• Previous studies have shown that arbuscular mycorrhizal fungi (AMF) can influence plant diversity and ecosystem productivity. However, little is known about the effects of AMF and different AMF taxa on other important community properties such as nutrient acquisition, plant survival and soil structure.
• We established experimental grassland microcosms and tested the impact of AMF and of different AMF taxa on a number of grassland characteristics. We also tested whether plant species benefited from the same or different AMF taxa in subsequent growing seasons.
• AMF enhanced phosphorus acquisition, soil aggregation and survival of several plant species, but AMF did not increase total plant productivity. Moreover, AMF increased nitrogen acquisition by some plant species, but AMF had no effect on total N uptake by the plant community. Plant growth responses to AMF were temporally variable and some plant species obtained the highest biomass with different AMF in different years. Hence the results indicate that it may be beneficial for a plant to be colonized by different AMF taxa in different seasons.
• This study shows that AMF play a key role in grassland by improving plant nutrition and soil structure, and by regulating the make-up of the plant community.
Another way in which AMF may influence plant communities is by improving soil structure and soil aggregation. Soil aggregate formation is an important determinant for ecosystem functioning because soil aggregates improve penetration of the soil by air and water and render the soil more resistant to erosion (Oldeman et al., 1991; Pimental et al., 1995; Rillig & Mummey, 2006). Correlative field studies and pot experiments with single plants indicate that AMF promote soil aggregation and soil structure stabilization by binding and enmeshing soil particles together into bigger aggregates (Oades, 1993; Schreiner et al., 1997; Andrade et al., 1998; Miller & Jastrow, 2000). However, it is still unclear whether AMF influence soil aggregation in complex plant communities consisting of several different plant species.
To date, most studies investigating effects of AMF on plant growth have been executed in one growing season. However, most plant species are perennial and long-lived. In order to assess the ecological significance of AMF, several growing seasons should be studied rather than just one (Read, 2002). For instance, different AMF may be beneficial to the plant at a specific age or under specific environmental conditions (e.g. during drought or under nutrient stress). Some studies indicated that AMF have the greatest impact on growth of juveniles (Jones & Smith, 2004; van der Heijden, 2004), but none has investigated the effects of AMF and different AMF taxa on plant growth during two growing seasons. Information on the temporal variability of plant benefit when colonized by different AMF is also important for understanding evolutionary relationships between plants and AMF. Plants may co-evolve with those AMF that are beneficial throughout the lifespan of a species.
In this study, calcareous grassland microcosms were established and inoculated with different AMF taxa, or inoculated with an autoclaved inoculum (control microcosms without AMF). The effects of these AMF treatments on plant diversity and plant community structure have been published previously (see experiment 1 of van der Heijden et al., 1998a). Here we present data from the same grassland microcosms and test whether AMF and different AMF taxa influence nutrient uptake (N and P), soil structure, soil aggregate stability and plant survival. In addition, we investigated whether there were any temporal effects of AMF on plant productivity, and if plant species benefited from the same or different AMF taxa in different growing seasons. We hypothesized that AMF influence ecosystem processes by providing additional nutrients and improving soil structure.
AMF are usually present in grassland, so it is difficult to manipulate AMF and AMF community composition under field conditions (see Read, 2002 for a discussion). Therefore we simulated experimental grassland communities in microcosms with soil, plant and AMF material collected from a calcareous grassland in Switzerland.
Materials and Methods
Experimental grasslands, plant and fungal material
Forty-eight microcosms simulating European calcareous grassland were established under glasshouse conditions in containers measuring 26.5 × 17 × 18 cm. These containers were filled with a soil mixture (5.92 kg DW; approx. 7 l soil) of sieved autoclaved soil obtained from calcareous grassland in Switzerland and autoclaved quartz sand (2 : 1 v/v). Two 3-kg subsamples of this soil mixture were taken at the start of the experiment, air-dried, and stored for later analysis (chemical analysis; soil aggregation). At the bottom of each container, a small hole was made for drainage. A sterilized drainage mat covered the bottom to prevent soil loss. The microcosms were inoculated with 100 g soil inoculum containing one of the four AMF taxa (four single AMF treatments); or a mixture of these four AMF isolates (mixed AMF treatment); or just an autoclaved (121°C, 30 min) soil mixture of these four AMF isolates (nonmycorrhizal control treatment). The four AMF used, all belonging to the genus Glomus (Phylum: Glomeromycota), were isolated from the same calcareous grassland in Switzerland: Glomus sp. (Basle Pi: AMF A); Glomus geosporum (Nicol. & Gerd.) Walker (BEG 18: AMF B); Glomus sp. (BEG 21: AMF C); Glomus sp. (BEG 19: AMF D). Inoculum of each isolate was propagated on Hieracium pilosella in pots filled with sterilized loamy sand. Further details about these AMF isolates are given by Streitwolf-Engel et al. (1997); van der Heijden et al. (1998b). In earlier work, AMF C (Glomus sp. isolate BEG 21) was described as being similar in morphology to Glomus laccatum Blaszkowski. However, soil harvested from microcosms inoculated with AMF C contained only spores similar in morphology to Glomus intraradices (F. Oehl and E. Sieverding, personal communication) and DNA sequences of the internal transcribed spacer (ITS) regions of these spores showed 97% homology on the ITS1 side and 93% homology on the ITS4 side with a G. intraradices species. It is unclear whether the potentially different identity of AMF C is caused by contamination of inoculum or by difficulties with identification of the initial inoculum, as the description of G. laccatum is based on spores extracted from the field that were never obtained in pure culture.
Each microcosm received a 100 ml filtered washing of soil inoculum from the mixed AMF treatment (without AMF propagules) and 280 ml filtered washing of field soil (also without AMF propagules), respectively, to correct for possible differences in microbial communities between the different inocula, and to include microbial communities from natural calcareous grassland (Koide & Li, 1989). A total of 0.5 kg soil of the mixed inoculum and 2.2 kg field soil was wet-sieved through a series of sieves to prepare the microbial wash. The finest sieve was 10 µm. In each microcosm, seedlings of 11 different plant species were planted. Seedlings were approx. 2 wk old when planted. In total, 70 plant individuals were planted in each microcosm. The number of individuals per plant species is given in Table 1. These numbers corresponded to their natural abundances in calcareous grassland in Switzerland (Leadley & Stöcklin, 1996). The identity of these 11 plant species is also given in Table 1. Note that the species name Sanguisorba officinalis was mistakenly used in previous publications from this experiment (van der Heijden et al., 1998a; van der Heijden, 2002); the correct name is Sanguisorba minor.
Table 1. Number of individuals of each plant species in microcosms at the start of the experiment and after 20 months, when microcosms were harvested
P values for anova or nonparametric Kruskal–Wallis H-test are given for each plant species. Different letters indicate a significant difference among means of the treatments according to Tukey's test or Kruskal–Wallis-test. Significant differences among treatments are shown in bold.
A nonparametric test was performed because of heteroscedasticity among AMF treatments.
The 70 individuals were planted at fixed distances (2.5 cm) from each other according to a predefined design. Seedlings that died within 4 wk after planting were replaced so that each microcosm contained 70 seedlings after 4 wk. Eight different planting designs were used, each design being assigned to one replicate of each AMF treatment. A randomized block design was used (see below) where each plant design constitutes a block. Thus plant individuals in microcosms belonging to one block had the same neighbours, while plants in different blocks had different neighbours. A total of eight blocks (eight planting designs) were present. This approach was chosen to avoid potential differences between treatments being confounded by neighbourhood interactions and initial plant species composition (van der Heijden, 2004). Planting and harvesting were performed in the order of block numbers: block 1 first and block 8 last. Microcosms were established from 19–26 April 1996 and maintained in the glasshouse during two growing seasons (April 1996–December 1997). The night temperature in the glasshouse was 19°C; day temperatures varied between 23 and 35°C depending on the weather conditions outside. Additional lighting was provided with high-pressure mercury vapour lamps (Philips HPL-N, 400 W) to a daylength of 16 h d−1 from May 1996–January 1997, and from April–December 1997. Additional lighting was provided when natural light levels reached <400 W m−2. The microcosm received a winter period from January–March 1997. The day and night temperature in the glasshouse during the winter period varied between 4 and 7°C. No additional lighting was provided during the winter period, reducing daylength to approx. 9 h d−1. Microcosms were watered three times a week with distilled water, and each microcosm was adjusted to equal soil water content every 2 wk by weighing. The microcosms received no fertilization.
Measurements and harvesting
Plant variables Aboveground plant parts (defined as shoots) were cut per plant individual, approx. 2.5 cm above the soil surface, from 8–17 July 1996; 9–18 September 1996; 9–19 December 1996; and 15–31 July 1997. This cutting simulates hay-making as carried out in natural calcareous grassland. The microcosms were finally harvested in December 1997, when plants were cut at the soil surface. Each plant individual was harvested separately; a total of 10 183 plants were harvested. The shoot biomass of Bromus erectus, Brachypodium pinnatum, Carex flacca, Festuca ovina, Lotus corniculatus, Sanguisorba minor and Trifolium pratense was determined in year 1 (1996) and year 2 (1997) by pooling weights per microcosm for the first three harvests (1996) and the last two harvests (1997). Several plant species (Centaurium erythrea, Hieracium pilosella, Prunella grandiflora, Prunella vulgaris) remained small or grew close to the soil surface, and hardly any biomass of these species was obtained during cutting before the final harvest. The biomass of these species in years 1 and 2 was estimated by dividing their total biomass by two. Total shoot biomass of the microcosms in years 1 and 2 was estimated by adding the biomass of the individual plant species. The total shoot biomass of plant species in the microcosms has been published previously (experiment 1 of van der Heijden et al., 1998a). The number of plant individuals of each plant species still present in each microcosm at the final harvest was recorded to determine plant survival. Biomass data for individual plant species were used to calculate the equitability index J (a measure of evenness; Begon et al., 1996).
At final harvest, the soil was separated from plant roots, carefully mixed, weighed and stored in closed plastic bags at 4°C (a subsample of 0.5 kg) or oven-dried at 70°C for 48 h (remaining soil). Roots were washed, mixed, cut into 1-cm parts and divided into two subsamples, and fresh mass was determined on both subsamples. Total root length was determined (Marsh, 1971) on three samples taken from one subsample; the other subsample was oven-dried (70°C) and weighed. Total root length and root DW were subsequently determined g−1 dry soil.
Dried plant material was ground in a ball mill and mixed thoroughly, and P concentrations were determined of the shoot biomass of each plant species and of the roots. Phosphorus concentrations were determined by the molybdate blue ascorbic acid method (Watanabe & Olsen, 1965). Ground plant material of the nonmycorrhizal control communities and of microcosms inoculated with AMF A + B + C + D was used to determine the concentrations of N, and of the isotope 15N, for each plant species and for the roots. The N and 15N concentrations were determined by isotope ratio mass spectrometry (ANCA SL, MS 20–20; PDZ Europe Ltd, Northwich, UK). The P and N contents of individual plants species and of the roots were calculated, and the sum of these gave total plant P and N contents per microcosm. For several samples (especially of the nonAMF treatment), not enough plant material was available to determine P and N concentrations. In this case, samples of one AMF treatment were pooled. In these cases, P and N contents were calculated by making use of average P and N concentrations of the treatment investigated. No plant material was available to determine P and N concentrations for C. erythrea and N concentrations for P. vulgaris and T. pratense, in the nonmycorrhizal microcosms.
Fungal variables After determining root length, the same roots were cleared with 10% KOH and stained with trypan blue in order to stain mycorrhizal structures inside the roots (Phillips & Hayman, 1970). The percentage of root length colonized by AMF was estimated on three subsamples by a modified line-intersection method (McGonigle et al., 1990). Generally, to assess AMF colonization 100 line intersections per root sample are scored for the presence of AMF (McGonigle et al., 1990). We followed a slightly different procedure, and measured a minimum of 50 line intersections for each of three root subsamples. We decided to increase the number of root subsamples because there was considerable variation between them. Hence a total of 150 line intersections were scored for the presence of AMF. In addition, we estimated AMF colonization levels for individual plant species using roots that were connected to individual plants, following the same procedure as described above. The soil of the microcosms contained a dense root system, and it was difficult to isolate roots of individual plant species. As a result, relatively coarse roots were collected (fine roots could not be isolated because they were intermingled with roots of other plants). Hence mycorrhizal colonization levels of individual plant species represent an estimate. Some plant species remained very small in the control microcosms without AMF, and not enough root material could be isolated for each individual plant species to assess mycorrhizal colonization from all replicates. Moreover, C. erythrea remained small in almost all microcosms, and not enough root material was obtained to determine mycorrhizal colonization for this species. Consequently, colonization levels are given for 10 instead of 11 plant species.
A subsample of soil was taken, and the length of fungal hyphae in this soil was determined by an aqueous extraction and membrane-filter technique (Jakobsen et al., 1992). Before determining hyphal length density, soil was stored at 4°C for a maximum of 4 months. Trypan blue was used to stain the hyphae. Hyphal length density was estimated per g soil (DW). Hyphal length density values represent all hyphae, dead and alive. Turnover of AMF hyphae can be rapid (Staddon et al., 2003), and the hyphal length density in this study represents an estimate because soil was stored for a maximum of 4 months. It is sometimes difficult to distinguish mycorrhizal hyphae from hyphae of other soil fungi. In order to estimate the hyphal length density of AMF, we assumed that the amount of nonmycorrhizal hyphae was the same in mycorrhizal and nonmycorrhizal microcosms.
Soil aggregation Three variables related to soil structure were measured in microcosms inoculated with the mixed AMF treatment or in the uninoculated control microcosm. First, 3 kg oven-dried soil was taken, and the proportion of particles/aggregates <1 mm; 1–2 mm; 2–3.15 mm; 3.15–5 mm; and >5 mm was determined by sieving and weighing (weights of respective fractions as percentage of total soil DW). This determination was also performed in two subsamples of soil that had been autoclaved and dried at the start of the experiment, and kept stored during the experiment (no AMF treatment). Second, 150 g soil particles/aggregates >1 mm were taken, rewetted and subjected to wet sieving on a 1-mm-mesh screen, for 10 min, in a wet-sieving apparatus with a stroke length of 34 min−1 (Degens et al., 1994). The proportion of nonaggregate soil particles (mainly stones) in the fraction that remained after wet sieving was determined by dissolving the aggregates in a sodium hexametaphosphate solution. This suspension was sieved on a 1-mm-mesh screen and the weight of nonaggregate soil particles determined. The weight of nonaggregate soil particles was subtracted from the weight of the wet-sieved soil aggregates, thus the proportion of water-stable aggregates could be determined (expressed as percentage water-stable aggregates). The proportion of water-stable aggregates is an indicator of aggregate stability (Degens et al., 1994). Finally, the rate of water percolation through a soil column with 10 g of oven-dried soil aggregates (1–2 mm aggregate fraction) was measured (as described by Siegrist et al., 1998). This is another, more dynamic, estimate of aggregate stability, for example, the more water flows through the soil column during a given time, the higher is the aggregate stability and the lower is the soil's tendency to surface sealing (Siegrist et al., 1998). Soil collected from the microcosms at the end of the experiment was collected and total soil DW was determined for each microcosm.
Soil chemical analysis The pH–H2O, amount of CaCO3 per 100 g dried soil, percentage organic matter, and soil P concentration (mg per 100 g) were determined in soil samples collected from the nonmycorrhizal control microcosms (eight samples); from microcosms inoculated with the mixed AMF treatment (eight samples); and from the dried soil mixture (two 3-kg samples), stored from the beginning of the experiment. Soil P concentrations were determined using lactic acid, sodium acetate and citrate acid as extracting solutions (Labor Dr FM Balzer, Wetter Amönau, Germany).
Experimental design and statistical analysis
The experiment was set up as a randomized block design where each AMF treatment was replicated eight times. Each replicate was assigned to one block, making a total of eight blocks. For each variable, an anova was performed to test whether treatments varied from each other (Proc GLM; spss ver. 10.1). The anova consisted of two factors: AMF (with six levels); and block (with eight levels). The factor AMF was treated as a fixed effect; the factor block as a random effect. A significant block × AMF effect was not expected, and the AMF effect was tested using the residual mean square as denominator to calculate the F ratio (Newman et al., 1997). The block effect was significant for the following variables: total biomass; root biomass; equitability index; total P content; total N content; P concentration of P. vulgaris; soil particle/aggregate fractions 1–2 mm, 3.15–5 mm, >5 mm; and soil weight. Tukey's test was performed to test which treatments varied from each other.
A repeated-measures anova was performed in order to test whether plant species responded differently to different AMF taxa in different years. The repeated anova consisted of two factors, AMF and block, as explained above. Biomass data from the first and second years were used as repeated measures. The nonmycorrhizal treatment was excluded from the repeated-measures anova to test specifically for differences between different AMF taxa.
If necessary, variables were transformed to meet the requirement of homoscedasticity of variance among treatments. A nonparametric test (Kruskal–Wallis H) was performed when variables could not be transformed. This was done for the following variables: percentage root length colonized by AMF; and number of plant individuals in microcosms for some plant species (Table 1). In these cases no block effect was included in the analysis. Pearson's correlation coefficients were calculated to test for correlation between two variables.
Temporal effects of AMF on plant growth
Total shoot biomass of the microcosms varied significantly among the different AMF treatments in year 1 (F5,35 = 8.3, P < 0.001) and year 2 (F5,35 = 27.5, P < 0.001). In year 1, the shoot biomass of microcosms inoculated with AMF A + B + C + D was highest and differed from noninoculated control microcosms (Fig. 1). However, in year 2 the shoot biomass of the noninoculated control microcosms was highest, and differed from microcosms inoculated with AMF (Fig. 1). The effect of the different AMF treatments on growth of individual plant species was variable, and not the same in both years (Fig. 1). For instance, shoot biomass production by Sanguisorba varied greatly among microcosms inoculated with different AMF in the first year of the experiment. However, this differential response was almost absent in the second year when plants were older (Fig. 1). Accordingly, the F ratio in an anova, executed without the nonmycorrhizal treatment, was much higher for shoot biomass data for the first year (F4,28 = 19.4, P < 0.001) compared with shoot biomass data for the second year (F4,28 = 0.67, P = 0.62). There was a significant year × AMF treatment interaction term for the shoot biomass of B. erectus, B. pinnatum, C. flacca, L. corniculatus and S. minor (Table S1 in Supplementary Material). This indicates that these species responded differently to different AMF taxa in years 1 and 2. There was no significant year × AMF treatment interaction term for shoot biomass of F. ovina and T. pratense (Table S1).
Plant productivity and plant evenness
Total biomass of microcosms (root plus shoot weight) varied significantly among AMF treatments (F5,35 = 10.59, P < 0.001). The total biomass of nonmycorrhizal control microcosms (mean = 115.7 g) was significantly higher than the biomass in microcosms inoculated with AMF A (103.1 g), C (99.9 g), D (94.3 g) and A + B + C + D (101.5 g); it was 14% higher compared with the average total biomass of the five treatments inoculated with AMF. Root biomass was, on average, 26% higher in control microcosms compared with microcosms inoculated with AMF (Fig. 2a). The root-weight ratio (RWR, root mass as fraction of total plant mass) varied significantly among AMF treatments (F5,35 = 3.79, P < 0.008). The RWR of the nonmycorrhizal control microcosms was, on average, 9.4% higher compared with microcosms inoculated with AMF. The specific root length (SRL, root length per unit root mass) did not vary significantly among AMF treatments, although it was higher in control microcosms (132.4 mg−1 root) compared with microcosms inoculated with AMF (126.1 mg−1 root).
Plant evenness (as measured by the equitability index) was generally higher in microcosms inoculated with AMF compared with control microcosms (Fig. 2b). Plant evenness also varied among microcosms inoculated with different AMF (Fig. 2b). For instance, plant evenness in microcosms inoculated with AMF A was significantly higher than in those inoculated with AMF B.
The number of plants surviving to the end of the experiment was influenced by AMF treatment for four out of 11 plant species (Table 1). For instance, microcosms inoculated with AMF A + B + C + D contained significantly more individuals of T. pratense compared with nonmycorrhizal microcosms. Moreover, C. erythrea was almost absent from microcosms inoculated with AMF C, while the highest number of plants was found in microcosms inoculated with AMF B. This indicates that survival of this species varied between microcosms inoculated with different AMF. Almost all plant individuals of B. erectus, B. pinnatum, C. flacca, F. ovina, P. vulgaris and S. minor survived the experiment. Species richness was the same in all treatments because at least some plant individuals of each plant species survived.
Mycorrhizal colonization, hyphal length density and root length
The degree to which plant roots were colonized by AMF varied among microcosms inoculated with different AMF taxa (Fig. 2c). Colonization levels of roots in microcosms inoculated with AMF B were significantly lower compared with colonization levels of roots in microcosms inoculated with AMF A or C. Plant roots in nonmycorrhizal control communities were not colonized in seven out of eight microcosms. One subsample of plant roots in one-control microcosms was colonized by AMF (2.1% of root length in this microcosm was colonized by AMF). Colonization levels of different plant species also varied, and depended on fungal identity (Table S2). Trifolium pratense was most heavily colonized, while colonization levels of C. flacca were lowest. Colonization levels of most plant species were highest in microcosms inoculated with AMF A, C or A + B + C + D, and lowest in microcosms inoculated with AMF B or D (Table S2).
Hyphal length density also varied among microcosms inoculated with AMF (Fig. S1). Hyphal length density did not correlate with the percentage of root length colonized by AMF (R2 = 0.01, P = 0.60, n = 40; data for nonmycorrhizal microcosms excluded).
Total root length varied among different AMF treatments (Fig. 2d). Nonmycorrhizal microcosms contained, on average, 34% greater total root length compared with the mean for microcosms inoculated with AMF. The total root length also varied between microcosms inoculated with different AMF taxa (Fig. 2d). Total root length and hyphal length density were negatively correlated (R2 = 0.21, P < 0.001, n = 47). Microcosms with high amounts of roots (such as the nonmycorrhizal treatment) contained low amounts of external hyphae, while microcosms with low amounts of roots (AMF D) contained large amounts of external hyphae.
Phosphorus and nitrogen nutrition
Total plant P content per microcosm varied among the different treatments (Fig. 2e). The average plant P content in microcosms inoculated with AMF was 0.92, while only 0.64 g P was found in nonmycorrhizal communities. Hence mycorrhizal communities contained, on average, 44% more P compared with communities without AMF (Fig. 2e). This shows that AMF enhanced overall P uptake by the whole plant community. Plant P content also varied between microcosms inoculated with different AMF taxa. Microcosms inoculated with AMF A and D contained significantly more P than microcosms inoculated with AMF B (Fig. 2e).
The total P concentration in plant tissue also varied significantly among treatments, being lowest in nonmycorrhizal microcosms and highest in microcosms inoculated with AMF D (Table S2). The shoot P concentration of individual plant species depended on AMF treatment, and was higher in microcosms inoculated with AMF compared with control microcosms without AMF (Table S2). Interestingly, the P concentration of the sedge C. flacca (which is thought to be a nonmycorrhizal plant and was hardly colonized by AMF in this study) was also enhanced in mycorrhizal microcosms (Table S2). Most plant species contained the highest P concentration in microcosms inoculated with AMF D (Table S2).
Shoot P content of individual plant species also varied significantly among AMF treatments (Fig. S2), and was significantly higher in microcosms inoculated with AMF for eight out of 10 plant species (Fig. S2). Most plant species had the lowest shoot P content in microcosms inoculated with AMF B (Fig. S2) when microcosms inoculated with different AMF taxa were compared. The shoot P content of seven out of 10 plant species was significantly positively correlated with the percentage of root length colonized by AMF (data not shown).
Soil P concentrations at the end of the experiment were significantly lower in communities inoculated with AMF A + B + C + D compared with control communities using citric acid or sodium acetate as P-extraction method (Table 2). No difference in soil P concentration was observed between both treatments when lactic acid was used to extract soil P (Table 2). The soil P concentration at the start of the experiment (5.4 mg per 100 g sodium acetate extractable P) was much higher compared with end concentrations.
Table 2. Mean values for different soil variables at the start of the experiment (no treatment) and after two growing seasons in microcosms inoculated with arbuscular mycorrhizal fungi (AMF) A + B + C + D (+AMF) or without AMF (NM)
For each soil variable, mean values for no treatment, +AMF and NM are given in subsequent columns. The last column (P) shows for each variable whether there was a significant difference between nonmycorrhizal microcosms or microcosms with AMF according to anova. Values for no treatment are given for comparative reasons only and were not included in the anova because only two soil samples were analysed (see Materials and Methods). Significant differences among treatments are shown in bold.
Three different extraction buffers (lactic acid, sodium acetate, citrate acid) were used to extract P from soil.
Each microcosm received 5.92 kg soil at the start of the experiment.
Total N content did not vary significantly between microcosms inoculated with AMF A + B + C + D (average = 1.57 ± SE = 0.06) and control microcosms without AMF (average = 1.50 ± SE = 0.07). However, the total N concentration of plant material (root plus shoots) varied significantly between the two treatments, and was higher in mycorrhizal microcosms (Table S2). Moreover, the N concentration of P. grandiflora and S. minor, and of roots, was significantly higher in microcosms inoculated with AMF A + B + C + D compared with nonmycorrhizal microcosms (Table S2). The N content of H. pilosella and S. minor was also significantly higher in mycorrhizal microcosms (data not shown).
The concentration of the isotope 15N varied among plant species. Average δ15N values for both legumes (L. corniculatus and T. pratense) were, respectively, 0.366 and 0.369. Values for nonlegumes were 0.368, 0.368, 0.370 and 0.371, respectively, in B. pinnatum, C. flacca, S. minor and F. ovina, up to 0.374, 0375 and 0.378, respectively, in H. pilosella, B. erectus and P. grandiflora.
Soil aggregation and soil chemical properties
The soil that was added to microcosms at the start of the experiment contained proportionally more and larger aggregates than soil from microcosms with or without AMF at the end of the experiment (Table 2). This indicates that aggregates were broken down and lost during this experiment. Aggregate loss depended on AMF treatment. Microcosms inoculated with AMF (A + B + C + D) had 3.2% more aggregates (>1 mm) than microcosms without AMF. The proportion of aggregates in classes 1–2, 2–3.15 and >5 mm was also significantly higher in soils of microcosms inoculated with AMF (Table 2). The percentage of water-stable aggregates, and the amount of water percolating in a given time through a column filled with soil particles/aggregates (another measure of aggregate stability), was higher in microcosms with AMF compared with those without AMF (Table 2), indicating that the presence of AMF promoted soil aggregate stability. There was a significant positive correlation between the amount of water-stable aggregates and the rate of water percolation (Pearson's correlation: n = 16, R2 = 0.41, P = 0.008), indicating that both measurements assessed the same type of aggregate stability.
The amount of aggregates (>1 mm) was negatively correlated with total root length (R2 = 0.63, P < 0.0001, n = 16) and not correlated with hyphal length density (R2 = 0.18, P = 0.10, n = 16).
The soil used in this experiment was rich in organic matter, contained considerable amounts of calcium carbonate and had a relatively high pH (Table 2). The pH of the soil and percentage organic matter did not vary between the two AMF treatments (Table 2).
Mycorrhizal contribution to plant community composition and plant survival
We observed that AMF enhanced plant diversity and changed plant community composition by stimulating the growth of subordinate plant species (van der Heijden et al., 1998a). Eight of the 11 plant species were almost completely dependent on the presence of AMF to be successful in the microcosms. The presence of AMF stimulated their growth (van der Heijden et al., 1998a) and enhanced P acquisition (present study). However, AMF did not necessarily enhance their survival. The majority of seedlings planted at the start of the present study survived, and for those plant species with low survival (Centaurium, Trifolium, Lotus and Hieracium), at least some individuals of each plant species were still present at harvest after 20 months (Table 1). Accordingly, species richness was not affected by AMF. Instead, AMF enhanced the evenness of plant communities (Fig. 2b).
Mycorrhizal contribution to P nutrition and biomass production
We reported previously that plant productivity and P uptake in experimental macrocosms simulating North American old-field communities increased when the number of AMF taxa increased (experiment 2 of van der Heijden et al., 1998a). However, increased mycorrhizal diversity (four compared with one AMF taxa) did not result in higher biomass or in increased nutrient acquisition in this experiment. This study is in accordance with results of an earlier short-term study (3 months), where no positive effect of AMF diversity was also found (van der Heijden et al., 2003). The AMF taxa used in this study belonged to the same genus (Glomus), while the experiment with old-field plant communities (experiment 2 of van der Heijden et al., 1998a) included five different AMF genera. Variations in growth effects by different AMF taxa appear to be largest at the genus level, not at the species or isolate level (Hart & Klironomos, 2002). Moreover, it has been indicated that Glomus species have a more ruderal life style and different life-history characteristics compared with other AMF genera such as Gigaspora (de la Providencia et al., 2005). This may indicate that it is more likely to find complementary effects of AMF diversity when different AMF genera (with different strategies) are present. Future studies that test whether AMF diversity promotes plant productivity should therefore include AMF taxa from different genera.
Effects of AMF on P uptake in plant communities are poorly documented. One study showed that AMF stimulated P acquisition, and this was positively correlated with plant biomass and hyphal length density in Canadian old-field communities (experiment 2 of van der Heijden et al., 1998a). Our data show that AMF contributed to P uptake in calcareous grassland communities, and that these effects can also be independent from plant biomass. The data indicate that AMF, not only plant roots, are responsible for P acquisition: nonmycorrhizal communities contained 32% greater total root length and had a higher total biomass. Despite this, mycorrhizal communities contained 43% more P. The additional P acquisition by AMF could be explained by several factors. First, the density of mycorrhizal hyphae was 4.5–14.8 times higher than that of roots (depending on AMF treatment). Consequently, hyphae could forage for nutrients in a given soil volume much more effectively than roots. Second, hyphae can enter soil pores inaccessible to roots (Smith & Read, 1997). This might be especially important in soils rich in aggregates, as used in this experiment. Third, AMF excrete enzymes such as phosphatases and can release P from some organic sources (Joner & Johansen, 2000). Fourth, kinetic parameters (Km and Vmax) for P uptake are much higher for mycorrhizal hyphae than for roots (Jakobsen et al., 2002), and this could also explain why plant communities inoculated with AMF contained more P than nonmycorrhizal controls. Fifth, several studies performed with single plants showed that AMF interact with root system architecture (Koide, 1991). Mycorrhizal colonization can reduce root weight, specific root length and root fineness, and these effects reduce the ability of plants to acquire nutrients. Finally, AMF altered plant diversity in these grassland communities (van der Heijden et al., 1998a), and differences in P uptake could be related to changes in plant species composition between mycorrhizal and nonmycorrhizal communities (because some plant species contain more P than others). However, as AMF increased the P concentration of all plant species, this possibility is unlikely.
Plant species from different functional groups coexisted in the microcosms: several grass species with dense and fine root systems; forbs with relatively thick roots; a plant species with dauciform roots (C. flacca), and two N-fixing legumes. Despite this diversity in resource acquisition strategies, AMF appeared to be the overruling factor determining P uptake. This study thus indicates that AMF play a key role in P nutrition in grassland.
This study also shows that P nutrition depends on the identity of AMF taxa present in a community. Almost all plant species contained the highest P concentration in microcosms inoculated with AMF D, indicating that this AMF was particularly efficient in supplying P. This was also observed in a short-term pot experiment where plants inoculated with AMF D also had the highest P concentration (van der Heijden et al., 2003). In another short-term experiment with three plant species this was not observed, perhaps because a different soil was used (van der Heijden et al., 1998b). Several other studies also showed that different AMF supply different amounts of P to the plant (Jakobsen et al., 1992; Koch et al., 2004; Munkvold et al., 2004; Smith et al., 2004). AMF also altered the distribution of nutrients among co-occurring plant species, as reported earlier (van der Heijden et al., 2003). Those plants that received most P from AMF also had the highest mycorrhizal dependency (van der Heijden, 2002; data not shown). This indicates that one mechanism by which AMF stimulate growth of mycorrhiza-dependent plant species is enhanced P supply.
The increase in P uptake did not result in increased biomass production, indicating that other factors controlled plant productivity and that ‘luxury’ P consumption occurred. It has been suggested that luxury P consumption is useful for the plant because it can be utilized to enhance seed quality (Koide et al., 1988) or at other times when P is limiting (Koide, 1991). However, aboveground P storage is only successful in the absence of grazing or hay-making/mowing. The aboveground biomass of the microcosms was harvested five times at regular intervals (to simulate hay-making as in natural grassland), and substantially higher amounts of P were removed from mycorrhizal compared with nonmycorrhizal microcosms (data not shown). Moreover, the fact that enhanced P uptake did not result in increased biomass production shows that P-use efficiency was reduced when AMF were present (the average P-use efficiency of biomass was 94.3 g g−1 in microcosms with AMF and 178.6 g g−1 in nonmycorrhizal microcosms).
Mycorrhizal contribution to N nutrition
A few studies have indicated that AMF enhance plant N nutrition (Tobar et al., 1994; Mäder et al., 2000; Jin et al., 2005). It has also been suggested that AMF acquire N from organic substances by enhancing decomposition (Hodge et al., 2001). In this study we did not observe that AMF enhanced the total amount of N in plant biomass (although the experiment lasted 20 months and AMF could have acquired N from the soil, which was rich in organic matter). However, the distribution of N among co-occurring plants was influenced by AMF, and some plants received more N when mycorrhizal. One other way in which AMF could improve N nutrition of the vegetation is by stimulating growth of legumes (that fix N through their symbiosis with rhizobia bacteria; van der Heijden et al., 2006). AMF had a large effect on growth of both legume species in this experiment. However, the biomass of both species was small compared with the total biomass, and these species had a negligible effect on the total N content of microcosms. The δ15N values of both legumes were close to zero, and comparable with those of other plants (approx. 0.37), and it was not possible to calculate whether biological N fixation had occurred (despite the observation of nodules in the legume roots).
Plant productivity in terrestrial grassland is usually limited by N and/or P availability (Chapin, 1980). The N : P ratio of plant material gives an indication of which nutrient limits plant growth (Koerselman & Meuleman, 1996). An N : P ratio >14 indicates that P is limiting, while a ratio <14 indicates that N is limiting growth (Koerselman & Meuleman, 1996). The N : P ratio of the plant species in this study varied between 0.7 and 7, indicating that N availability might be the factor limiting growth in the microcosms. As expected, plants that grew in microcosms inoculated with AMF had a lower N : P ratio (because of higher P supply) compared with those that grew in nonmycorrhizal microcosms, indicating that N limitation was even greater in mycorrhizal microcosms. However, other unknown factors could also have limited biomass production in the microcosms.
Mycorrhizal contribution to soil structure
We observed that AMF reduced the loss of soil aggregates and enhanced soil aggregate stability in the microcosms. Soil structure was thus improved by the presence of AMF. This is, to our knowledge, the first experimental demonstration that AMF contribute to soil aggregation in multispecies grassland communities. Our work confirms field studies that used correlative approaches to show that AMF play a pivotal role in improving soil structure (Oades, 1993; Miller & Jastrow, 2000). Interestingly, mycorrhizal microcosms were slightly heavier compared with nonmycorrhizal microcosms (Table 2). Although not significant, this observation deserves further investigation as it may indicate that AMF reduced soil loss by preventing soil erosion during watering.
It is difficult to elucidate the precise mechanisms responsible for the effects of AMF, because many factors (plant species composition, plant nutrition, root system size, soil structure, etc.) are simultaneously affected by AMF. The effects of AMF on soil aggregation could be independent of the hyphal network, and caused by indirect effects of AMF. AMF altered plant species composition in the microcosms which, in turn, could influence soil aggregation – it is known that different plant species differ from each other in influencing aggregate formation (Degens et al., 1994).
Temporal variation in plant growth response to AMF
This study showed that growth responses of plants to different AMF were temporally variable and plant species-dependent. Lotus and Trifolium performed best with one AMF in the first growing season, but grew best with a mixture of several AMF taxa in the second. In contrast, Sanguisorba and Brachypodium responded differently to different AMF in their juvenile state (first season), while such differential effects were much weaker in the second season, when plants were older. The precise mechanisms responsible for these observations are unclear. Juvenile plants start from seed and are often nutrient-limited because they have hardly any nutrient reserves. This is in contrast to older plants that have extensive root systems and can rely on stored nutrients. Hence it is not surprising that juveniles of several plant species (e.g. Sanguisorba and Brachypodium) responded much more strongly to AMF than adult plants (Fig. 1). This observation also indicates that, in those studies where perennial plants are being studied, observed AMF effects on plant growth probably represent an upper limit because such experiments are usually performed with seedlings.
Moreover, it has been shown that particular plant–AMF combinations are most efficient in stimulating plant growth and nutrient acquisition (van der Heijden et al., 1998b; Klironomos, 2003). Such optimal plant–AMF combinations may be temporally variable and age-dependent. The fact that adult plants and seedlings of the same plant species harbour different AMF communities (Husband et al., 2002) suggests that plants have the ability to select specific AMF. Perhaps selection of specific AMF types is based on their symbiotic performance (Kiers & van der Heijden, 2006). Moreover, it may be disadvantageous for a plant to specialize on certain AMF if this benefit is temporally variable. This could explain why host specificity in the arbuscular mycorrhizal symbiosis has rarely been observed (Law, 1988; Fitter, 1990; cf. Bidartondo et al., 2002).
In summary, this study demonstrates that arbuscular mycorrhizal fungi play a key role in grassland by influencing plant productivity and promoting plant nutrition, plant survival, soil structure and soil stability. It builds on our earlier work, in which we showed that AMF and AMF communities influence plant community structure and ecosystem productivity (van der Heijden et al., 1998a). Several of the community characteristics affected by AMF are important for ecosystem functioning (Loreau et al., 2001), pointing to the ecological significance of AMF. This study also suggests that indirect effects of AMF on ecosystems (e.g. improved soil structure) could change plant community dynamics. Recent work shows that AMF can increase plant P supply, irrespective of increased plant production (Smith et al., 2004). Our work shows that such results also apply to complex plant communities consisting of plant species with a diverse range of nutrient-acquisition strategies. The fact that different AMF provided different benefits in different growing seasons suggests that plant–fungal interactions are even more complex than previously shown. Our work points to the significance of microbial diversity, and emphasizes that ecological patterns seen aboveground are, at least in part, driven by belowground interactions (Klironomos, 2002; Wardle, 2002; De Deyn et al., 2003; Reynolds et al., 2003; Vandenkoornhuyse et al., 2003; Bardgett, 2005; Bonkowski & Roy, 2005; Stinson et al., 2006; van der Heijden et al., 2006). These finding are also relevant for sustainable agriculture, where soil organisms may be useful for plant production (Mäder et al., 2002).
We would like to thank Matthias Brunner, Marcel Gutter, Carla Langlotz, Dominik Refardt, Atilla Regös and Rolf Studer for practical assistance. The reviewers provided many helpful suggestions. This research has been supported by a grant from the Dutch Science Foundation (grant 016.001.023; Vernieuwingsimpuls, 2000) and by the Priority Program Environment from the Swiss National Science Foundation.