse, 2004). The diversity of AMF potentially influences plant fitness, community structure, biodiversity and ecosystem variability (van der Heijden et al., 1998). Because of these beneficial effects on plant and soil performance, AMF are crucial for the reclamation and restoration of degraded ecosystems. It has been suggested that the success of ecosystem reforestation efforts is likely to depend on the establishment of mycorrhizas, and AMF should receive special attention in indigenous tree seedling production and restoration (Wubet et al., 2003a).

AMF are obligate biotrophic fungi, and they cannot be axenically cultured in the laboratory. Conventional means of studying the diversity of AMF in natural ecosystems are based on the fungal structures within plant roots and visual inspection of spore architecture in rhizosphere soils. However, hyphal morphology within roots can only test whether roots are colonized by AMF, and it is impossible to identify these intraradical structures below the family or genus level from morphological data (Merryweather & Fitter, 1998). Also, spore morphology in rhizosphere soils can only test which AMF species occur in soils. As AMF species are diverse in sporulation condition, sporulation ability and colonization ability, even spores in soils are not always active. Hence, the species and the quantity of AMF from soils cannot entirely reflect the species and the quality from the corresponding roots (Clapp et al., 1995). Molecular techniques, especially nucleic acid-based approaches, have the potential to revolutionize our understanding of AMF. With molecular techniques, the species composition of active AMF communities colonizing roots can be investigated qualitatively and quantitatively (Alkan et al., 2006; Hijri et al., 2006).

In recent years, there has been a growing demand for the re-establishment of vegetation in degraded semi-arid and arid ecosystems using AMF. This is also the case of the area of Yuanmou (101°35′–102°06′E, 25°23′–26°06′N). Yuanmou is a typical hot and arid valley in southwest China. The vegetation of Yuanmou is called valley-type savanna vegetation by plant ecologists, and is predominantly composed of grasses and bushes, with a few trees (Jin & Ou, 2000). Human activities such as overgrazing and felling have intensively disturbed the savanna vegetation and it is now facing desertification (Jin & Ou, 2000; Liu, 2003). Thus, considerable efforts are required for ecological restoration in this area. Because of the importance of AMF for vegetation re-establishment in the fragile and degraded ecosystems, it is necessary to understand comprehensively the status of AMF in this hot and arid ecosystem. Previously, we have investigated extensively the hyphal colonization, spore density, species composition and diversity of AMF in this area using morphological methods (Li et al., 2007a, b). However, the species composition of active AMF communities colonizing roots was absent in our previous studies. In addition, nothing is known about the molecular diversity of AMF in this savanna ecosystem, and very little in similar ecosystems worldwide. In this paper, we further investigate species composition and molecular diversity of active AMF communities associated with three representative plant species in the two different habitat types. Our primary purposes are to understand the species composition of AMF colonizing each plant species, and to elucidate the distribution patterns of AMF communities related to host-plants and habitats.

Materials and methods

Site description

The study was carried out in Yuanmou County, southwest China. The climate of Yuanmou is dry and hot, with an average annual temperature of 21.9 °C, a mean annual rainfall of 629 mm and an annual evaporation of 3729 mm (Liu, 2003). In the present study, three different plant species and two different habitats were chosen for survey. Of three plant species, Bothriochloa pertusa and Heteropogon contortus are representative plant species of savanna vegetation. Cajanus cajan, an introduced tropical grain legume with excellent drought tolerance, is very prevalent in this hot and arid ecosystem. Two habitats were chosen, representing a fallow land (FL) and an undisturbed land (UL), which were located adjacently in a sloping field (about 5 × 8 km). The FL was fallowed from the cultivated land for 5 years. It was recolonized by diverse indigenous grasses and herbs, and by regrowth of some residual crops. Additionally, some exotic species were transplanted into the FL. The UL is undisturbed valley-type savanna vegetation, which is dominated by grasses and bushes with rare trees.

Root sampling

Fine roots of B. pertusa, H. contortus and C. cajan were sampled from the two habitats. Five replicates were collected randomly from each of the three plant species in each habitat. The distance of five replicates was >30 m from each other. Roots were washed under running tap water to remove attached soil and debris. Samples were stored at −70 °C for subsequent assessment of root colonization and molecular analyses.

Assessment of arbuscular mycorrhizal colonization

Samples of fine roots were cleared in 10% w/v KOH and were stained with 0.5% acid fuchsin (Li et al., 2007b). Mycorrhizal colonization was quantified using the magnified intersection method (McGonigle et al., 1990).

DNA extraction from roots

The genomic DNA was extracted from roots by the improved CTAB method (Li, 2000). Aliquots (1 g) of each root sample were homogenized in liquid nitrogen and then in 1 mL of CTAB extraction buffer (2% CTAB, 20 mM EDTA, 100 mM Tris-HCl, pH 8.0, 1.4 M NaCl, and 0.2%β-mercaptoethanol). The mixture was incubated at 65 °C for 1 h, and centrifuged at 8050 g for 10 min. Then, the supernatant was mixed with an equal volume of chloroform :isoamyl alcohol mixture (24 : 1, v : v), and was centrifuged at 2360 g for 30 min. Thereafter, the upper phase was mixed with 2/3 vol. of isopropanol, DNA was precipitated at −20 °C for at least 1 h and centrifuged at 8050 g for 5 min. The pellet was rinsed with 200 μL ice-cold (−20 °C) 70% ethanol, centrifuged for 5 min, dried, and then dissolved in 50 μL sterile distilled water. The quality and quantity of DNA in the samples were checked on a 1% agarose gel, and diluted 100 times for PCR.

Nested-PCR amplification

The target region for PCR was the partial sequences of large subunit rRNA genes (LSU rDNA). A two-step procedure (nested PCR) was conducted. The first amplification was performed using the universal fungal primer pair LR1/FLR2 (van Tuinen et al., 1998; Trouvelot et al., 1999). The 20 μL reaction volume contained 2 μL 10 × PCR buffer, 0.2 mM dNTPs, 0.5 μM each primer, 1 U Taq polymerase (Takara, Japan) and 1 μL DNA template. The PCR program was as follows: initial denaturation at 94 °C for 3 min, 30 cycles of denaturation at 94 °C for 1 min, annealing at 58 °C for 1 min and extension at 72 °C for 1 min, followed by a final extension at 72 °C for 10 min. Aliquots of 3 μL were run on an agarose gel to estimate the quantity of PCR products. Then, 1 μL of this reaction was used directly as a template for the second PCR.

The second PCR was conducted using the Glomeromycetes-specific primer pair 28G1/28G2 (da Silva et al., 2006) under the above-described reaction conditions. Nested PCR products were separated by 2% agarose gel and purified with an E.Z.N.A. Gel Extraction Kit (Omega Bio-tek) according to the manufacturer's instruction.

Cloning and construction of LSU rDNA libraries

The PCR products were cloned into pMD18-T vector (Takara) and transformed into competent cells of Escherichia coli DH-5α strain following the manufacturer's protocols. Four LSU rDNA libraries were constructed from all samples of each plant species in each habitat (three of the replicates were used to construct three respective libraries, the remaining two replicates were combined as one library), and a total of 24 libraries were generated. In each library, clones containing the converted DNA fragments were selected by blue/white screening, and white colonies were picked and grown overnight at 37 °C in liquid Luria–Bertani medium (5 g of yeast extract, 10 g tryptone, 10 g NaCl, add deionized water to a final volume of 1 liter, pH 7.0) containing ampicillin (100 μg mL−1). Thereafter, these clones were identified by PCR with the primer pair 28G1/28G2.

Restriction analyses and sequencing

PCR products of approximately 100 positive clones in each library were screened by restriction fragment length polymorphism (RFLP) with the restriction enzymes HaeIII, HinfI and TaqI according to the manufacturer's protocols. RFLP gel scans were analyzed carefully with the labimage software (version 2.62) and then clustered with ntsys-pc software (version 2.11a). One clone representative of each RFLP type detected in each library was selected and sequenced in both directions using vector primers on an ABI 3730 automatic sequencer.

Sequence analyses and construction of phylogenetic tree

DNA sequences were proofread and trimmed to remove amplicon primer sequence with the seqman program from the lasergene software Package (DNAStar Inc.). The sequence data obtained were compared with those available at the public GenBank database using blast tool to determine whether sequences were derived from Glomeromycota, and then checked for chimeric sequences using ribosomal database project II (RDP-II) software (Cole et al., 2003). Because of pre-existing knowledge of the intra- and interspecific genetic variation of AMF, sequences were grouped into the respective operational taxonomic units (OTUs) using the dotur program with the sequence similarities (95–100%), before sequences were used for phylogenetic analysis (Schloss & Handelsman, 2005). The representative sequences from each OTU and some published Glomeromycota sequences were aligned using clustalw 1.83 and the alignment was edited manually using bioedit program. Phylogenetic relationships were estimated with a Bayesian approach using mrbayes version 3.0b4 (Ronquist & Huelsenbeck, 2003). Molecular evolutionary models for Bayesian analysis were determined by modeltest 3.06 (Posada & Crandall, 1998). Bayesian analysis was performed with Markov Chain Monte Carlo simulations. Four chains were run, each for 2 000 000 generations, and were sampled every 1000 generations, starting with a random tree. Bayesian topology was constructed after the exclusion of the first 10% of trees (burin), and posterior probabilities were estimated for the remaining sampled generations, using Mortierella polycepha (AF113464) as the outgroup.

Statistical analyses

To fulfill the assumption of normality and homogeneity of variances before anova, data on percentage of AM colonization were transformed by arcsine x1/2 and spore densities were transformed by log (x+1). Means given in tables were untransformed. Means were separated using the least significance difference (LSD) test at the 0.05 probability level using spss software (version 12.0). All AMF community analyses were performed using the RFLP database. The nonparametric tests of related samples were performed to infer whether or not composition of the AMF communities differed among host-plant species and between the two habitats. The relationship between AM colonization and OTUs was determined by Pearson's correlation analysis. The Shannon–Wiener diversity indices (H′) of AMF communities were calculated using the dotur program (Schloss & Handelsman, 2005). The similarity indices (θ) between AMF communities based on membership and structure were estimated by sons program (Schloss & Handelsman, 2006). A hierarchical cluster analysis using Ward's method and squared Euclidean distance was used to determine the similarity among AMF communities with respect to the number and abundance of OTUs.


Root colonization by AMF

All surveyed samples were colonized by AMF and formed typical arbuscular mycorrhizal structures. Intra- and intercellular hyphae, vesicles and arbuscules were abundant in the root tissues. The intensity of total colonization for each plant species was moderate to high, and ranged from 37% to 75% (Table 1). The spore density ranged from 878 to 1682 per 100 g soil.

Table 1.   Means for AMF colonization in plant roots in the two habitats
HabitatsPlants%RLH%RLV%RLA%RLCSD (per 100 g soil)
  1. Means followed by the different letters (a–d) in each column are significantly different according to the LSD test at the 0.05 probability level. SD, spore density.

FLB. pertusa35.47 ± 5.43 cd9.73 ± 3.04 a0.27 ± 0.27 c38.53 ± 6.18 cd1682 ± 207 a
C. cajan71.47 ± 4.07 a4.53 ± 1.22 a31.74 ± 4.25 a75.33 ± 3.65 a1430 ± 103 a
H. contortus32.13 ± 5.49 d4.93 ± 1.46 a2.80 ± 1.42 c37.07 ± 6.17 d1163 ± 266 a
ULB. pertusa55.20 ± 4.90 b6.66 ± 0.42 a9.73 ± 1.47 b58.53 ± 3.98 b878 ± 123 ab
C. cajan50.40 ± 3.47 bc6.00 ± 1.01 a16.93 ± 4.63 b53.67 ± 3.77 bc1353 ± 351 ab
H. contortus53.61 ± 6.40 b9.34 ± 2.04 a11.33 ± 1.40 b56.13 ± 4.36 b966 ± 407 b

In the FL, the percentages of root length with hyphae (%RLH), arbuscules (%RLA) and the total colonization (%RLC) varied significantly among the three plant species and showed significant differences, which were mainly driven by C. cajan. In contrast, in the UL, there were no significant differences in AMF colonization (Table 1).

The AMF colonization of the same plant species differed significantly between habitats, except the percentage of root length with vesicles (%RLV) (Table 1). For the herbaceous plants B. pertusa and H. contortus, %RLH and %RLC in the UL were higher than those in the FL. For the woody plant C. cajan, the results were contrary to the herbaceous plants.

Amplification of AMF LSU rDNA sequences from roots

Partial LSU rDNA sequences from all root samples were successfully amplified by nested PCR. The length of PCR products by nested PCR varied with the AMF genera, about 520 bp for Acaulospora and Scutellospora, and about 580 bp for Glomus. A total of 2353 clones from 24 libraries were screened by RFLP, and 428 clones were subsequently sequenced (one clone of each RFLP type in each library was sequenced). A total of 393 sequences (2267 clones) belonging to AMF were obtained, excluding non-AMF and chimeric sequences (86 clones). These non-AMF and chimeric sequences were excluded from community structure analysis as well as sequence analysis. The 393 AMF sequences were grouped into 100 OTUs based on sequence similarities (95–100%) with the dotur program. The number of AMF sequences and the number of OTUs from each plant species are shown in Table 2. One representative sequence from each OTU was deposited into GenBank under accession nos GQ149129GQ149228.

Table 2.   The number of clones, RFLP types, AMF sequences, OTUs and Shannon–Wiener diversity indices from the three plant species
HabitatsPlantsNo. clones
for RFLP
No. RFLP groupsNo. clones for
No. clones for
and chimera
No. AMF clonesNo.
  • *

    OTUs determined at the 0.05 level by dotur program. The numbers within parentheses are the corresponding data in four libraries.

FLB. pertusa395434343039571.18
(99, 99, 97, 100)(17, 7, 11, 8) (17, 7, 11, 8) (99, 99, 97,100)  
C. cajan3929191892390292.52
(98, 100, 97, 97)(36, 12, 19, 24) (34, 12, 19, 24) (96, 100, 97, 97)  
H. contortus3985353512396121.45
(99, 99, 100, 100)(28, 8, 8, 9) (26, 8, 8, 9) (97, 99, 100, 100)  
ULB. pertusa38967675823366252.38
(100, 96, 98, 95)(6, 33, 10, 18) (5, 27, 9, 17) (99, 75, 97, 95)  
C. cajan3991051051031398482.67
(100, 100, 99, 100)(37, 36, 11, 21) (35, 36, 11, 21) (99, 100, 99, 100)  
H. contortus38069695458322241.88
(95, 93, 95, 97)(29, 7, 25, 8) (18, 7, 21, 8) (65, 93, 67, 97)  

Correlation analyses

Correlation analysis indicated that the total AM colonization was weakly positively correlated with OTUs (r=0.48, P<0.05). However, there was no statistically significant relationship between spore density and OTUs.

Phylogenetic analyses

modeltest indicated that TrN+I+G was the best-fit model for the sequence data. The topology of the phylogenetic tree as well as the results of blast indicated that the 393 AM fungal sequences belonged to the genera Glomus, Acaulospora and Scutellospora. Of the 393 sequences (2267 clones), most of them appeared to represent Glomus species, only one sequence (one clone) representing Acaulospora and seven sequences (eight clones) representing Scutellospora. The phylogenetic tree showed that all Glomus sequences from the present study, as expected, grouped together (the posterior probability=1.0) (Fig. 1).

Figure 1.

 Phylogenetic relationships from mrbayes analysis for AMF colonizing the roots of the three different plant species in two different habitat types, based on TrN+I+G model, rooted tree by Morterella polyephala as the outgroup. The Bayesian posterior probabilities (≥0.5) are shown above the branches. GenBank accessions (in bold) with triangles, circles and pentagrams indicated that AMF sequences obtained from Bothriochloa pertusa, Cajanus cajan and Heteropogon contortus, respectively. Open symbols and closed symbols indicate FL and UL, respectively. RA denoted that the relative abundance of each OTU in all AMF clones (a total of 2267 AMF clones).

A comparison of our sequences with those in the NCBI database using blast revealed that most of AMF sequences obtained from the roots of the three plant species exhibited low similarity to the sequences published previously. In this study, 318 (about 81%) of the 393 sequences and 84 (about 84%) of the 100 OTUs were novel (<95% similarity with respect to other published sequences in the nucleotide NCBI database). Both the sequence similarity (<95%) to sequences of identified AMF available in GenBank and the posterior probability (<0.9) in the phylogenetic tree were too low for us to relate them precisely to morphological species.

The diversity of AMF community

The AMF diversity, expressed by the Shannon–Wiener index, was higher in the UL (H′=3.19) than that in the FL (H′=2.55). For a given plant species, the diversity of AMF communities colonizing its roots was also higher in the UL than that in the FL. The AMF diversity of the woody plant C. cajan was higher than that of the herbaceous plants (Table 2). Regardless of habitats, Shannon–Wiener indices (H′) of B. pertusa, C. cajan and H. contortus were 2.19, 3.07 and 2.29, respectively.

The distribution of AMF

The results of nonparametric tests of K-related samples showed that there was a significant difference in the composition of the six AMF communities (χ2=51.77, d.f.=5, P<0.001). The similarity of AMF composition between any two AMF communities was very low (Table 3). Only 1 OTU (1%) shared with six AMF communities, but the relative abundance of this OTU accounted for over one-sixth (16.72%) of all AMF clones (a total of 2267 AMF clones) (Fig. 1).

Table 3.   Similarity values for the comparison between any two AMF communities based on membership and structure of AMF
  1. The numbers in bold are the average similarities of three effective replicates for each plant. The numbers within parentheses are SEs. FLBP, Bothriochloa pertusa in the FL; FLCC, Cajanus cajan in the FL; FLHC, Heteropogon contortus in the FL; ULBP, B. pertusa in the UL; ULCC, C. cajan in the UL; ULHC, H. contortus in the UL.

FLBP0.35 (0.16)
FLCC0.19 (0.06)0.35 (0.10)
FLHC0.26 (0.09)0.15 (0.05)0.26 (0.17)
ULBP0.03 (0.02)0.03 (0.02)0.01 (0.01)0.41 (0.05)
ULCC0.04 (0.02)0.03 (0.02)0.03 (0.02)0.18 (0.06)0.29 (0.04)
ULHC0.08 (0.04)0.05 (0.02)0.03 (0.02)0.14 (0.06)0.16 (0.06)0.56 (0.11)

The obvious difference in the AMF community composition occurred over host-plant species. Within a habitat, the AMF composition differed significantly among the three plant species (χ2=25.32, d.f.=2, P<0.001 for the FL; χ2=17.55, d.f.=2, P<0.001 for the UL). The similarity values (θ) of AMF communities for the comparison between the B. pertusa/C. cajan, B. pertusa/H. contortus and C. cajan/H. contortus were 0.19, 0.26 and 0.15 in the FL and 0.18, 0.14 and 0.16 in the UL, respectively (Table 3). The number of shared OTUs for the three plant species was 3 in the FL, and 6 in the UL, respectively (Fig. 1). The relative abundance of the these shared OTUs varied significantly among host-plant species.

The AMF community composition also differed significantly between the two habitats. The results of nonparametric tests of two related samples showed that there was significant difference between the two habitats (Z=−2.51, P<0.05). The similarity index of AMF communities between the two habitats was very low (θ=0.07, SE=0.02) according to the analysis by sons program. There were only 9 of 100 OTUs in common between the two habitats. The proportion of each shared OTU was very different between the two habitats.

Cluster analyses of AMF communities

Cluster analysis based on the similarity of AMF communities with respect to the number and abundance of OTUs showed that the six AMF communities were divided into two distinct clusters. Each cluster contained only three AMF communities from each habitat (Fig. 2a). Similar results were also obtained in cluster analysis based on the effective replicates, in which the AMF communities associated with the plant species within the same habitat generally had a high degree of similarity (Fig. 2b). Even the fungal communities associated with the same plant species were divided into different clusters.

Figure 2.

 (a) Dendrogram of cluster analysis based on the similarity of the six AMF communities. (b) Dendrogram of cluster analysis based on the similarity of AMF communities for the effective replicates. FLBP, Bothriochloa pertusa in the FL; FLCC, Cajanus cajan in the FL; FLHC, Heteropogon contortus in the FL; ULBP, B. pertusa in the UL; ULCC, C. cajan in the UL; ULHC, H. contortus in the UL.


Primer specificity

The selection of a suitable set of primers is most crucial for the studies of AMF communities that colonize the roots. To characterize the natural community of AMF, a set of primers would be expected to amplify the majority of AMF and to exclude plant DNA and other fungi. The primer NS31/AM1 covering a c. 500-bp central fragment of the small subunit (SSU) rDNA gene has been used widely in the molecular diversity of AMF in the previous studies. However, some studies showed that sequences related to other fungi besides AMF were occasionally amplified by NS31/AM1 (Daniell et al., 2001; Rodríguez-Echeverría & Freitas, 2006). In our previous study, all sequences amplified from one sample of B. pertusa using NS31/AM1 were Ascomycetes and Basidiomycetes (data not shown).

It has been demonstrated that the LSU rDNA gene is a suitable molecular marker for clarifying taxonomic and phylogenetic relationships in Glomeromycota (da Silva et al., 2006). The extent of the variable D1–D2 region of LSU rDNA enables not only analysis of AMF biodiversity by systematic sequencing but also the design of taxon-discriminating primers, which can then be used to monitor AMF in roots from the field (Gollotte et al., 2004). A new primer pair 28G1/28G2 was designed to evaluate the phylogenetic relationship of Glomeromycota (da Silva et al., 2006), and the phylogenetic analysis based on the LSU rDNA sequences was congruent with that based on SSU rDNA obtained by Schwarzott et al. (2001). In the present study, the primer 28G1/28G2 was used to amplify the partial LSU rDNA sequences from the roots of the three plant species in the two habitats. To our knowledge, this was the first time the primer 28G1/28G2 was used to characterize the AMF communities associated with roots. In the present study, 86 (or 3.65%) of the 2353 clones were non-AM fungal and chimeric sequences, which indicated that the primer specificity of 28G1/28G2 was suitable for field-collected roots.

Phylogenetic analysis of AMF sequences

The target sequences in the present study span the D1–D2 region of LSU rDNA. This region is far more variable than any region of the SSU rDNA, and is known to provide information to estimate higher-level phylogenetic relationships (van Tuinen et al., 1998; Kjøller & Rosendahl, 2000). Taking into account the high genetic diversity of AMF among spores of the same species, and even from single spores (Kuhn et al., 2001), we treated sequences with a similarity value of >95% of the LSU rDNA as belonging to the same AM fungal type (or OTU). At present, there is no consensus on at what level of divergent sequences can be treated as the same AMF type. For the whole internal transcribed spacer (ITS)+5.8S, sequences with a similarity value of >93% were treated as the same AM fungal type (Wubet et al., 2003b), and for SSU rDNA, >97% were documented (Vandenkoornhuyse et al., 2002; Öpik et al., 2006). Based on the knowledge that the variability of LSU gene was between SSU and ITS, the sequences were treated as the same OTU at a similarity of >95% in the present study.

The topology of the phylogenetic tree as well as the results of blast indicated that the AMF sequences obtained from the roots of the three plant species were assigned to the genera Glomus, Acaulospora and Scutellospora. However, the majority of sequences and OTUs were novel (<95% similarity with respect to other published sequences in the NCBI nucleotide database). The results were consistent with some previous reports, in which the majority of AMF sequence types detected from natural ecosystems showed no sequenced relatives among described and named fungi (Husband et al., 2002; Wubet et al., 2003b; Rodríguez-Echeverría & Freitas, 2006). There are four possible reasons why most sequences are novel. Firstly, most known morphospecies of AMF have not been sequenced, and their LSU rDNA sequences are unavailable in GenBank database. Secondly, spores of the same AMF species, and even single spores, frequently contain a mixture of different ribosomal gene sequences (Kuhn et al., 2001). In addition, the number of AMF sequences available in GenBank is presently too small to interpret efficiently the inter- and intraspecific genetic variation of AMF (Wubet et al., 2003b). Thirdly, most of fungal types obtained in the field have not been isolated in culture, or may not be presently cultivatable (Fitter, 2005), for which the sequence data are unavailable. Finally, it is possible that some nonsporulating AM fungal species exist in plant roots (Clapp et al., 1995). Although these nonsporulating species can be found by molecular methods, their sequences are not homologous with any known morphospecies. For these reasons, we are uncertain whether each OTU represents a single morphospecies, or whether a morphospecies includes >1 OTU. Species designation will be possible only after sequencing of more identified morphospecies of AMF.

Glomus is the dominant genus

In the present study, the majority of sequences detected in the roots of the three plant species, belonged to the genus Glomus, which was in accordance with our previous report based on spore morphology (Li et al., 2007a). Other reports on molecular diversity of AMF also showed that Glomus was the dominant genus of AMF communities in different ecosystems (Daniell et al., 2001; Wirsel, 2004; Hijri et al., 2006). It is probable that this observation was not just caused by bias of the molecular analyses, for example by primers that neglected some non-Glomus groups of Glomeromycetes. The predominant Glomus may well reflect the real composition of AMF in this hot and arid ecosystem, from both morphological and molecular evidences.

The molecular diversity of AMF

Although it was not possible to equate OTUs directly to morphospecies, the number of OTUs should be a conservative estimate of the number of AMF morphospecies present (Vandenkoornhuyse et al., 2002). The total number of 100 OTUs obtained from the roots of the three plant species in the two habitats was far greater than our expectation, given that only about 190 AMF species have been described worldwide so far. Furthermore, the number of OTUs obtained from B. pertusa (30), C. cajan (74) and H. contortus (31) was much higher than in other reports, in which 19 OTUs were obtained from Collinsia sparsiflora (Schechter & Bruns, 2008), 22 AM fungal types from Prunus africana (Wubet et al., 2003b) and 14 from Solidage gigantean (Vallino et al., 2006).

In the present study, the molecular diversity of AMF communities based on LSU rDNA in the two habitats was found to be high (H′=2.55 for the FL, H′=3.19 for the UL, both from three plant species). High AM fungal diversity was reported in the dry Afromontane forest (H′=2.58 from one plant species based on ITS) (Wubet et al., 2003b) and in the tropical forest (H′=2.33 from two plant species based on SSU) (Husband et al., 2002). The higher AMF diversity in the present study existed potentially due to the following reasons: (1) The higher AMF diversity may reflect high colonization in all the plants surveyed. (2) High root weights (1 g) were used for DNA extraction, whereas in some studies, only several root pieces of 1–2 cm from each replicate were used for DNA extraction (Stukenbrock & Rosendahl, 2005; Vallino et al., 2006). Generally, more roots may obtain more sequence types of AMF that associated with roots. (3) There were a large number of clones for RFLP analysis (approximately 400 clones for a plant species in each habitat). It has been suggested that AMF richness increases with sample size and number of clones screened within an ecosystem (Öpik et al., 2003). Our results indicated that the molecular diversity of AMF was far higher than expected, supporting the hypothesis that in semi-natural grassland ecosystems the AM fungal diversity is particularly high (Vandenkoornhuyse et al., 2002).

Patterns of AMF distribution related to host-plant species and habitats

Our results clearly demonstrated that the three plant species within a habitat were colonized by significantly different AMF populations. Similar results were also found in other reports, in which the coexisting plant species were associated with divergent AMF communities (Vandenkoornhuyse et al., 2002; Gollotte et al., 2004). Furthermore, different abundances of the shared OTUs in the roots of the three plant species within a habitat were detected, indicating that there was a preference between host-plant species and some AMF (Vandenkoornhuyse et al., 2002; Gustafson & Casper, 2006). Schechter & Bruns (2008) concluded that host preference could have a strong influence on AMF composition.

The composition of the AMF community also changed drastically between the two habitats. Even in the same plant species, the composition and the dominant taxa of AMF colonizing its roots differed significantly with habitats. The large difference between the AMF communities in the two habitats might be caused by soil factors, land-use types and plant community composition. Firstly, soil factors, such as pH, nutrient content, total soil C and N, moisture and temperature, are known to influence AM fungal spore distribution (Husband et al., 2002). In the area, organic C, total N and total P of the UL were higher than those of the FL, but available P was contrary (Li et al., 2007b). Secondly, it was shown that the AMF communities changed with land use intensity (Oehl et al., 2003). In the present study, the FL was fallowed from the cultivated land for 5 years. The UL has never been cultivated. Our previous reports based on spore morphology showed that the AMF community composition varied considerably across different land-use types in this hot and arid ecosystem (Li et al., 2007a). Thirdly, it is known that the plant community structure affects diversity and community composition of AMF (Johnson et al., 2003). In the present study, there was an obvious difference in plant community composition between the two habitats.

In the present study, both the host-plant species and the habitat affected the AMF community composition significantly. Which one had a greater effect on AMF community? Based on the results in the present study, for a given host-plant species, the similarity value of AMF communities between the two habitats was less than that between it and the other plant species within a habitat, for example, for B. pertusa, the similarity value between the two habitats was 0.03, while the similarity value between B. pertusa/C. cajan and B. pertusa/H. contortus within the FL was 0.19 and 0.26, respectively (Table 3). Cluster analyses also indicated that the similarity of AMF communities colonizing different plant species within a habitat was higher than that colonizing the same plant species from different habitats (Fig. 2). These results from the molecular level further supported our previous hypothesis based on morphological evidences that AMF communities were more influenced by habitats compared with host preference (Li et al., 2007a).


We would like to thank Dan-Dan Zhao, Kai Wang and Chang-Cong Liang for helping to collect samples. We also thank Tao Sha, Prof. Hang-Bo Zhang, Prof. Li Yu, Dr Bao-Yu Tian and Dr Wei Wang for their help with the phylogenetic analysis. We specially thank the anonymous referees for their valuable suggestions and comments on our manuscript. This study was financially supported by the National Natural Science Foundation of China (NSFC30770052, NSFC40763003 and NSFC30900963), and the National Basic Research Program of China (973 Program 2007CB411600).

Authors' contribution

L.-F.L. and T.L. contributed equally to this work.