Molecular analysis of arbuscular mycorrhizal fungi colonising a semi-natural grassland along a fertilisation gradient

Authors

  • Juan C. Santos,

    1. Department of Cryptogamic Botany, Swedish Museum of Natural History, PO Box 50007, SE-104 05 Stockholm, Sweden;
    2. Department of Forest Mycology and Pathology, Swedish University of Agricultural Sciences (SLU), PO Box 7026, SE-750 07 Uppsala, Sweden
    Search for more papers by this author
  • Roger D. Finlay,

    1. Department of Forest Mycology and Pathology, Swedish University of Agricultural Sciences (SLU), PO Box 7026, SE-750 07 Uppsala, Sweden
    Search for more papers by this author
  • Anders Tehler

    1. Department of Cryptogamic Botany, Swedish Museum of Natural History, PO Box 50007, SE-104 05 Stockholm, Sweden;
    Search for more papers by this author

Author for correspondence: Juan C. Santos Tel: +46 8 51954161 Fax: +46 8 5154221 Email: juan.santos@nrm.se

Summary

  • • The community of arbuscular mycorrhizal fungi (AMF) colonizing the roots of Festuca pratensis and Achillea millefolium was characterized in a Swedish pasture at different times, along a gradient of fertilization.
  • • The small subunit ribosomal RNA gene was subjected to PCR and denaturing gradient gel electrophoresis (DGGE), sequencing and phylogenetic analysis.
  • • The sequences found in this study clustered in 10 discrete sequence groups, seven belonging to Glomus, two to Scutellospora and one to Diversispora.
  • • A negative correlation was observed between soil mineral nitrogen and the number of AMF sequence groups in the roots. The frequency of occurrence of AMF in roots decreased dramatically between June and September. No plant-host specificity could be detected.

Introduction

Most land plants have root systems that form symbioses with arbuscular mycorrhizal fungi (AMF) belonging to the phylum Glomeromycota (Schüßler et al., 2001b). These fungi are obligate biotrophs and the mycorrhizal associations they form have a variety of possible effects on their plant hosts. Increased uptake of phosphorus and other growth-limiting nutrients, such as nitrogen, has traditionally been considered the main benefit for mycorrhizal plants. Recent studies have shown that AMF can also play important roles such as the reduction of pathogen infections (Newsham et al., 1995a; Borowicz, 2001), improvement of water relations (Davies et al., 1993), and limiting the uptake of heavy metals (Leyval et al., 1997).

So far, fewer than 200 species of these fungi have been described. Because of the widespread distribution of such a relatively low number of AMF species among a large number of host species, potentially as many as 250 000, fungal specificity has traditionally been considered to be low. However, evidence is now accumulating that challenges this assumption (Fitter, 2005). Description of new AMF species has mainly been based on spores obtained with trap cultures in the field using a reduced number of plant species. These AMF species are usually less environmentally demanding and less host-specific than naturally occurring taxa, and thus are less representative of the indigenous flora of natural soils (Miller et al., 1985). The application of molecular tools to the identification of AMF in the field, particularly in plant roots, has revealed a hidden diversity with many detected sequences that cannot be related to known taxa (Helgason et al., 1999; Vandenkoornhuyse et al., 2002). As morphological identification of AMF species based on hyphal morphology is highly imprecise, and the spore bank in the soil does not reflect the distribution of species in roots (Clapp et al., 1995), the molecular identification of the fungi in planta remains the most realistic approach with which to carry out ecological studies.

Since the first molecular study on AMF (Simon et al., 1992), and especially since the development of primers with improved success in specific amplification of the glomalean SSU rRNA gene (Helgason et al., 1998; Redecker, 2000), a large number of publications on AMF molecular ecology have appeared. A diverse range of molecular techniques have been applied to the study of AMF, including polymerase chain reaction–restriction fragment length polymorphism (PCR–RFLP, Helgason et al., 1999; Daniell et al., 2001; Husband et al., 2002a; Vandenkoornhuyse et al., 2002); single stranded conformation polymorphism (SSCP, Kjoller & Rosendahl, 2000; Nielsen et al., 2004); terminal (t)-RFLP (Vandenkoornhuyse et al., 2003); denaturing gradient gel electrophoresis (DGGE, Kowalchuk et al., 2002; Öpik et al., 2003; de Souza et al., 2004, 2005; Ma et al., 2005); and temperature gradient gel electrophoresis (TGGE, Cornejo et al., 2004). The DGGE approach has recently been widely applied in microbial ecology as a sensitive and rapid technique for profiling microbial communities (Muyzer & Smalla, 1998). This method has been shown to be a viable alternative in the study of AMF communities in comparison with the more traditional approach using cloning (Kowalchuk et al., 2002), but has seldom been considered in ecological studies. In this study we aimed to evaluate the power of DGGE in describing the intraradical AMF community in a seminatural Swedish grassland.

Glomalean fungi not only affect plants at the individual level; they can also influence the plant community as a whole. Experiments in laboratory mesocosms (Grime et al., 1987) suggest that mycorrhizal symbiosis may promote seedling survivorship and species richness of grassland communities; more recent studies in mesocosms and field plots (van der Hejden et al., 1998) suggest that the belowground diversity of arbuscular mycorrhizal fungi may influence both the diversity and productivity of plant communities. The composition and diversity of the plant community have also been shown to influence the structure of the AMF community (Burrows & Pfleger, 2002; Johnson et al., 2003).

The traditionally managed grasslands in the mainland of Sweden are located on noncalcareous soils and dominated by perennial plant species, most of them clonal (Glimskär, 1999). Changes affecting the plant community structure can be expected to influence the community structure of AMF colonizing roots. The increasing use of fertilizers in seminatural grasslands, together with changes in management, have produced a dramatic decrease in plant diversity in northern Europe (Austrheim et al., 1999). Root colonization and sporulation of AMF have been reported to be negatively affected by increasing amounts of fertilizer (Jensen & Jakobsen, 1980). The importance that AMF can have for population size and variation in clonal plants is demonstrated by Streitwolf-Engel et al. (2001).

Very few studies have focused on AMF in Northern Europe. Nielsen et al. (2004) characterized AMF species colonizing roots of aquatic plants in southern Sweden, and Öpik et al. (2003) studied AM communities colonizing roots of Pulsatilla species in Estonia, but we are not aware of any other molecular studies of AMF in boreal plant communities. In the present study we characterize the composition and changes in the AMF community in a grazed grassland along a gradient of fertilization and in relation to seasonal variation. Host-plant preference of AMF has been reported to occur between different AMF taxa (Helgason et al., 2002; Vandenkoornhuyse et al., 2002, 2003; Gollotte et al., 2004), and root architecture has been hypothesized to be related to the presence of AMF species with potentially different functions (Newsham et al., 1995b). In order to test for possible host specificity and to detect a larger number of AMF taxa, we analysed two common plant species with a widespread distribution in the pasture: Achillea millefolium, a herb with a stoloniferous, poorly branched root system and Festuca pratensis, a grass with fasciculated, highly branched roots.

Materials and Methods

Sampling and DNA extraction

Root and soil samples were collected from a grazed field with a scattered occurrence of shrubs (Juniperus communis L.) at Hönsgärde in Uppland, Sweden (59°45′ N, 17°57′ W). The site includes a protected graveyard from the Bronze Age and has a flora typical of the grasslands in mid-Sweden, hosting some threatened plant species and many red-listed fungal species. During recent years part of the field has been fertilized, resulting in a gradient of soil N and P concentrations. The soil is a sandy loam, pH 5.1–6.2.

Plots, each 1.5 m diameter, were established along four different transects running from more fertilized parts in the west side of the field to less fertilized in the east. Five plots were sampled along each transect (20 plots altogether). The transects were 40 m long and ran parallel to each other with a separation of approx. 6 m. One plant of both Achillea millefolium L. (Asteraceae) and Festuca pratensis Huds. (Poaceae) was collected from each plot in June (spring) and September (autumn) 2003. Mineral N and soluble P were measured in the soil around the roots sampled in June. Soil parameters were measured using the methods of Mulvaney (1996) for N and Egnér et al. (1960) for P at the Soil Fertility and Plant Nutrition Department, SLU in Uppsala.

The roots were rinsed thoroughly in tap water and cut into 1-cm pieces. For each sample, 0.1 g root material (representing approx. 5–8 cm root length) was placed in a 2-ml screw-capped propylene tube with half the volume filled with 2.5 mm zirconia–silica beads (Biospec Products Inc., Techtum Lab, Umeå, Sweden). The tubes were shaken at 6500 rpm for 30 s in a Mini-BeadBeater (Biospec Products Inc.) to homogenize the roots. DNA was extracted using a DNeasy Plant Kit (Qiagen, Crawley, UK) with a final elution volume of 100 µl.

PCR and DGGE analysis

PCR amplifications used the AM1 primer (Helgason et al., 1998) and the NS31-GC primer described by Kowalchuk et al. (2002), which corresponds to the universal eukaryotic NS31 primer (Simon et al., 1992) plus a GC-clamp sequence.

A PCR amplification kit (puRETaq Ready-To-Go, Amersham Biosciences, Uppsala, Sweden) was used with a final reaction volume of 25 µl. As template, 2 µl extracted DNA was used in all reactions. The thermocycling program used was as follows: 94°C for 2 min; 35 cycles of 92°C 30 s, 60°C 60 s, 72°C 45 s (with an extension of 1 s per cycle); and 72°C for 5 min. Reaction mixtures were overlaid with a drop of mineral oil and run in a Perkin Elmer 480 thermal cycler (Perkin-Elmer, Norwalk, CT, USA). Reaction yields were estimated in a 1.2% agarose gel, after 10 min staining in a 0.5 mg l−1 ethidium bromide bath and 10 min destaining in MilliQ water (Millipore B. V., Etten-Leur, the Netherlands).

The PCR products were analysed using DGGE with a DCODE Universal Mutation Detection System (Biorad, Sundbyberg, Sweden). The gels were cast with a manual gradient former (Model 475, Biorad) with the following characteristics: 6% (wt : vol) polyacrylamide (37.5 : 1 acrylamide-bis-acrylamide), 1 mm thick, 16 × 16 cm. A linear gradient of 25–40% denaturant was used, where 100% denaturant acrylamide is defined as containing 7 m urea and 40% formamide. Before polymerization, a 10-ml top gel with no denaturants was added in order to obtain well polymerized gel slots. Electrophoresis was run in 0.5 × TAE buffer (40 mm Tris–acetate, 1 mm ethylenediaminetetraacetic acid) for 16 h at 65 V and 60°C. Gels were stained in MilliQ water containing 0.5 mg l−1 ethidium bromide for 10 min and destained in MilliQ water for another 10 min before UV transillumination.

Recovery of DNA after DGGE

All the detected DGGE bands were excised from the acrylamide gel for DNA isolation and put in 2-ml screw-capped polypropylene tubes with approx. 0.5 ml 2.5 mm zirconia–silica beads (Biospec) and 250 µl sterile MilliQ water. In order to disrupt the acrylamide gel pieces, vials were shaken with a homogenizer (Fast Prep 120, Biospec) at 5500 rpm for 20 s, kept on ice for 1 min and then shaken again in the same way. The supernatant was then centrifuged briefly at low speed to pellet the acrylamide. A volume of 2 µl supernatant was used for a further PCR amplification using the same primers and cycling as above. Representative bands were run again in a new DGGE to confirm band migration.

Sequencing

Before sequencing, PCR products were purified with the QiaQuick PCR purification kit (Quiagen) with a final elution volume of 30 µl. Sequencing reactions were carried out in a PRISM 310 (Applied Biosciences) automatic sequencer as described in the manufacturer's instructions, using both AM1 and NS31 (Simon et al., 1992) as sequencing primers. Sequences were registered in the European Molecular Biology Laboratory (EMBL) database under accession numbers AJ866185–270.

Data analysis

A search for sequences similar to those from this study was done with the blast tool (Altschul et al., 1997) provided in GenBank. Sequences with high similarity and a wide selection of AMF taxa, including representatives of the major clades described by Schüßler et al. (2001b), were downloaded. A multiple alignment was performed using the ClustalX program (Thompson et al., 1997). The result was refined by eye and confronted against the alignment constructed by Schüßler et al. (2001b) deposited in the EMBL database (accession number ALIGN_000208). Endogone pisiformis Link and Mortierella polycephala Coem. were used as outgroup, as both have been shown to belong to a possible sister group to the ‘Symbiomycota’, in which the phylum Glomeromycota is included (Tehler et al., 2003; Lutzoni et al., 2004). The region with most variation, and thus with most relevant phylogenetic information, lies 70–300 bp from the 5′ end of the amplicon. That region was used in the following steps of the phylogenetic analysis.

Maximum parsimony (MP), as implemented in paup* ver. 4.0b10 for Macintosh (Swofford, 2002), was used to analyse phylogenetic relationships. A heuristic search with 500 replicates and random addition of sequences was performed using tree bisection reconnection as swapping algorithm, with the MulTrees in effect and saving no more than 500 trees at each replicate. Gaps were treated as missing data, and all characters were unordered and of equal weight. An MP bootstrap analysis (1000 replicates) was performed with 50 random taxon additions per bootstrap replicate and the MulTrees option in effect to save no more than 10 trees at each replicate of random addition sequence. A parallel bootstrap analysis was carried out using the neighbour-joining (NJ) method (1000 replicates) with the Kimura two-parameter substitution model.

Generalized linear models were used for analysis of the relationships between explanatory variables (host, season, N and P) and presence of AMF sequence groups using the Poisson distribution and the log function as the link. The analysis was carried out with the software sas ver. 8. In the first model, effects of season and hosts were investigated. As measurements of N and P were measured only in June, models to study the effects of these parameters as explanatory variables were restricted to the occurrence of AMF in June. As the model includes both N and P, the tests for one explanatory variable at a time are adjusted for the other.

Results

PCR and DGGE analysis

A total of 80 roots was subjected to DNA extraction. PCR products of the expected size (550 bp plus 39 bp from the GC tail of the clamped primer) were obtained from a total of 58 samples (72.5% of roots). All the bands identified using DGGE were excised and reamplified. Only 13 bands were so weak that they could not be reamplified. The observed bands that produced sequences of glomalean origin showed a migration range between 33% and 36% of denaturant in the gradient under the electrophoretic conditions used. Two bands produced sequences of basidiomycete origin (99% similarity with isolate Eocronartium muscicola, AY123323 and 97% with Asterophora parasitica, AJ496255, respectively, using a BLAST query in GenBank). These two nonglomalean bands were easily distinguished in the gels as they migrate further in the gel, thus appearing outside the normal denaturant gradient range for glomalean bands.

Certain lanes in some gels produced double bands. This is a well known and common problem when using DGGE, where two bands with the same sequence appear very close to each other (Janse et al., 2004). Although double bands produced very similar sequences, only one of the bands produced a sequence of good quality. Only nonredundant sequences within the same sample and sequences of good quality were included in the phylogenetic analysis and in the analysis of the AM community structure.

The 80 root samples produced a total of 86 bands with sequences of glomalean origin. Bands producing similar sequences showed identical positions in the acrylamide gels (Fig. 1c). Although some groups produced patterns with nearly identical positions in the gel, the five most common sequence types could be discerned easily by their migration patterns.

Figure 1.

Denaturing gradient gel electrophoresis (DGGE) gels of PCR products from the small subunit rRNA from arbuscular mycorrhizal fungi (AMF). (a) Migration patterns of single bands isolated from field samples representing all 10 different detected phylotypes. (b) Schematic drawing of DGGE patterns from (a). (c) Banding patterns from root samples with different AMF communities. Lane 6 shows no PCR product. (d) Schematic drawing of the DGGE pattern from (c). Labelled bands are those that could be reamplified and sequenced. Faint bands that failed to be reamplified or produced bad quality sequences are not labelled.

Phylogenetic analysis

The 290 characters of the alignment had 121 characters that were parsimony-informative. The analysis found 51 956 different trees as short as 576 steps. The MP and NJ analyses produced trees with basically the same topology (Fig. 2). This topology is largely in agreement with those published previously (Redecker, 2000; Schüßler et al., 2001a; Vandenkoornhuyse et al., 2002; Öpik et al., 2003; Tehler et al., 2003). All the families from the phylum Glomeromycota received high support in the analyses, except for Archaeosporaceae and Glomeraceae, where Glomus group A and Glomus group C appear paraphyletic. The 86 sequences analysed were clustered in 10 discrete groupings or phylotypes with support in the MP and NJ bootstrap analysis (> 80% in at least one of the analyses). The pairwise sequence similarities within the groupings vary from 97.5 to 100%.

Figure 2.

Phylogram representing one of the most parsimonious trees showing the different sequences of arbuscular mycorrhizal fungi (AMF) obtained from band isolates using a PCR–DGGE (denaturing gradient gel electrophoresis) analysis. Two Zygomycetes (Endogone pisiformis and Mortierella polycephala) are used as outgroup taxa. Shadow areas indicate diverse taxonomic groups as described by Schüßler et al. (2001a). Bootstrap values above branches are from the maximum parsimony analysis (1000 bootstraps) and under branches from the neighbour-joining analysis (1000 bootstraps); these are shown only when > 70% in at least one of the analyses. Sequence groups on the right (e.g. Glo8, Scu1) identify clusters with sequence similarity > 97.5%. Band identifiers from this study are in bold type and relate to host (A, Achillea; F, Festuca); time (j, June; s, September); plus a sequence number. Bands with identical sequences are represented by ▪, ▴ or • as follows: ▪ in Glo8: Aj101, Aj124, As143; Fs011; Glo10: Aj031, Aj099, As135, As145, Fj006, Fj009, Fj050, Fj051, Fj057, Fj066, Fj074, Fj082; Glo3: Aj022, Aj042, Aj092, Aj097, Aj103, As024, Fj002, Fj048, Fj049, Fj059, Fj080, Fs003, Fs083; Glo2: Aj095, Aj114, As041, Fj020, Fj073, Fj076; Glo4: Aj025, Aj027; Scu2: As 138, Fs146. ▴ in Glo8: Aj028, Aj094, Aj112, Aj149, As030, As139, Fj007, Fj008, Fj055, Fj056, Fj062, Fj068, Fj078, Fs085, Fs088; Glo2: Aj104, Fs090; • in Glo2: Aj100, Aj126. Values to the right show numbers of root samples containing the particular sequence groups.

Seven of these sequence groups belong to the genus Glomus, two to the genus Scutellospora, and one to the Diversispora. Most of the sequence groups found show high similarity to previously described root-derived sequences belonging to unknown glomalean species. Only Glo8 is clearly related to the species complex Glomus intraradices/Glomus fasciculatum. The sequence type Scu1 shows high similarity to Scutellospora dipurpurescens, but this relationship is not supported in the analysis. Both Glo3 and Div1 are similar to isolates where spores are known and have been sequenced, but not determined taxonomically.

Analysis of community structure in roots

Five of the groupings (Glo8, Glo10, Glo3, Glo2 and Glo4) accounted for 93% of the observed sequences. All occurred in both plant species. All five sequence groups could be detected in both June and September, except for Glo4, which could not be detected in September. Up to three different groups could be detected in a single root sample.

No patterns of AMF specificity were obvious with respect to the two host-plant species investigated in the present study. Some rare phylotypes appeared in just one of the host species, but they occurred at such low frequencies that no conclusions can be drawn from this data set.

Sequences of AMF were detected in 90% of June samples, but only in 55% of the samples taken in September (Fig. 3). The season effect is significant (P < 0.05) in the statistical analysis (Table 1a).

Figure 3.

Arbuscular mycorrhizal fungal (AMF) community in roots of Festuca and Achillea in June (spring) and September (autumn). Histogram represents total number of detected bands in the PCR–DGGE (denaturing gradient gel electrophoresis) analysis that produced sequences belonging to the five most common phylotypes.

Table 1.  Analysis of parameter estimates from generalized linear models: models using host and season consider data from both spring and autumn; models using mineral nitrogen and phosphorus as independent variables consider only data from spring
Variabledfχ2P
  • *

    , Statistically significant at P < 0.05.

Host10.070.8
Season16.610.01*
Host × season10.030.85
Mineral nitrogen14.970.03*
Phosphorus10.510.47

The soil parameters analysed along the fertilization gradient showed values ranging from 19 mg soluble P and 4.5 mg mineral N per kg soil to 170 mg P and 300 mg N per kg soil at the more fertile end of the gradient (Fig. 4).

Figure 4.

Relationship between soil parameters and number of observed arbuscular mycorrhizal fungal (AMF) sequence groups per sample in June (spring). (a) Total mineral nitrogen; (b) soluble phosphorus. See Table 1 for statistical analysis of the relationships between the variables.

A significant negative correlation between the frequency of observed sequences per root sample and the levels of mineral N was recorded in June (P < 0.05) (Table 1b). High levels of N in the soil were consistent with the distribution of nitrophilic vegetation in the pasture. The number of sequence groups per root did not show any significant correlation with levels of P, although it showed the same decreasing tendency as in the case of N (Table 1b).

Discussion

The diversity of glomalean fungi

This study is one of the first using PCR–DGGE to characterize ecological changes in AMF communities in grasslands, and is the first molecular survey of AMF in a terrestrial ecosystem in Sweden. The only two other studies using DGGE to study glomalean communities have characterized the AMF-colonizing roots of Ammophila arenaria (Kowalchuk et al., 2002) and of two closely related species of Pulsatilla (Öpik et al., 2003), in the latter case combining DGGE with cloning. The number of detected sequence groups revealed in our study is similar to that found in similar biotopes using molecular methods. Öpik et al. (2003) found a total of 14 sequence groups by carrying out pot experiments and field sampling in a dry meadow in Estonia. Scheublin et al. (2004) found 15 sequence groups by looking at five plant species in a dune grassland in Holland. Vandenkoornhuyse et al. (2002) found an unexpectedly high number of sequence groups, 24 in total, with an intensive sampling of two plant species in a grassland in Scotland.

The composition of the AMF community in our study is dominated by taxa belonging to the genus Glomus. This seems to be a common pattern in similar studies, with many Glomus sequences, and just a few or none belonging to Scutellospora and Acaulospora (Fig. 2). No sequences of species belonging to the family Acaulosporaceae were detected in this study, although Acaulospora sequences have been detected in a parallel study in the same pasture (J.C.S., unpublished data). As expected, no sequences from representatives belonging to the families Paraglomaceae or Archaeosporaceae were detected either, as the AM1 primer does not amplify sequences from these two divergent families (Redecker et al., 2000). This primer also mismatches single priming sites in at least some taxa belonging to Glomus group B (Husband et al., 2002a). No taxa belonging to this group were observed in our study. This could be caused by its absence from our samples; however, it is rarely represented in studies using the AM1 primer and appears to be better represented in AMF studies using other primer combinations (Wubet et al., 2003a; 2003b; Nielsen et al., 2004; Uhlmann et al., 2004).

The five most abundant sequence groups in our study seem to represent quite ubiquitous AMF taxa observed in a range of diverse habitats. For example, sequences closely related to the group Glo8 (G. intraradices/G. fasciculatum) have been found in a tropical forest in Panama (Husband et al., 2002a; 2002b); a wetland in Germany (Wirsel, 2004); a natural dune grassland in Holland (Scheublin et al., 2004); a grassland in Estonia (Öpik et al., 2003); and arable land (Daniell et al., 2001) and temperate forest (Helgason et al., 2002) in Britain. Most of the studies cited above have also found several of the other common Glomus sp. groupings described here. It is worth noting that the community most similar to that in our study, in terms of sequence similarity of sequence types, is that described from grasslands in Estonia by Öpik et al. (2003), which, apart from being a very similar ecosystem, is also geographically closest. Most sequence groups detected have been described only from roots, and it has been suggested that no spore production occurs in these species and that they rely on vegetative strategies for colonization and dispersal (Clapp et al., 1995). This could be true, but it also has to be considered that most of the taxonomically characterized AMF sequences in databases come from a handful of cultivable isolates, and only a few field studies have attempted to both sequence and taxonomically characterize spores.

PCR–DGGE analysis of AMF communities

The banding patterns of the different sequence groups in our study agree with those reported in the literature (Kowalchuk et al., 1997; Öpik et al., 2003; Ma et al., 2005). The five most common phylotypes showed a distinct banding pattern that permitted identification before sequencing. Amplicons of nonglomalean origin produced bands that migrated beyond the typical glomalean range, in agreement with the results of Ma et al. (2005). DGGE has usually been considered a straightforward method that can be used to describe diversity, circumventing tedious cloning and sequencing, but analysis of banding patterns alone in environmental samples can easily lead to an artificial overestimation of the real diversity in the community. Excision and sequencing of bands is always recommended as a control procedure to check for possible artefacts (Ferris et al., 1996), such as the presence of double bands (Janse et al., 2004). This was probably the case in some of our bands that did not produce usable sequences. Some other bands produced ambiguous nucleotide signatures at single positions, probably because they harboured two very similar sequences with the same melting profile. This probably has no effect on the final results, as it would not affect the number of sequence groups detected in the same sample.

Seasonal variation

Previous studies have shown that AMF community composition can change considerably throughout the year (Husband et al., 2002b; Vandenkoornhuyse et al., 2002). In our study, the frequency of AMF sequences in the roots decreased dramatically in the September sampling, when we were also able to detect more infrequently occurring sequence types. This could be caused by a shift in community composition, but could also be caused by the absence of dominant spring sequence types that would otherwise obscure the detection of less common sequences or of weaker bands in the acrylamide gels.

Host specificity

In our study we could not see any clear pattern of host specificity, as the most common glomalean sequence groups were present in both plant species. The PCR–DGGE approach used in this study can only characterize the AMF community in terms of presence or absence, and this may mask quantitative differences in colonization of different plant species by particular AM taxa. No information is available about the relative abundance of different fungal taxa in the roots. The only possible way of gaining this with the present method would be to measure relative band intensity in each gel lane, but this is unlikely to yield reliable estimates as it would require untested assumptions, for instance that the fungal DNA is equally likely to be extracted and amplified in different species, and that those species have equivalent number of copies of the rRNA gene. In order to test whether certain plant species are preferentially colonized by particular glomalean taxa, a method able to measure relative abundance of fungal taxa would be required. Cloning has been used to do this, by considering that the percentage of a clone with a particular sequence type in a clone library reflects the relative abundance of that particular AMF sequence type in an environmental sample. Apart from the assumptions mentioned above when using DGGE, further assumptions are required when using cloning. These are that all the DNA is equally ligated and transformed in the cloning process (Helgason et al., 1999; Scheublin et al., 2004). Quantitative, real-time PCR may ultimately provide a solution to quantification of AMF dynamics under field conditions, but development of such methods will require careful calibration for different taxa.

Although we do not measure the degree of root colonization in the present study, the reduced presence of glomalean sequences in autumn probably reflects changes in the amount of root colonization. Seasonal changes, with high colonization levels in spring followed by lower levels in autumn, have been found by Titus & Leps (2000). In another study of a seminatural grassland, also in mid-Sweden but using microscopic quantification of total colonization instead of molecular methods, Eriksson (2001) found no differences in colonization rates in A. millefolium between April and August. Sanders & Fitter (1992) could not find any consistent patterns of colonization, even when comparing the same community during a consecutive 2-yr period. This lack of common patterns could indicate that season is perhaps not the main factor explaining the apparent decline in the presence of AMF in our study. In any case, a more intensive sampling distributed throughout the season, with measures of both colonization and species composition, and other ecological factors such as root turnover in the host, would be needed to obtain a better understanding of the seasonal dynamics of AMF communities.

Effects of soil fertilization

The negative correlation between N in the soil and the frequency of sequences per root is in agreement with the results of Johnson (1993), showing a decrease in AMF diversity in response to fertilizer application. In our study we were not able to detect a qualitative shift in the composition of the AMF community in response to N, as shown by Jumpponen et al. (2005) in an N-fertilization experiment in a tallgrass prairie ecosystem in North America. At present it is not completely clear how increased amounts of N affect the composition of AMF in the soil. Most studies have used spore identification and quantification to monitor fertilizer-related changes and are subject to the inherent drawbacks associated with this method, as noted above. Measurements of root colonization have shown both increases and decreases, or no effect at all, in relation to increasing amounts of N in the soil.

Conclusions

Great care must be taken when attempting to establish relationships between environmental factors and the diversity of AMF, as basic knowledge concerning the genetics of the fungi is still lacking, and the sample sizes in most molecular studies are often, of necessity, small. If the true numbers of AMF taxa and the degree of specificity of their interactions with different plant species are higher than previously assumed, then more detailed knowledge is necessary to gain insight into the ecological consequences that changes in AMF diversity can have (Fitter, 2005). Detailed molecular studies of the fungi colonizing different plant species are necessary to determine whether individual AMF taxa do indeed have differential interactions with different plant host species. In the present study we were able to identify taxa colonizing A. millefolium and F. pratensis using molecular methods, but found no evidence of host-specific differences, although the fungi characterized were similar to those identified in similar community studies. The results may reflect the fact that these two plant species are colonized by similar AMF taxa, but the lack of host-specific differences may also be caused by the use of primers amplifying only a subset of the taxa present, insufficient sample size, or sampling at the wrong depth. Although we chose two species with contrasting types of root system, both are perennials and have similar general growth requirements. The choice of other, annual species with more specialized growth requirements may have revealed greater differences. We found evidence of a quantitative, but not a qualitative, effect of N on numbers of AM taxa in the roots; and a seasonal effect, consistent with other studies, with reduced frequency of occurrence of fungi in September compared with June.

Recent studies of the microdistribution of different AMF taxa in roots (Scheublin et al., 2004), at different times of year (Husband et al., 2002a; Vandenkoornhuyse et al., 2002) and at different depths in soils (Oehl et al., 2004), have shown a hidden diversity and more complex patterns in AMF communities than previously supposed. More complete monitoring of the community composition of AMF is a first step towards understanding their ecology, and requires not only the development of methodologies able to detect the whole range of AMF groups, but also more ambitious sampling strategies, both in terms of space and time.

Acknowledgements

Financial support from The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) is gratefully acknowledged. We thank Lennart Norell from the Unit of Applied Statistics and Mathematics, SLU for helping us with the statistical analysis. We also thank Ursula Eberhardt, Anders Glimskär and Andy Taylor for their helpful advice.

Ancillary