‘… what is the optimal sampling scheme to sufficiently characterize a community of AM fungi? Plant community ecologists have explored an analogous question for decades; mycorrhizal researchers must now do the same.’
Quantitative advance: DNA-based detection methods
Our ability to address community assembly in non-cultivable microscopic organisms has been greatly enhanced by high throughput DNA-based methods, which make it possible to gather sufficient information about the occurrence of such organisms. That said, few studies have sought to describe the AM fungal communities associating with a considerable proportion of a plant community: those of Davison et al. (2011) based on cloning and Sanger sequencing (212 root samples from 11 plant species, total of 2486 sequences) and Öpik et al. (2009) using 454-sequencing (10 composite root samples of 10 plant species, 126 655 sequences), and it is unsurprising that these data were re-analysed in the two new studies. Recurring re-analysis of these data indicates the hunger of the research community for suitable datasets, and the existing information has now been impressively complemented by the study of Montesinos-Navarro et al. (103 samples from 35 plant species, 1523 sequences). A challenge in future studies is to reach even larger species pools, congruent with the high diversity of most plant communities.
Qualitative advance: network analysis
What can network analysis tell us that is new about mycorrhizal communities? First, network analysis addresses the structure of symbiotic interactions at the community level. Second, network analysis aims to detect and generalize patterns of interactions, and describes these using standardized metrics that allows independent case studies to be compared (as done by Montesinos-Navarro et al.). Third, by measuring network structure, the mechanisms shaping communities can be inferred; a potential application that is nicely reviewed by Fontaine et al. (2011) and Chagnon et al. In the field of AM research, description of a fungal community interacting with the entire plant community (or a significant proportion of it) represents a conceptual leap forward – until now, most studies have only considered the AM fungal communities associating with one or several host plant species at a particular location (cf. Dumbrell et al., 2010; Öpik et al., 2010; Caruso et al., 2012).
How to sample?
Regardless of the analytical method employed, the information yielded by datasets, and their comparability, is still inevitably influenced by sampling design, and this point deserves particular attention. Different levels of nestedness in plant–AM fungal networks were observed in the Mexican semi-arid sites and the Estonian forest (Montesinos-Navarro et al.; nestedness is the trend of specialized species to associate with partners that form well-defined subsets of the partners with which non-specialized species associate). Could this result have been influenced by the sampling design? At the Mexican site, the number of plant individuals sampled per species reflected species abundance in the field, and 16 of the 35 species were represented by a single individual. It is reasonable to imagine that the number of plant–fungal interactions would be lower for infrequently sampled plant species, resulting in more observed ‘specificity’ in the network. However, the Estonian plant community was sampled using factorial design so that an approximately equal number of samples per plant species was obtained. Furthermore, only the most common species in the community were sampled. If the rare plant species at this location – possibly interacting with rare or specific fungi – had been sampled, would the network structure have been more similar to that observed in the Mexican site? This poses an important methodological question that needs to be resolved: what is the optimal sampling scheme to sufficiently characterize a community of AM fungi? Plant community ecologists have explored an analogous question for decades (Greig-Smith, 1964 and onward); mycorrhizal researchers must now do the same.
At a conceptual level, a conundrum emerges in network analysis terminology: the use of the terms ‘specialist’ and ‘generalist’. Both studies detect some level of nestedness in plant–AM fungal interactions, as expected for a mutualistic network (Fontaine et al., 2011). This means that for each partner type there are species that associate with a large (interaction generalists, IG) or a small number of partners (interaction specialists, IS), and that, on each side, IS species interact with a subset of those partners interacting with IG species. This also means that interactions between IS and IG partners prevail in the network (Fig. 1). The re-analysed datasets provide information about the habitat preferences of the plants and fungi. The emerging picture is that IS plants and their IG fungal partners are both generalists in terms of habitat preference (habitat generalists, HG; i.e. an HG–HG relationship). On the contrary, the IG plants are habitat specialists (HS), and their partners include most fungi in the species pool (HS – (HS + HG) relationship). Thus, specialists in terms of partner choice in a specific network may be generalists in other terms (e.g. habitat preference), and – not to be forgotten – they may even be IG in other places and times. Therefore, context should be indicated when using the terms ‘specialist’ and ‘generalist’ in order to ensure the intended comparisons are made. However, on the positive side, by distinguishing specialization in terms of biotic interactions and towards different environmental factors, we can gain more detailed insights about the ecology of studied organisms.
Network structure: what does it tell us?
Montesinos-Navarro et al. showed that plants exhibit stronger nestedness than AM fungi. This means that the proportion of IS species in the community was higher for plants. Such a pattern was observed in both the Mexican and Estonian sites. What mechanisms might be generating such an interaction pattern? Does this mean that plants can afford to be more selective towards their partner AM fungi than the fungi towards the plants? Considering the obligate nature of the symbiosis for the fungal partner, but not for most of the plant species, this could be a plausible conclusion. Or is it related to the fact that, although the actual number and delineation of Glomeromycota species remains (and needs) to be clarified, there are c. 1000 times more AM-plant than AM-fungal species (Hodge et al., 2010)? Chagnon et al. highlight the phylogenetic and evolutionary directions that should be investigated in order to better understand such patterns.
If observed network structure is significant and not caused by random processes (Fontaine et al., 2011), we need to know which traits are shared by IG or IS plants and fungi or partners belonging to the same module. As we have seen earlier, Estonian IS plants and IG fungi share the trait of being HG. It would also be interesting to know the habitat preferences of the Mexican plant and fungal species. Future studies can establish the generality of IS plants being HG and interacting with HG fungi, but some additional support is already provided by the observation of an invasive plant also exhibiting specialism towards HG fungi (Moora et al., 2011). Looking more closely at the IG fungi in the Öpik et al. (2009) dataset, we see that the most common among them belong to Glomeraceae and include species within the unsolved Glomus intraradices irregulare group. The global distribution of these molecular operational taxonomic units (MOTUs) is remarkably wide (for distribution maps see Fig. 4 in Öpik et al., 2010), and individual case studies show that at least some of them are locally abundant. By contrast, the IS fungi, for example those in modules 2 and 3 of Chagnon et al., include species in the families Acaulosporaceae, Diversisporaceae, Gigasporaceae, and some Glomeraceae. These MOTUs tend to be less abundant locally (Öpik et al., 2009) and less widespread globally (Öpik et al., 2010). This suggests a tendency for IG fungi to be widespread and locally abundant, and for IS fungi to be narrowly distributed and less abundant. Whether these patterns are artefactual – caused by low sampling intensity recording apparent, but not true rarity – or related to life history traits of fungi, such as hyphal growth rate and root colonization rate, sensitivity to disturbance, colony size, etc. remains to be disentangled.
Moreover, the extent to which AM symbiosis in general, and co-occurrence of certain plant and fungal partners in particular, shapes global vegetation patterns remains a poorly studied question. In order to address it we need to look at the relationships between the functional traits of the interacting species. While characterization of plant functional traits is already relatively advanced (e.g. Kattge et al., 2011), systematic characterization of AM fungal functional traits is still required as suggested by Chagnon et al.
Unfortunately, many of the fungi detected in DNA-based studies are not identified to the species level (they bear MOTU codes at best). An aspect inhibiting linkage of ecological patterns with autecological species information is the fact that a large proportion of cultured species have not been sequenced as already suggested in this journal by Hibbett et al. (2009). We thus need to characterize simple functional traits for cultured Glomeromycota species, and sequence more species that are maintained in culture collections. Furthermore, potential functional traits of AM fungi can be characterized using protein-encoding genes (Gamper et al., 2010), fungal population structure and size of individuals can be described (Stukenbrock & Rosendahl, 2005), and genomics/proteomics can be used to provide culture-independent information about fungal traits (Tisserant et al., 2012).
Among fungal ecologists there is tangible impatience to move beyond descriptions of diversity patterns in mycorrhizal fungi towards explaining these patterns with predictive power. In this issue of New Phytologist, two studies provide evidence that AM fungal–plant networks are nested, as are other mutualistic networks, and that they can contain some degree of modularity. In other words, there are fungi and plants in the community interacting with a subset of potential partners (nestedness), and there can be reciprocal specialization between AM plants and fungi (modularity). However, in order to explain the meaning of these patterns, there remains a need to better link community-level and autecological information about AM fungi and plants. This can be achieved by sequencing more cultured species (to aid identification), describing functional traits of cultures, describing population properties, and applying omics, comparative and phylogenetic methods to gain functional information about Glomeromycota without relying on cultures. Last but not least, information about AM fungal and host plants’ traits have to be linked to better understand the functioning of mycorrhizal communities.
The authors would like to thank M. Zobel, J. Davison and M-A. Selosse for helpful comments and discussion. The overviewed work of the authors is funded by Estonian Science Foundation grants 7738 and 9050, targeted financing grant SF0180098s08 and the European Regional Development Fund (Centre of Excellence FIBIR).