Direct and indirect influences of 8 yr of nitrogen and phosphorus fertilization on Glomeromycota in an alpine meadow ecosystem


Authors for correspondence:
Huyuan Feng
Tel: +86 931 8912537
Lizhe An
Tel: +86 931 8912560


  • We measured the influences of soil fertility and plant community composition on Glomeromycota, and tested the prediction of the functional equilibrium hypothesis that increased availability of soil resources will reduce the abundance of arbuscular mycorrhizal (AM) fungi.
  • Communities of plants and AM fungi were measured in mixed roots and in Elymus nutans roots across an experimental fertilization gradient in an alpine meadow on the Tibetan Plateau.
  • As predicted, fertilization reduced the abundance of Glomeromycota as well as the species richness of plants and AM fungi. The response of the glomeromycotan community was strongly linked to the plant community shift towards dominance by Elymus nutans. A reduction in the extraradical hyphae of AM fungi was associated with both the changes in soil factors and shifts in the plant community composition that were caused by fertilization.
  • Our findings highlight the importance of soil fertility in regulating both plant and glomeromycotan communities, and emphasize that high fertilizer inputs can reduce the biodiversity of plants and AM fungi, and influence the sustainability of ecosystems.


Arbuscular mycorrhizal (AM) fungi in the Phylum Glomeromycota generally form mutualistic associations with the roots of most plants in natural conditions (Smith & Read, 2008). It is well accepted that AM fungi can increase soil resource capture as well as improve the stress tolerance of their hosts in return for plant photosynthates (Smith & Read, 2008). Moreover, AM fungi can exert strong effects on plant communities (Hartnett & Wilson, 1999; Vogelsang et al., 2006) and rhizosphere microorganisms (Zhang et al., 2010) and, consequently, influence many ecosystem processes (Rillig, 2004). Although their physiological benefits and ecological importance are well established, the ecological determinants that regulate the abundance and composition of AM fungal communities are not well understood.

It has been suggested that glomeromycotan communities are structured by the composition of plant communities (Johnson et al., 2004; Börstler et al., 2006; Hausmann & Hawkes, 2009) and soil factors, such as pH (Dumbrell et al., 2010) and the relative availability of soil resources (Egerton-Warburton et al., 2007; Alguacil et al., 2010). Some field studies of environmental gradients have reported a positive correlation between the species richness of plants and AM fungi (Landis et al., 2004; Wu et al., 2007), whereas others have shown no relationship (Börstler et al., 2006; Öpik et al., 2008) or a negative relationship (Antoninka et al., 2011). Such inconsistent observations suggest that communities of plants and AM fungi do not always respond similarly to environmental variation.

The tight linkage between mycorrhizas and plant nutrition suggests that AM fungi may be very sensitive to changes in the availability of soil resources. The functional equilibrium model predicts that the enrichment of soil resources by fertilization will reduce plant allocation to roots (Brouwer, 1983) and mycorrhizas (Johnson et al., 2003) because, once soil resource limitation is eliminated, plant competition for above-ground resources (i.e. light; Hautier et al., 2009) will become stronger, so that plants should allocate more biomass to shoots and leaves rather than below ground (Johnson, 2010). Empirical field studies support these predictions by showing that fertilization reduces significantly the biomass of plant roots (Bloom et al., 1985) and AM fungi (Johnson et al., 2003, 2008). Furthermore, fertilization has been shown to influence the species composition and reduce the species richness of plants (Rajaniemi, 2002; Dickson & Foster, 2011) and AM fungi (Egerton-Warburton et al., 2007; Alguacil et al., 2010). However, none of these studies has measured simultaneously the effects of fertilization on communities of both plants and AM fungi. In order to understand the mechanisms by which soil fertility influences the abundance, diversity and species composition of AM fungi, the dynamics of both plant and glomeromycotan communities must be measured synchronously across a soil nutrient gradient, and the proportions of variance attributed to changes in soil and plant properties should be differentiated. This type of analysis can be accomplished using a structural equation model (SEM) which quantifies the relative strength of relationships among variables in complex systems and tests for hypothesized causal relationships among the variables (Grace, 2006).

Fertilization with both nitrogen (N) and phosphorus (P) often reduces AM fungal abundance (Bååth & Spokes, 1989; Sylvia & Neal, 1990). This could result from direct fungal responses to increased availability of soil N and P, as well as changes in soil pH; however, AM fungi may also respond indirectly to fertilization through changes in the plant community. As AM fungi are obligate biotrophs, a decrease in root substrate can be expected to reduce glomeromycotan abundance. Alternatively, shifts in the plant community composition towards plant species that support fewer mycorrhizas (Johnson et al., 2008) could account for the reduction in AM fungal extraradical hyphae caused by nutrient enrichment. Fertilization is also expected to change the composition of AM fungal communities and, again, this may occur by direct and indirect mechanisms. Glomeromycotan taxa that are the most aggressive colonists in a high-fertility environment may be favored directly by fertilization (Douds & Schenck, 1990; Johnson, 1993). Alternatively, AM fungi may change indirectly in conjunction with the changing composition of the host plant community. It is important to stress that direct and indirect mechanisms are not mutually exclusive.

We studied simultaneously the responses of plants and AM fungi to 8 yr of fertilizer treatments to test the following hypotheses: H1, fertilization will reduce AM fungal abundance (functional equilibrium hypothesis); H2, fertilization will reduce the diversity of both plant and glomeromycotan communities; H3, fertilization will have both direct and indirect (i.e. plant-mediated) effects on AM fungal abundance and community composition; H4, the predicted effects of fertilization on Glomeromycota will be manifested at both a whole-community scale and at the scale of an individual plant species. Glomeromycotan community composition, root colonization and extraradical hyphae were measured in mixed root samples and in roots of Elymus nutans (Poaceae), the most competitive plant species. The analysis of AM fungi in mixed roots describes mycorrhizal responses to fertilization at the whole-community level, whereas the analysis of E. nutans roots reveals mycorrhizal responses that are independent of plant community changes.

Materials and Methods

Site description and experimental design

This study was conducted at the Walaka experimental site of the Research Station of Alpine Meadow and Wetland Ecosystems of Lanzhou University, which is located in the eastern Tibetan Plateau of China (35°58′N, 101°53′E; 3500 m above sea level). The climate of this region is humid-alpine with a mean annual temperature of 1.2°C (11.7°C maximum in July and − 11°C minimum in January) and a mean annual frost period of c. 270 d. The mean annual precipitation is 620 mm (most in summer) and the mean annual cloud-free solar radiation is c. 2580 h. The subalpine meadow soil has relatively low P availability (< 2 mg available P kg−1 dry soil). The native vegetation consists mainly of Arctic alpine and Chinese Himalayan plants, and is dominated by Kobresia setchwanensis and some grasses, such as E. nutans (Supporting Information Table S1 provides a full species list). The experimental site (c. 8 ha) has been overgrazed in the past, but has been fenced and only grazed in winter and early spring (October to April in the following year) since 2001.

The long-term fertilization experiment was established on a flat field in the Walaka experimental site on March 2002. Twenty-five 6 × 10-m2 plots composed of five fertilization levels with five replicates were distributed in five columns and five rows with a randomized block design (Fig. S1). Each plot was separated from the others by a 1-m buffer strip. The fertilization treatment was generated with different amounts of (NH4)2HPO4 fertilizer applied annually from 2002 at the beginning of the growing season (usually in the middle of May). Fertilizer applications of 0, 30, 60, 90 and 120 g m−2 yr−1 are hereafter referred to as F0, F30, F60, F90 and F120, respectively. The corresponding N and P inputs of the five fertilization levels are as follows: F0, control; F30, 6.4 g N and 7 g P m−2 yr−1; F60, 12.7 g N and 14.1 g P m−2 yr−1; F90, 19.1 g N and 21.1 g P m−2 yr−1; F120, 25.4 g N and 28.2 g P m−2 yr−1. The original goal of this experiment was to determine the fertilizer application rate that would generate maximal grass yield, and also to test the relationship between productivity and diversity and other ecological questions (e.g. Luo et al., 2006; Yang, 2011).

Sampling procedure and plant analysis

Samples were collected on 20 May, 10 July and 5 September 2010. Nine soil cores (diameter, 3.8 cm; depth, 25 cm) were taken randomly from each plot, mixed adequately as one sample in a sealed bag and transported in an ice box to the laboratory within 36 h. The 25 experimental plots were sampled during three sampling dates for a total of 75 samples. Fine live roots were separated carefully from each soil sample, washed cleanly and used for DNA extraction and the determination of AM fungal colonization; the remaining soil was air dried, sieved (1 mm) and used for spore isolation and soil chemical analyses. To analyze the colonization and composition of the AM fungal community in the roots of single plant species, we sampled a dominant plant species E. nutans at the second sampling date. Three individuals of E. nutans were chosen randomly from each plot, excavated and pooled as one sample. A subsample of E. nutans fine roots was separated from each sample and used for DNA extraction and the determination of AM fungal colonization. The shoots and remaining roots of each E. nutans sample were separated and dried (80°C for 48 h) for the determination of tissue N and P contents of E. nutans using the methods described below.

The above-ground vegetation was sampled in two 0.25-m2 quadrats in each plot at the end of August 2010. Species richness and abundance (based on the number of individuals per species; we regarded a ramet as an individual for the clonal species) were estimated, and individual plants in each quadrat were clipped to the soil surface. The height of E. nutans individuals was measured on three randomly selected individuals in each quadrat. Concurrent with the above-ground vegetation sampling, 13 soil cores (diameter, 3.8 cm; depth, 25 cm) were taken randomly from each plot, and the roots were removed using a 1-mm sieve and washed cleanly for the measurement of root biomass. All shoot and root samples were dried at 80°C for 48 h, and weighed. The dry plant samples were then ground to a fine powder for the determination of the tissue N and P contents.

Analyses of AM fungal colonization, extraradical hyphae and spores

Root samples were stained with trypan blue and the percentage root length colonized by Glomeromycota (%RLC) and the percentage colonized by arbuscules (%AC) were quantified using the magnified intersection method (McGonigle et al., 1990). Extraradical hyphae of each soil sample were extracted and stained with trypan blue using the methods of Brundrett et al. (1994). Hyphae of AM fungi were distinguished from other fungal hyphae at 200× magnification according to morphology (Miller et al., 1995) and staining color, and the hyphal length was measured by a line intersection method and changed to the hyphal length density (m g−1 dry soil) (Brundrett et al., 1994). Spores of AM fungi were separated from 50 g of air-dried soils by wet sieving and sucrose centrifugation (Brundrett et al., 1994). A dissecting microscope was used to sort spores according to morphology and color, and the abundance of each morphotype was recorded. Morphological identification of spore morphotypes followed Liu et al. (2009).

Molecular analysis of AM fungi colonizing roots

One hundred root samples, consisting of 75 mixed and 25 E. nutans samples, were used for molecular analysis. Root DNA was extracted using a Plant DNA Extraction Kit following the manufacturer’s protocol (Tiangen Biotech, Beijing, China). The extracted DNA was diluted with double-distilled H2O (1 : 10) and subjected to a nested PCR with a first primer combination of GeoA2–Geo11 and a second glomeromycotan-specific primer pair NS31–AML2 (Liu et al., 2011). The first PCR was carried out in a final volume of 25 μl with 2 μl extracted DNA dilution and 0.5 μM of each primer using a Taq PCR Kit (New England Biolabs, Ipswich, MA, USA) with the following cycling conditions: 94°C for 2 min, 30 × (94°C for 30 s, 59°C for 1 min and 72°C for 2 min) and 72°C for 10 min. The first amplification product was diluted with double-distilled H2O (1 : 100) and 2 μl of this dilution was used as a template for the second PCR amplification. The second PCR conditions were the same as those of the first PCR, but with the following cycles: 94°C for 2 min, 30 × (94°C for 30 s, 58°C for 1 min and 72°C for 1 min) and 72°C for 10 min. All PCR products were examined on 1.5% (w/v) agarose gels with ethidium bromide staining to confirm the product integrity. The second PCR products were purified using the Gel and PCR Clear Up System (Promega), and the expected DNA fragments (c. 560 bp) were obtained. Purified DNA fragments were ligated into pGEM-T vector (Promega) and cloned into Escherichia coli DH5α according to the manufacturer’s recommended protocol, resulting in 100 clone libraries. For each clone library, 48 putative positive transformants were picked randomly and immersed in 30 μl of double-distilled H2O, and subjected to three cycles of freezing and thawing for the preparation of plasmid templates (Liu et al., 2011). Inserts were re-amplified with primers NS31–AML2 using the same conditions as in the second PCR. A total of 4452 positive PCR products was then screened using restriction fragment length polymorphism (RFLP) with restriction enzymes HinfI and Hin1II (Fermentas, Vilnius, Lithuania). RFLP patterns were only compared within the same sample group (samples from plots of the same fertilization level at the same sampling time, for example, five samples from control plots at the first sampling time, were considered as a sample group). One representative clone of each RFLP type in each sample group was sequenced by the Major Biotech Company (Shanghai, China) using the vector primer T7, for a total of 468 sequences. All DNA sequences were edited and compared with public databases using BLAST (Altschul et al., 1997), and the non-glomeromycotan and, possibly, chimeric sequences were eliminated from the dataset. The remaining 352 sequences of AM fungi were submitted to the GenBank database under the accession numbers JN009130JN009481. All glomeromycotan sequences were aligned using ClustalW, and clustered to species-level groups according to 97% sequence similarity using the furthest neighbor algorithm in the Mothur program (Schloss et al., 2009). Each species-level group was regarded as an AM fungal phylotype. To elucidate the phylogenetic relationships between the AM fungal phylotypes obtained in this study and the published AM fungal sequences, up to three representative sequences from each phylotype, the most closely related sequences from GenBank and the representative sequences of major families of Glomeromycota were used in the Bayesian phylogenetic analysis according to the method described elsewhere (Liu et al., 2011).

Soil properties, plant N and P concentrations

The soil moisture content was measured gravimetrically (dried at 105°C for 12 h), and the soil pH was measured in 1 M KCl (1 : 5 w/v). Soil organic carbon (C) and total N concentrations were analyzed using the CHNS-analyzer system (Elementar Analysensysteme GmbH, Hanau, Germany) with the burning method at 450 and 1250°C, respectively. Soil available P was extracted using a Mehlich-3 extractant and measured using the molybdate-blue colorimetric method (Mehlich, 1984). Soil available N (NO3-N + NH4-N) concentrations were analyzed using a FIAstar 5000 Analyzer (FOSS, Hillerød, Denmark). The N content of shoots and roots was measured using the same method as employed to measure soil total N. The P content of shoots and roots was measured using the molybdate-blue colorimetric method after digestion with sulfuric acid (Lu, 1999).

Statistical analysis

The matrix of plant community composition was calculated using the mean number of individuals or ramets of each species in two sampling quadrats in each plot, and the community composition of AM fungi was analyzed on the basis of the clone numbers of each phylotype in a root sample. The shoot biomass of the plant community was calculated from the mean weight of above-ground plants in two quadrats. The raw data that were measured at three sampling times (including soil characteristics, %RLC, %AC, extraradical hyphal length density, spore density, spore community and the AM fungal community in mixed roots) were pooled and calculated, using the means to represent the status of each variable during the whole growing season.

All statistical analyses were carried out using R version 2.12.2 (R Development Core Team, 2011) or SPSS 13.0 (SPSS Inc., Chicago, IL, USA). Before analysis, all data were tested for normality; only the soil N : P data did not meet the normality distribution, and so these data were loge(1 + x) transformed; moreover, raw community data for plants and AM fungi were square root transformed in order to downweigh the importance of abundant species or phylotypes. The effects of the fertilization treatment on the soil characteristics, AM fungal colonization, extraradical hyphal length density, species richness of plants and AM fungi, and plant N and P contents were tested by one-way ANOVA. To assess the responses of plants and AM fungi to the N and P availability and soil pH, which might be changed significantly by fertilization treatments, we first generated proxies of N and P availability by principal component analysis (PCA), using the scores of the first principal component (PC1) of data matrices which included available soil N, shoot N and root N or available soil P, shoot P and root P (hereafter referred to as N or P availability score). We used the combined plant and soil nutrient information, rather than the available soil N or P alone, to define the N or P availability because the plant nutrient status is important in determining mycorrhizal fungi, and this method has been employed previously (e.g. Cox et al. (2010) using foliar N, root N and soil solution nitrate to define the N availability). Linear regression analysis was performed using the plant or AM fungal variables against the N and P availability scores and soil pH.

To elucidate the influence of fertilization treatments on the community composition, all community matrices (including communities of AM fungi in mixed and E. nutans roots, spore communities and plant communities) were ordinated (two-dimensional solution) using nonmetric multidimensional scaling (NMDS) with the Bray–Curtis dissimilarity measurement. To explore the correlations between each community and the corresponding environmental variables, soil and/or plant variables were fitted as vectors onto the NMDS plots using the function ‘envfit’ from the ‘vegan’ library of the R package (Oksanen, 2011). The relationships between any two communities (i.e. spore and plant communities) or between a community matrix and a matrix of N or P variables were measured using Mantel tests, in which the matrices of community structure and N or P variables were represented by the Bray–Curtis distance and Euclidean distance, respectively (Oksanen, 2011). In addition, to determine AM fungal indicator species for each fertilization treatment, we conducted indicator species analyses (species with Indval values  0.6 are strong indicators) for each AM fungal community using the function ‘indval’ from the ‘labdsv’ library of the R package (Roberts, 2010).

To determine the direct and indirect effects of fertilization on AM fungi, we used AMOS 5.0 (SPSS Inc.) to construct and test a SEM. In this model, we used AM fungal extraradical hyphal length density and the root : shoot biomass ratio to represent the status of glomeromycotan abundance and plant allocation to roots, respectively; the plant and glomeromycotan communities were represented by the scores of the first component of PCA of the community composition of plants and AM fungi, respectively. We tested how well the model fitted the data using the maximum likelihood χ2 goodness-of-fit test, Jöreskog’s goodness-of-fit index (GFI) and the Bollen–Stine bootstrap test (in these tests, high P values indicate that the data fit the model well).


Relationship between fertilization treatments and N and P availability

Eight years of fertilization increased significantly the soil available P concentration and reduced the soil N : P ratio, and, in all but the F30 treatment, increased significantly the soil available N concentration compared with the unfertilized control (Table 1). Fertilization treatment did not change significantly the total N (F4 = 0.888, = 0.489), organic C (F4 = 0.398, = 0.808) or soil moisture (F4 = 0.061, = 0.993); moreover, the soil pH decreased steadily with increasing fertilizer inputs from 7.01 to 6.44 (Table 1), even though the soil pH was not changed significantly by fertilization treatment (F4 = 2.497, = 0.075). Plant tissue N and P contents at both community and single-species scales responded differently to different levels of fertilization, but it was clear that higher tissue N and P contents (except the shoot N) and a lower tissue N : P ratio occurred in plots with higher fertilizer inputs (F60, F90, F120) compared with control plots (Table S2). The ordinations of N and P matrices for both mixed root samples (community-level analysis) and E. nutans samples (individual-level analysis) successfully explained the variation of N and P availability among treatments (Fig. S2): for the community-level analysis, the PC1 of N and P matrices explained 68.1% and 78.4% of the variation in N and P availability, respectively; for the E. nutans analysis, the PC1 of N and P matrices explained 61% and 83.8% of the variation of N and P availability, respectively. The loading values of PC1 of all matrices were positive (Fig. S2), suggesting that the PC1 scores increased with increasing N or P availability. In addition, we found that the relationships between each plant and soil N or P variable (shoot, root and soil N or P concentrations) and plant and AM fungal species richness, or AM fungal extraradical hyphal length density, were similar to the results derived from scores of N or P availability (data not shown). These results indicate that the PC1 scores are excellent proxies for N and P availability in our case.

Table 1.   Soil characteristics in different fertilization treatments
 Total N (%)Organic C (%)Available N (mg kg−1)Available P (mg kg−1)Available N : available PMoisture (%)pH
  1. Data are means ± SE (= 5). F0, F30, F60, F90 and F120 represent (NH4)2HPO4 fertilizer applications of 0, 30, 60, 90 and 120 g m−2 yr−1, respectively. Significant differences across treatments within each variable were determined using Tukey’s honestly significant difference (HSD) test ( 0.05) after one-way ANOVA and are indicated by dissimilar letters.

F00.3 ± 0.011.7 ± 0.0617.6 ± 1.41c2.8 ± 0.37e6.8 ± 1.03a30.0 ± 1.647.0 ± 0.13
F300.3 ± 0.011.6 ± 0.0321.7 ± 1.50bc42.4 ± 2.34d0.5 ± 0.04b29.8 ± 0.616.9 ± 0.14
F600.3 ± 0.011.7 ± 0.0533.3 ± 2.10b79.1 ± 3.14c0.4 ± 0.03b30.1 ± 0.676.8 ± 0.07
F900.3 ± 0.011.7 ± 0.0855.6 ± 4.27a145.1 ± 8.34b0.4 ± 0.03b30.1 ± 1.046.7 ± 0.10
F1200.3 ± 0.011.7 ± 0.0547.1 ± 3.51a189.3 ± 4.99a0.3 ± 0.01b30.6 ± 1.166.4 ± 0.20

Plant community responses to fertilization

Fertilization reduced significantly the species richness (F= 70.96, < 0.001) and individual plant density (F4 = 84.20, < 0.001), and increased the shoot biomass (F4 = 6.20, = 0.002; Table 2). Although the root : shoot biomass ratio decreased by > 30% across the fertilization gradient, the root biomass (F4 = 1.35, = 0.288) and root : shoot biomass ratio (F4 = 2.48, = 0.077) did not differ significantly among treatments (Table 2). Most species, particularly the sedges and leguminous species, were excluded in plots with higher fertilizer inputs (i.e. F60 to F120 plots; Table S1), whereas E. nutans became increasingly dominant (F4 = 59.68, < 0.001) and taller (F4 = 26.61, < 0.001) with fertilization (Table 2). Linear regression analysis showed that plant species richness was negatively correlated with the scores of both N and P availability, and positively related to soil pH (Fig. 1a). In addition, the root : shoot biomass ratio was negatively related to the scores of P availability (r2 = 0.218, = 0.01), but not N availability (r2 = 0.07, = 0.114). NMDS ordination of plant communities yielded a continuum of points from low to high fertilization along the first axis, and the soil available P (r2 = 0.913, < 0.001) and available N (r2 = 0.665, < 0.001) were very important in structuring the plant communities (Fig. 2a).

Table 2.   Changes in plant properties for whole plant communities and for Elymus nutans after long-term fertilization treatment
 Plant community levelElymus nutans
Species richnessDensity (individuals 0.25 m−2)Shoot biomass (kg m−2)Root biomass (kg m−2)Root : shoot biomass ratioRelative abundance (%)Height (cm)
  1. Data are means ± SE (= 5). F0, F30, F60, F90 and F120 represent (NH4)2HPO4 fertilizer applications of 0, 30, 60, 90 and 120 g m−2 yr−1, respectively. Significant differences across treatments within each variable were determined using Tukey’s honestly significant difference (HSD) test ( 0.05) after one-way ANOVA and are indicated by dissimilar letters.

F027.6 ± 1.0a183.0 ± 9.3a0.5 ± 0.03b0.2 ± 0.030.5 ± 0.079.4 ± 1.0d57.7 ± 3.7c
F3016.9 ± 1.9b131.1 ± 11.3b0.7 ± 0.05ab0.2 ± 0.040.3 ± 0.0414.8 ± 1.8cd75.2 ± 3.3b
F609.8 ± 1.0c64.2 ± 6.2c0.8 ± 0.05a0.3 ± 0.030.3 ± 0.0327.0 ± 3.0c89.0 ± 2.2a
F904.9 ± 0.5d28.7 ± 2.3d0.9 ± 0.02a0.3 ± 0.040.3 ± 0.0551.3 ± 5.1b99.2 ± 2.2a
F1204.4 ± 0.9d27.8 ± 4.2d0.8 ± 0.10a0.2 ± 0.010.3 ± 0.0470.8 ± 4.1a90.2 ± 3.5a
Figure 1.

Linear regressions of plant species richness (a), extraradical hyphal length density of arbuscular mycorrhizal (AM) fungi (b) and AM fungal phylotype richness in mixed roots (c) and Elymus nutans roots (d) versus nitrogen (N) and phosphorus (P) availability scores and soil pH. N and P availability scores were derived from the first principal component of an N matrix (including soil available N, shoot N and root N) and a P matrix (including soil available P, shoot P and root P), respectively (see Materials and Methods). N and P availability scores increase with increasing fertilizer inputs (see Fig. S2).

Figure 2.

Nonmetric multidimensional scaling (NMDS) patterns of community compositions of plants (a), arbuscular mycorrhizal (AM) fungal spores (b), AM fungi colonizing mixed roots (c) and AM fungi colonizing Elymus nutans roots (d); points are marked according to fertilization treatments. The stress value reflects how well the ordination summarizes the observed distances among samples (lower than 20% can be ecologically interpretable and useful). Soil and/or plant variables are fitted as vectors onto each ordination plot (for (a) all vectors are shown; for (b)–(d), only seven vectors with higher r2 values are shown). Soil AN, soil available N; Soil AP, soil available P.

Effects of fertilization on intra- and extraradical hyphae and spore density of AM fungi

High levels (F90 and F120) of fertilization reduced significantly %RLC and %AC in both mixed and E. nutans roots, whereas a low level (F30) of fertilization did not reduce or even increased (%RLC in mixed roots) root colonization by AM fungi in both root systems (Fig. 3a,b). In addition, few or even no arbuscules were observed in roots from most F90 and F120 plots. The extraradical hyphal length density of AM fungi decreased gradually across the fertilization gradient (Fig. 3c), and responded negatively to the scores of both N and P availability, and positively to soil pH (Fig. 1b). The density of AM fungal spores peaked in soils with the F30 treatment, whereas the spore densities in other fertilized soils were similar to the control (Fig. 3d); moreover, the spore density did not show significant correlations with N (r2 = 0.039, = 0.172) and P (r2 = 0.01, = 0. 627) availability scores or soil pH (r2 = 0.103, = 0. 07).

Figure 3.

The root length colonization (%RLC; a) and arbuscular colonization (%AC; b) by arbuscular mycorrhizal (AM) fungi in mixed and Elymus nutans roots, AM fungal extraradical hyphal length density (c) and spore density (d) varied across a fertilization gradient. F0, F30, F60, F90 and F120 represent (NH4)2HPO4 fertilizer applications of 0, 30, 60, 90 and 120 g m−2 yr−1, respectively. Bars represent means ± SE (= 5). Significant differences between columns (for (a) and (b), comparison between columns of the same root system) were determined using Tukey’s honestly significant difference (HSD) test ( 0.05), and are indicated by dissimilar letters above the bars.

Morphological and molecular identification of AM fungi

Eleven AM fungal spore morphotypes were found in this study (Table 3). Species accumulation curves of spore morphotypes (Fig. S3a), based on individual spore counts, indicated that a large proportion of the total AM fungal spore diversity was captured for all treatments. All spore morphotypes, except Glomus macrocarpum, were present in all treatments, and the most abundant species was Diversispora spurca (Table 3).

Table 3.   Compositions of spore communities of arbuscular mycorrhizal (AM) fungi across the experimental fertilization gradient
 Glomus mosseaeGlomus constrictumGlomus etunicatumGlomus macrocarpumGlomus sp.Acaulospora tuberculataAcaulospora scrobiculataScutellospora pellucidaScutellospora dipurpurescensPacispora scintillansDiversispora spurcaSpecies richness
  1. Data (means ± SE, = 5) are spore numbers per 50 g dry soils. F0, F30, F60, F90 and F120 represent (NH4)2HPO4 fertilizer applications of 0, 30, 60, 90 and 120 g m−2 yr−1, respectively. Significant differences of the abundance of the same species (column) across different treatments were determined using Tukey’s honestly significant difference (HSD) test ( 0.05) after one-way ANOVA and are indicated by dissimilar letters.

F08.7 ± 1.4a10 ± 0.3b3.7 ± 0.2b00.9 ± 0.24.8 ± 1.2bc2.7 ± 0.94.1 ± 1.56.7 ± 1.9a0.1 ± 0.1c15.9 ± 3.49.2 ± 0.2
F308.9 ± 1.7a23.6 ± 3.2a15.0 ± 1.2a0.1 ± 0.10.5 ± 0.210.0 ± 0.3a5.2 ± 0.52.0 ± 0.54.3 ± 0.4ab0.1 ± 0.1c24.6 ± 4.69.2 ± 0.5
F606.8 ± 0.9ab9.7 ± 1.3b7.9 ± 1.4b0.1 ± 0.10.5 ± 0.210.5 ± 0.8a3.7 ± 1.11.9 ± 0.43.3 ± 0.3ab1.1 ± 0.2ab28.7 ± 6.09.8 ± 0.2
F907.9 ± 2.1ab11.9 ± 1.1b7.6 ± 1.2b00.8 ± 0.37.9 ± 1.4ab3.1 ± 0.52.2 ± 0.62.7 ± 0.6b1.3 ± 0.3a24.5 ± 3.69.6 ± 0.2
F1202.3 ± 0.7b7.3 ± 1.3b4.7 ± 0.9b00.5 ± 0.23.7 ± 0.5c2.7 ± 0.21.5 ± 0.31.8 ± 0.5b0.3 ± 0.2bc28.4 ± 6.69 ± 0.4

A total of 38 AM fungal phylotypes, based on 2775 AM fungal clones (c. 62%), was identified in all root samples (Table 4, Fig. S4). Although the numbers of clones of AM fungi varied in the treatments, inspection of species accumulation curves of AM fungal phylotypes in mixed and E. nutans roots showed that the curves from F60, F90 and F120 were closer to asymptotic than the curves from F0 and F30 (Fig. S3b,c), suggesting that a higher proportion of the diversity of AM fungi in roots was captured in the high fertilizer treatments. Of the 38 detected AM fungal phylotypes, 10 phylotypes were related to sequences of described species, 19 to uncultured AM fungi, one to a spore-derived sequence and eight phylotypes were previously undescribed (< 97% identity with published sequences). Thirty-seven and 24 phylotypes were detected from mixed and E. nutans roots, respectively and, except for Phylo-38 (related to Acaulospora scrobiculata), all of the phylotypes detected in E. nutans were also detected in the mixed root samples (Table 4). Taken as a whole, Phylo-1 (related to G. intraradices/fasciculatum) was the most abundant phylotype in both mixed and E. nutans root systems.

Table 4.   Relative abundances (%, proportion of clone numbers) of each arbuscular mycorrhizal (AM) fungal phylotype and indicator species of fertilizer treatments detected in mixed and Elymus nutans roots in different fertilization treatments on the Tibetan Plateau Thumbnail image of

Diversity and composition of glomeromycotan communities across the fertilization gradient

Spore morphotype richness was not changed significantly by fertilization treatment (F4 = 0.93, = 0.470), and did not show a significant correlation with the score of N (r2 = 0.039, = 0.172) or P (r2 = 0.01, = 0.627) availability. However, the phylotype richness of AM fungi in mixed (F4 = 30.47, < 0.001) [correction added after online publication 14 March 2012: in the preceding text, the F4 value has been corrected to read as 30.47] and E. nutans (F4 = 5.99, = 0.002) roots was highly affected by treatment, and both showed significant negative correlations with the scores of N and P availability (Fig. 1c,d). The phylotype richness of AM fungi in both mixed and E. nutans root systems was interrelated (= 0.692, < 0.001; Pearson correlation), whereas neither was correlated significantly with the spore species richness. In addition, the species richness of AM fungi in mixed roots (= 0.664, < 0.001), but not the spore species richness (= − 0.051, = 0.809), was positively correlated with plant species, suggesting that plant diversity is an important determinant of AM fungal diversity in roots.

Although spore species richness was similar among treatments, spore community composition correlated with most soil and plant variables and also with plant richness (Fig. 2b). The community composition of AM fungi in mixed (Fig. 2c) and E. nutans (Fig. 2d) roots was also clearly affected by the fertilizer treatments. Fifteen soil and plant variables fitted as vectors onto the ordination of the AM fungal community in mixed roots showed that 11 variables were correlated significantly; of these, the soil available P (r2 = 0.932, < 0.001) and available N (r2 = 0.726, < 0.001), and the plant richness (r2 = 0.882, < 0.001) and root N (r2 = 0.702, < 0.001), were the two most important soil and plant variables, respectively (Fig. 2c). In addition, the AM fungal community in E. nutans roots was correlated significantly with most soil and plant nutrient variables, especially the soil available P (r2 = 0.904, < 0.001), root N (r2 = 0.719, < 0.001) and root P (r2 = 0.700, < 0.001; Fig. 2d).

Indicator species analyses of the community of AM fungi in mixed and E. nutans roots revealed some significant (Indval ≥ 0.6,  0.05) indicator phylotypes for each fertilization treatment. Indicator species for each treatment were often different between the AM fungal community in mixed and E. nutans roots, except for two uncultured Glomus species (Phylo-2 and Phylo-25), which were indicators of F0, and Archaeospora trappei (Phylo-32), which was an indicator of F120, in both mixed and E. nutans root systems (Table 4). Although there was no significant indicator spore morphotype for each treatment (all Indval 0.5), the abundance of Scutellospora dipurpurescens was negatively correlated with the N (r2 = 0.225, = 0.01) and P (r2 = 0.378, < 0.001) availability scores, and positively correlated with the soil pH (r2 = 0.389, < 0.001).

Relationships among N and P availability and communities of plants and AM fungi

Mantel tests using AM fungal communities in mixed roots against the matrices of plant community, plant biomass and N or P variables showed that all tested matrices were related significantly to the fungal community (Table 5); the plant community achieved the highest Mantel r score, followed by P availability, with the plant biomass the lowest. Similarly, the communities of spore and root-inhabiting AM fungi in E. nutans were related significantly to N or P availability and plant community, and the most related matrices were plant community and N availability, respectively (Table 5). Moreover, the AM fungal community in mixed roots showed significant correlations with that in E. nutans roots (Mantel = 0.6196, = 0.001) and the spore community (Mantel = 0.4618, = 0.001).

Table 5.   Mantel tests using community matrices of Glomeromycota based on spore and molecular identification with matrices of plant, N variables and P variables to explain the effects of plant, N availability and P availability on the community composition of arbuscular mycorrhizal (AM) fungi
MatrixComponent variablesAM fungal community in mixed rootsAM fungal community in Elymus nutans rootsSpore community
Mantel rP valueMantel rP valueMantel rP value
  1. Soil AN, soil available N; Soil AP, soil available P.

Plant communitySpecies composition0.7020.0010.47920.0010.33980.001
Plant biomassShoot biomass, root biomass, root : shoot biomass ratio0.16170.032−0.0030.531
N availabilitySoil AN, shoot N, root N0.47000.0010.49500.0010.28190.004
P availabilitySoil AP, shoot P, root P0.59150.0010.46950.0010.30240.002

Our SEM successfully elucidated the causal relationships among our experimental variables in the mixed root samples. The model fitted the data very well (maximum likelihood, χ2 = 0.528, = 0.467; Bollen–Stine bootstrap, = 0.436; GFI, = 0.988) and accounted for 84% and 87% of the variance in the extraradical hyphal length density and community composition of AM fungi, respectively (Fig. 4a). The effect of fertilization on plant community composition (λ = − 0.87) was stronger than its effect on the composition of AM fungi in mixed roots (λ = − 0.07); however, the plant community showed a very strong effect on the AM fungal community (λ = 0.86) as well as the extraradical hyphal length density (λ = 0.42). Moreover, fertilization led directly to negative changes in the root : shoot biomass ratio (λ = −0.44) and extraradical hyphae of AM fungi (λ = − 0.42; Fig. 4a). Taken as a whole, the effect of fertilization on the composition of the AM fungal community was mainly by indirect paths (mediated through host plants), whereas the effect of fertilization on extraradical hyphae was by both direct and indirect paths (Fig. 4b).

Figure 4.

Fertilization influences directly and indirectly communities of plants and Glomeromycota, root : shoot biomass ratio and extraradical hyphae of arbuscular mycorrhizal (AM) fungi. (a) A structural equation model (SEM) showing the causal relationships among fertilization, plant and AM fungal variables. The community composition of AM fungi includes data from the mixed root samples only. The width of arrows indicates the strength of the causal effect. The numbers above the arrows indicate path coefficients (λ ≥ 0.05 indicates significant pathway). Bold and dashed lines indicate significant and nonsignificant pathways, respectively. R2 values represent the proportion of variance explained for each variable. (b) Direct, indirect and total effect coefficients of fertilization on plant and AM fungal variables in this SEM.


Our long-term field experiment supports our research hypotheses and highlights the negative effect of high levels of fertilizer on Glomeromycota. As predicted by the functional equilibrium model, fertilization reduced significantly the AM colonization of roots and the extraradical hyphal length density (H1; Fig. 3). Furthermore, fertilization reduced the species richness of plants and Glomeromycota (H2; Fig. 1). By measuring simultaneously the responses of plants and AM fungi to fertilization, we determined that Glomeromycota respond to fertilization both directly and indirectly (H3). The path coefficients of our SEM (Fig. 4) indicate that changes in the composition of AM fungi are largely mediated by shifts in the plant community composition, whereas the reduction in extraradical AM fungal hyphae is associated with both changes in the plant community composition and factors that are independent of the plant community and root : shoot biomass ratio. Our findings corroborate other studies showing that AM fungi have a certain degree of ecological specificity for their hosts (Helgason et al., 2002; Vandenkoornhuyse et al., 2003; Liu et al., 2011). Half as many AM fungal species were observed in E. nutans roots compared with the mixed root samples (Table 4). Consequently, the fertilizer-induced loss of plant diversity should be expected to generate dramatic changes in the community composition of AM fungi as E. nutans becomes an increasingly dominant plant. The strong direct path from fertilization to extraradical hyphal length density and the weaker indirect path through the root : shoot biomass ratio (Fig. 4) suggest that root quality may be more important than root quantity in determining the responses of fungal abundance to fertilization. This idea is supported because there is no correlation between root biomass and extraradical hyphal length density of AM fungi (data not shown). It is well established that fertilization reduces soluble carbon in the apoplast of plant roots (Schwab et al., 1991), and that the development of arbuscules and other AM fungal structures in plant roots is tightly linked to carbon exchange within the apoplast (Parniske, 2008; Kiers et al., 2011). This mechanism appears to be operating at both the scale of an individual plant species and the whole plant community (H4), because both E. nutans roots and mixed roots showed a progressive reduction in AM fungal colonization with fertilization (Fig. 3a,b).

Previous field studies in North America and Germany have shown that, when P is not limiting, N fertilization reduces AM fungal colonization, extramatrical hyphal lengths and spore biovolume (Johnson et al., 2003; Blanke et al., 2005). Our results corroborate these findings and add an additional dimension by analyzing simultaneously plant and fungal community responses to fertilization. In our experiment, the highest level of fertilization caused the number of plant species to decline by more than six-fold (Table 2), and reduced the number of AM fungal phylotypes by one-half (Table 4). This response is predicted by resource competition theory (Tilman, 1988), and such changes in diversity and composition of plant communities caused by fertilization have been well documented in ecosystems worldwide (e.g. Hautier et al., 2009; Dickson & Foster, 2011). It is likely that above-ground competition for light is a major mechanism for this response in our study system, because fertilization increased significantly the height of the winning plant species (E. nutans; Table 2). Our finding of reduced phylotype richness of AM fungi with fertilization is consistent with other root DNA-based studies of Glomeromycota in N and/or P fertilized fields (e.g. Santos et al., 2006; Alguacil et al., 2010); however, the underlying mechanism explaining why increased N and P availability reduces the species richness of mycorrhizal fungi is mostly unknown. It has been shown that the competitive ability for host carbohydrates varies among AM fungal species (Cano & Bago, 2005; Bennett & Bever, 2009) and, in some cases, competition can be sufficiently strong to exclude some AM fungal taxa from host roots (Hepper et al., 1988). Consequently, a mechanism for reduced species richness and shifts in the community composition of AM fungi across the fertilization gradient can be partially explained by the enhanced carbohydrate competition among fungal species as plants allocate less carbon to mycorrhizal symbioses. We detected indicator AM fungi in our low- and high-fertility treatments (Table 4): Phylo-2 and Phylo-25 (undescribed Glomus phylotypes) were specific to F0, whereas Phylo-32 (related to A. trappei) was specific to F120; these results suggest that some glomeromycotan species prefer to inhabit soils with particular fertility levels.

Spore communities were generally less responsive to the nutrient treatments and plant community changes compared with the AM fungi inside plant roots (Tables 3, 5); nevertheless, spore abundance decreased gradually from F30 to F120, suggesting that high soil fertility can also depress sporulation by AM fungi. In contrast with the significant reduction in species richness of AM fungi inside plant roots, the spore richness remained constant across the fertilization treatments (Table 3). This result is consistent with some studies (e.g. Johnson, 1993 (N and P fertilization); Mathimaran et al., 2007 (P fertilization)), but differs from the study by Egerton-Warburton et al. (2007), in which spore richness in four of five grassland sites declined with N enrichment, but the remaining site was increased by N or N and P fertilization. Spore abundance of some AM fungal species was reduced significantly in the F120 treatment (Table 3), and it is possible that, over time, these species will disappear from the high-fertility plots. The striking difference in the species composition of AM fungi in roots and spores is also noteworthy and has been reported previously (Clapp et al., 1995; Börstler et al., 2006). It has been suggested that the phenology of AM fungi may generate distinct root and spore communities, and this may help partition fungal niches in time and space (Pringle & Bever, 2002). Alternatively, the ecological and evolutionary forces that structure spore communities of AM fungi may be entirely different from the forces structuring AM fungal communities inside roots. It is interesting to note that both spore density and colonization (%RLC) in mixed roots peaked in F30 rather than in F0 plots (Fig. 3), supporting the idea that severe limitation of N and P could restrict AM fungal growth, such that slight enrichment of these nutrients would be beneficial to fungal fitness in terms of sporulation and root colonization (Amijee et al., 1989; Treseder & Allen, 2002).

The reciprocal influences of plants and soil organisms on each other have been shown to explain many patterns in ecological communities (Reynolds et al., 2003). Our findings show that fertilization affects the reciprocity among plants and AM fungi, and that E. nutans and A. trappei (Phylo-32) come to dominate fertilized plots in our grassland system. Future studies are necessary to determine whether or not this outcome arises from a positive feedback between E. nutans and A. trappei, or whether each species is simply the best adapted to an enriched fertilized environment. Regardless of the mechanism, it is likely that fertilizer-induced changes in communities of plants and Glomeromycota generate changes in soil stability and nutrient cycling because the extraradical mycelium of AM fungi plays a critical role in creating and maintaining stable soil aggregates (Miller & Jastrow, 2000), and AM fungal species differ in their production of external hyphae (Cano & Bago, 2005). Furthermore, decomposition dynamics varies tremendously among plant species; grasses adapted to low-fertility soil tend to have more recalcitrant litter than those adapted to nutrient-rich soil (Wedin, 1995; Li et al., 2011).

In summary, our study shows that increasing the N and P availability in an alpine meadow ecosystem reduces the abundance of Glomeromycota and the species richness of both plants and AM fungi. To our knowledge, this study provides the first molecular-based evidence that increased N and P availability alters glomeromycotan communities at both the community and single-plant-species scales. Our results suggest that the enrichment of natural ecosystems with high levels of N and P may reduce dramatically above- and below-ground biodiversity, and potentially jeopardize the sustainability of alpine meadow ecosystems. Future studies are needed to assess how fertilization influences the total soil bacterial and fungal communities, as well as some particularly important functional groups such as N-cycling microbes, before we can fully appreciate the feedbacks among above-ground and below-ground communities and their soil conditions.


We are grateful to Dr Zhongling Yang for facilitating the sample collection and vegetation investigation. We thank Professor Alastair Fitter and three anonymous reviewers for helpful comments on earlier versions of the manuscript. This research was supported by the National Basic Research Program of China (2012CB026105), National Natural Foundation of China (40930533, 31170482), State Key Laboratory of Frozen Soil Engineering, Chinese Academy of Sciences (SKLFSE200901), The Major Project of Cultivating New Varieties of Transgenic Organisms (2009ZX08009-029B) and PhD Programs Foundation of the Ministry of Education of China (2010021111002). N.C.J. acknowledges financial support from the US National Science Foundation (DEB-0842327) and the Fulbright Commission.