Initial effect of removal treatments
The negative effects of the disturbed control and all removal treatments on total mycorrhizal colonization 5 months after the start of the experiment suggest that the immediate effects of the treatments could be attributed to the physical disturbance associated with the initial removal operation. This is not surprising since this initial effect is one of the main artefacts of the field removal approach (Díaz et al. 2003). In the specific case of AMF, several authors have previously reported the negative relationship between physical disturbance and colonization (Jasper, Abbott & Robson 1989; An et al. 1993; Merryweather & Fitter 1998). This effect has been attributed to the fact that physical disturbance of the soil is often followed by increased mineralization of soil organic matter, which results in nutrient flushes that in turn can reduce mycorrhizal colonization (Smith & Read 2008). In this experiment, no effect of disturbance on nutrient availability (as determined by the use of ion exchange resin bags) was observed. Therefore, our results suggest that the soil and rhizosphere manipulations associated with the removal and disturbed control treatments may have directly disrupted the extraradical mycelium (Evans & Miller 1988; Read & Birch 1988). This in turn might have negatively affected intraradical AMF colonization. These results highlight the importance of including in removal experiments a disturbed control treatment that accounts for the initial side effects attributed to physical and biotic alteration of the soil caused by the act of removal (Díaz et al. 2003).
Resilience of AMF colonization in the face of disturbance
Seventeen months after the initial removals, effects of treatments or initial disturbance on AMF colonization were no longer detectable, and this was still the case after 29 months. This suggests that AMF communities are highly resilient in the face of disturbances of the canopy and the soil physical structure such as those applied in our experiment. Similarly, rapid recovery of AMF colonization after disturbance has been previously observed (Merryweather & Fitter 1998), suggesting that short-term experiments on mycorrhizal community structure should be interpreted with caution. Although these results may overlook changes in mycorrhizal composition in roots, it has been shown that changes in AMF composition in roots can be also related with changes in the percentage of colonization because AM fungi vary in the amount of root they colonize, either alone or in mixtures (Jansa, Smith & Smith 2008).
Resilience of AMF communities under different PFT treatments
Once they had recovered from the disturbance associated with the initial removal operation, no difference in AMF colonization was observed under different PFT manipulations, even under those that involved dramatic changes in above-ground biomass and alteration of soil nutrient availability in the understorey. The fact that AMF colonization indeed decreased in the first months after the initiation of the experiment and then recovered rules out the possibility that this lack of response could be attributed to resistance or inertia of AMF communities, or to inability of our measurement protocols to detect differences. It could be argued that the similar AMF colonization among different PFT treatments could be due to the capacity of the remaining PFTs to compensate for the lost biomass of the removed PFTs. However, in the same experimental setup, Gurvich (2005) found no significant above-ground biomass compensation of the removed biomass by remaining PFTs. Although it has been shown that changes in AMF composition in roots are also related with changes in the percentage of colonization because AM fungi vary in the amount of root they colonize (Jansa, Smith & Smith 2008), we cannot exclude the possibility of the occurrence of changes in mycorrhizal composition and functioning without changes in the overall rate of colonization.
Despite the general assumption that AMF are generalists, their effects on plant species vary according to different plant–fungal combinations (e.g. van der Heijden et al. 1998; Scheublin, van Logtestijn & van der Heijden 2007). In addition, differential sporulation of AMF species grown with different host plants is well documented (Johnson, Tilman & Wedin 1992; Bever et al. 1996; Eom, Hartnett & Wilson 2000; Lovelock, Andersen & Morton 2003; Lovelock & Ewel 2005), suggesting some level of specificity in this association. Nevertheless, Lovelock, Andersen & Morton (2003) found that in tropical forest, although there was a species effect on AMF spore communities, no effect of plant life history groupings was observed. Although we found some seasonal changes in Gigasporaceae spore abundance, the lack of a consistent effect of PFTs on the most abundant AMF functional groups (Glomeraceae and Acaulosporaceae), as well as the other spore variables observed in our experiment, are in line with those observations and suggest that AMF-community differences between host plants might disappear beyond the scale of host species. In other words, despite some level of host specificity or preference for some AMF functional group or taxa, the removal of certain hosts or groups of hosts (such as PFTs) seems unlikely to lead to local extinction of AMF species or to changes in the relative abundance of AMF functional groups. This is probably because in natural communities the availability of compatible hosts is highly unlikely to reach low enough levels to exclude AMF species from root colonization, except for great disturbances or drastic changes in native communities (e.g. from natural forest to crops; Helgason et al. 1998). The possibility exists that the behaviour of AMF in the face of plant removals could have been to lay down resting spores as an immediate response to the manipulations, thus explaining the lack of differences between treatments. However, if the spores indeed responded in this way, after 3 years in the soil the resting AMF spores would have germinated to reach a host or would have perished as a result of the action of the complex natural soil food web, including fungivorous invertebrates such a nematodes and collembola. In fact, some transient effects were observed: after 17 months spore evenness was significantly enhanced by all removal treatments, especially no shrubs treatment. This effect seems to be related to the biomass removed since there was a significant correlation between this variable and spore evenness. Whatever the mechanisms behind these effects, the AMF spore community showed resilience to them, since the differences were absent after a relatively short period. Moreover, no AM morphotaxon was absent in any treatment. This agrees with a recent study that imposed a different kind of disturbance to the soil fungal community (application of the fungicide benomyl), and found that no AM DNA-based taxa were eliminated under this treatment in comparison with the control (Helgason et al. 2007). Moreover, in natural communities AMF spore diversity did not consistently change during conversion of tropical forest to grassland and other types of communities (Johnson & Wedin 1997; Picone 2000; Violi et al. 2008). In grassland microcosms, Johnson et al. (2004) found no effect of plant biomass on AMF communities. In a grassland removal experiment, Wardle et al. (1999) found that after 3 years decomposer soil biota were not strongly affected by removal of several PFTs (C4 grasses, C3 annual grasses, C3 perennial grasses, clovers and dicotyledonous weeds), and showed no significant relationship with plant biomass or productivity. Our experiment suggests that this is also true in the case of plant communities consisting of more contrasting PFTs (i.e. herbaceous and woody species).
Our findings contrast with those observed in similar field removal experiments for ericoid mycorrhizas and ectomycorrhizas. In the arctic tundra of Alaska, the removal of the shrub Betula nana decreased ericoid mycorrhizal colonization, and the removal of Ledum palustre decreased ectomycorrhizal colonization and affected ectomycorrhizal morphotype composition (Urcelay et al. 2003), while in northern Arizona, the removal of AMF-infected shrubs increased ectomycorrhizal colonization even though it did not affect composition (McHugh & Gehring 2006). These results suggest that ericoid and ectomycorrhizal fungi (Phyla Ascomycota and Basidiomycota) are more responsive to changes in vegetation than AMF (Glomeromycota). This is consistent with the widely known fact that Ascomycota and Basidiomycota are considerably more host-specific than Glomeromycota (Kottke et al. 2008; Smith & Read 2008).
In our experiment, despite the lack of differences in total mycorrhizal colonization in the longer term, percentage of colonization by vesicles was not affected by initial disturbance but was consistently enhanced by removal of graminoids after 17 and 29 months of removal. Most graminoids, and the dominant Stipa eryostachia in particular, showed a low percentage of vesicle colonization (Urcelay & Battistella 2007). Therefore, the higher proportion of roots of PFTs other than graminoids present in the no graminoid plots might account for the higher vesicle colonization observed in this treatment. It is known that AMF species belonging to the Gigasporaceae do not form vesicles, therefore it raises the possibility that the observed reduction in vesicle colonization could be a consequence of a reduction of Gigasporaceae in the no graminoid plots. When we examined this possibility, Gigasporaceae showed significantly higher values in no graminoid, but also in no shrub and to a lesser degree disturbed control in comparison with control and no perennial forb.
It has been recently shown in synthetic community experiments that plant effects on soil chemical properties and microbial communities can be highly idiosyncratic. For example, plant composition affects chemical properties but not microbial communities in sandy-soil communities in the Netherlands, but shows the opposite effects in chalk soil communities in the UK (Bezemer et al. 2006). In our experiment, no effect of disturbance or treatments on phosphate availability was observed, but removal of annual forbs promoted an increase of ammonia in the shorter term while removal of shrubs promoted lower nitrate availability (Fig. 5B) and soil temperature (Gurvich 2005) in the longer term. This decrease of nitrate availability could be attributed to the absence of the nitrogen-fixing legume Acacia caven, which is the most abundant shrub (Table S1). The lack of major treatment effects on AMF communities suggests that no nutrient- or temperature-mediated effect of PFTs removals existed. Although soil properties such as nutrients (Johnson et al. 2003; Smith & Read 2008) and temperature (Heinemeyer et al. 2003; Staddon et al. 2003) are known to affect AMF colonization and community structure in synthetic experiments, the magnitude of the changes observed in our field experiment might have not been strong enough to affect the fungal variables measured.
Possible asymmetry in the mutual effects of AMF and plants
Overall, and after applying removal treatments on a native shrubland during approximately 3 years, our results do not provide consistent evidence for direct or indirect effects of different PFTs on AMF colonization or spore community parameters. The removal of PFTs with more conservative trait syndromes (e.g. shrubs; Díaz & Cabido 1997) did not affect mycorrhizal parameters in a way consistently different from that of more acquisitive syndromes (e.g. fast-growing annual forbs). This is inconsistent with the hypothesis of differential effects of plants with different resource-use syndromes (conservative vs. acquisitive; Diaz et al. 2004) on soil microbial communities. On the other hand, and except for some transient effects, the effect of different removal treatments was not directly proportional to the biomass removed, which is inconsistent with the biomass-ratio hypothesis. These results add evidence to the suggestion that below-ground organisms are generally less responsive and/or respond more slowly to changes in vegetation than above-ground organisms (Wardle 2002, 2006). This could be particularly true for those organisms that are not highly host- or substrate-specific, such as AMF. Whether the resilience showed by mycorrhizal colonization and spore composition in the soil relates to resilience of AMF composition in roots or extraradical mycelium remains unanswered. This is because there may be no direct relationship between spore diversity in the soil and AMF taxa colonizing roots (Sanders 2004).
Because root mycorrhizal colonization is a dynamic variable that mainly relates to nutrient acquisition and plant productivity on the one hand, and mycorrhizal spores relate to seedling establishment after disturbance on the other hand (van der Heijden & Scheublin 2007), the ecological interpretation of the results reported here should be framed within these processes. Considering the facts that (i) AMF range from mutualistic to parasitic (Johnson, Graham & Smith 1997; Smith & Read 2008); (ii) they differentially affect the growth of plant species (van der Heijden et al. 1998; van der Heijden, Wiemken & Sanders 2003; Maherali & Klironomos 2007; Scheublin, van Logtestijn & van der Heijden 2007) or groups (e.g. forbs, C4 grasses and C3 grasses, Hetrick, Kitt & Wilson 1988; Wilson & Hartnett 1998); (iii) their removal has been shown to affects plant community structure in the field at shorter time scales than the one considered here (Gange, Brown & Sinclair 1993; Newsham et al. 1995; Hartnett & Wilson 1999; O′Connor, Smith & Smith 2002) and (iv) functional groups of AMF can have different effects on plant growth (Maherali & Klironomos 2007), our findings suggest that the hypothesis of a possible asymmetry in the plant–AMF interactions at the community level deserves explicit testing. This reinforces the need to carry out studies on the feedback effects between plants and AMF in natural communities in order to understand above- and below-ground interactions caused by changes in land use and vegetation composition.