Introduction

Long-term development of ecosystems comprises both progressive and retrogressive stages and involves interactions between soil pedogenesis and plant communities (Walker and Chapin 1987; Wardle 2002; Richardson et al. 2004; Peltzer et al. 2010). Despite a wealth of literature on vegetation change during ecosystem development (see Peltzer et al. 2010 and references therein) there has been relatively little attention paid to the role of mycorrhizal fungi. This is somewhat surprising given that mycorrhizal fungi are a major interface between soils and most plant species, and thus act directly in both pedogenesis and the development of plant communities, as well as modulating interactions between soil processes and plants. Recognising this critical role, Professor T. Walker, a pioneer of pedogenic research along chronosequences (Stevens and Walker 1970; Walker and Syers 1976), frequently noted that he would choose to work on mycorrhizas if he were to start his research career afresh.

The processes of ecosystem development have most commonly been assessed using long-term soil chronosequences, i.e., space-for-time sites differing in age but having similar parent material, climate, and regional species-pools (Jenny 1980). The early phases of ecosystem development (i.e., typically lasting a few hundred years) have been well documented, revealing strong and somewhat predictable shifts in plant community composition, biodiversity, and ecosystem processes such as primary productivity, biomass accumulation, nutrient cycling and decomposition (Walker and Chapin 1987; Wardle 2002). In contrast, far less is known about longer term changes in ecosystems, including the later decline or retrogressive phase (Wardle et al. 2004; Peltzer et al. 2010). Several studies have examined interactions between plants and soils throughout ecosystem development and retrogression (Jenny 1980). Soil nitrogen (N) is nearly absent at the start of primary succession, but increases through biological N-fixation of atmospheric di-nitrogen and plant decomposition, and subsequently declines during ecosystem retrogression, due to diminishing abundance of N-fixers and changes to less bioavailable forms of N. In contrast, phosphorus (P) is primarily rock-derived, and is initially relatively plentiful in primary succession in forms that are readily available to plants (e.g., calcium phosphate). However, P declines during ecosystem development through leaching, transformation to more recalcitrant pools, accumulation in microbial biomass and other mechanisms (Walker and Syers 1976). For example, at the well-studied Franz Josef soil chronosequence in New Zealand (NZ), total mineral soil N (in the top 100 mm) peaks at c. 9 g kg−1 within 500 years of ecosystem development and then declines to c. 3 g kg−1 thereafter, whereas inorganic P declines from c. 800 to <100 mg kg−1 by the oldest stages studied (Richardson et al. 2004). As a consequence of these different trajectories for N and P, nutrient limitation of primary production is expected to shift from N-limitation during early succession, to co-limitation by N and P during the mature phase, and finally to P limitation during retrogression; these predictions are confirmed in nutrient concentrations in leaves, the nutrient resorption proficiencies of litter, the physiology of roots, and nutrient limitation of leaf physiological processes including photosynthesis and respiration (Richardson et al. 2005; Turnbull et al. 2005; Whitehead et al. 2005; Peltzer et al. 2010; Holdaway et al. 2011).

At the latest stages of ecosystem development, and in the absence of rejuvenating disturbance (sensu Peltzer et al. 2010), ecosystems can become retrogressive. Retrogression is driven by diminishing nutrient availability leading to declines in ecosystem processes or properties such as primary productivity, plant height and vegetation biomass. In general, there are predictable, coordinated shifts in soils and plant communities during ecosystem development, converging on nutrient-poor soils, stress-tolerant soil biota (Doblas-Miranda et al. 2008), and highly conservative plant strategies (Vitousek 1998; Richardson et al. 2004). Because soil chronosequences represent strong environmental gradients of soil properties, nutrient availability, and plant species composition and strategy, they are ideal systems for testing ideas about the long-term effects of interactions between above- and below-ground processes. A critical component of these interactions is nutrient acquisition and uptake by plants and their associated biota.

The vast majority (c. 94 %) of plant species are mycorrhizal (Brundrett 2009), and most plants depend on mycorrhizal fungal symbionts for their nutrient uptake (Lambers et al. 2008). The mycorrhizal symbiosis includes different types of mycorrhizas, the most common of which are arbuscular mycorrhizas associated with c. 74 % of plant species (Brundrett 2009). Ectomycorrhizas and ericoid mycorrhizas are associated with relatively few plant species (2 % and 1 %, respectively), but include widespread and dominant plant species of temperate, boreal and some tropical forests. Orchids form their own type of mycorrhiza and include another 9 % of plant species (Brundrett 2009). Little is known of how orchid mycorrhizas respond to ecosystem development (but see Diez 2007; McCormick et al. 2012) and they never dominate ecosystems, so we do not include them in detail here. Three other types, ectendo-, arbutoid and monotropoid mycorrhizas significantly overlap with ectomycorrhizas in fungal identity and we consider them in that section (Brundrett 2004).

Plants allocate substantial, but variable, quantities of carbon to mycorrhizal fungi, averaging c. 10–20 % of net primary productivity (Leake et al. 2004; Hobbie 2006; Hogberg and Read 2006). In turn, the vast majority of plant uptake of several major nutrients is obtained via mycorrhizal fungi (Leake et al. 2004). As a consequence, mycorrhizal fungi can mediate ecosystem processes at the root–soil interface, including nutrient acquisition and uptake. Despite the importance of mycorrhizal fungi to plant nutrient uptake, few studies have directly studied mycorrhizal interactions along the major, well-characterised retrogressive chronosequences (Peltzer et al. 2010; Turner et al. 2012a, b), and those few have included only limited measurement of mycorrhizal fungal communities (e.g., Jehne and Thompson 1981; Holdaway et al. 2011).

For this review, we focus on understanding the links between fungal communities and long-term soil development throughout ecosystem development spanning primary succession through to retrogression. There have been a large number of studies of secondary successions of mycorrhizal fungal communities on planted trees (e.g., Fleming et al. 1984, 1986) or following disturbance (e.g., Gassibe et al. 2011; Kipfer et al. 2011). From these studies, broad concepts of ‘early-stage’ and ‘late-stage’ fungi have developed. These ‘stages’ in their original meaning refer to shifts in fungal communities over decadal time frames (typically over the lifespan of individual trees) and may have limited relevance to the much longer timescales involved with pedogenesis and ecosystem development. In secondary successions, previous vegetation has strong ecological legacy effects via soil properties and persistent fungal spores. We have therefore generally excluded studies of secondary-succession because of their limited relevance to ecosystem development and pedogenesis, and to avoid confusion over what comprises ‘early-stage’ versus ‘late-stage’. Further, for clarity, we use the word ‘colonisation’ to refer to the establishment of plants or fungi within a new habitat or landscape, and ‘infection’ to refer to fungal growth within or on plant roots.

Shifts among types of mycorrhizas

We begin by considering shifts among types of mycorrhizas (ectomycorrhizas, arbuscular mycorrhizas, ericoid mycorrhizas), and consider shifts in fungal community composition within these types in subsequent sections. Each of these types contains considerable variation in both plant and fungal physiology. Nonetheless, each type is also characterised by a particular phylogenetic grouping of fungi or plant components (e.g., fungi in the Glomeromycota form arbuscular mycorrhizas, most plants in the Ericaceae form ericoid mycorrhizas) and have some ecological traits in common (Read 1993; Read and Perez-Moreno 2003; but see Koele et al. 2012). Further, plants within a mycorrhizal type can frequently share mycorrhizal fungal symbionts, potentially influencing plant interactions and succession (e.g., Bever et al. 1997, 2010; Dickie et al. 2004; Nara 2006b; Mangan et al. 2010; Spence et al. 2011).

Read (1993) proposed a predictable sequence of mycorrhizal types during primary succession and subsequent ecosystem development in temperate and boreal forests. He suggested that non-mycorrhizal plant species colonise bare, high-P substrates in early succession, that these are replaced by arbuscular mycorrhizal plant species, followed by ectomycorrhizal trees with an arbuscular mycorrhizal understory, followed by ectomycorrhizal trees with an ericoid mycorrhizal understory, and finally to dominance by ericoid mycorrhizal vegetation. Lambers et al. (2008) augmented this model, noting that non-mycorrhizal cluster-rooted species (predominantly in the Proteaceae plant family) dominate on the oldest, most highly P-limited ecosystems. Despite these predicted trends in the mycorrhizal status of dominant vegetation, it has been recognised that all mycorrhizal types can occur across stages of ecosystem development and retrogression (Read 1993; Cázares et al. 2005; Lambers et al. 2008). For example, ectomycorrhizal and ericoid mycorrhizal vegetation may achieve peak dominance in mature and retrogressive ecosystems, but both types can also be present very early in primary succession (Cázares et al. 2005; Hobbie et al. 2005). Equally, arbuscular mycorrhizal vegetation is not eliminated during ecosystem development, and can either persist or even dominate on very old soils (Lambers et al. 2008).

Peltzer et al. (2010) identified nine well-studied long-term soil chronosequences that capture all stages of ecosystem development from primary succession through to retrogression. Using vegetation descriptions along these nine chronosequences, and overlaying data on plant mycorrhizal status predicted from plant phylogeny (Brundrett 2009; Koele et al. 2012), we find that few sequences follow a clear progression or shift in the dominance of mycorrhizal types (Table 1). For example, ectomycorrhizal plants dominate throughout the Glacier Bay sequence (Alaska, USA) including during the earliest stages of succession (Helm et al. 1999). Conversely, the Hawaiian (USA) and Franz Josef (NZ) chronosequences lack ectomycorrhizal plant species and are dominated by arbuscular mycorrhizal plants across most stages of ecosystem development. The absence of ectomycorrhizal plants in Hawaii may reflect biogeographic dispersal barriers, but Leptospermum in the oldest sites at Franz Josef is consistently arbuscular mycorrhizal (Holdaway et al. 2011) despite being ectomycorrhizal elsewhere in New Zealand (Moyersoen and Fitter 1999). Given the known ability of fungi to disperse both from Australia (Moyersoen et al. 2003) and elsewhere to New Zealand it seems unlikely this reflects biogeographic barriers to fungal dispersal. Instead, it has been suggested that climate, soils, or other factors may determine mycorrhizal status (Weijtmans et al. 2007). Ectomycorrhizal trees are also present in the San Joaquin Valley chronosequence, but never dominate (Brenner et al. 2001). Plant species that occur on retrogressive stages can be associated with arbuscular mycorrhizas (e.g., Metrosideros in Hawaii (USA), Lepidothamnus and Quintinia in Franz Josef (NZ), Juniperus in Northern Arizona (USA)), ectomycorrhizas (e.g., Picea in Arjeplog (Sweden) and Nothofagus in Waitutu (NZ)) or ericoid mycorrhizas (e.g., Vaccinium and Empetrum in both Glacier Bay (USA) and Arjeplog (Sweden); Table 1). Even where shifts in mycorrhizal type occur, a single type may dominate for much of the sequence despite strong turnover in plant community composition (i.e., host plant species) and major shifts in soil properties including nutrient status and physical characteristics.

Table 1 Summary of mycorrhizal types along nine long-term chronosequences that include distinct retrogressive stages and for which detailed vegetation data are available based on predicted mycorrhizal type by plant species. Mycorrhizal plant groups in bold are significant in that they account for the majority of plant biomass; those in plain text are present but minor
Full size table

The failure of most well-studied retrogressive chronosequences to follow the hypothesised trajectories of Read (1993) and others raises the question of whether the hypothesis should be rejected. Most literature on mycorrhizal fungi concerns relatively young soils and short-term chronosequences in the Northern Hemisphere, which may provide a biased view of mycorrhizal ecology (Dickie and Moyersoen 2008). In addition, the most commonly-studied secondary successions (i.e., abandoned agricultural systems dominated by arbuscular mycorrhizal species succeeding to forests dominated by ectomycorrhizas) conflate agricultural abandonment and old-field succession with ecosystem development. On the other hand, long-term soil chronosequences are not a representative sample of landscapes (Porder et al. 2007), but are space-for-time model systems for understanding long-term processes such as pedogenesis whilst holding other state factors (sensu Jenny 1980) steady. In some cases, a mycorrhizal type may not be available in the regional species pool (e.g., the absence of ectomycorrhizal plants in Hawaii) or other ecological factors may overwhelm pedogenesis as a determinant of mycorrhizal dominance. For example, the cold climate of Arjeplog (Sweden) and Glacier Bay (Alaska) may explain the absence of a stage dominated by arbuscular mycorrhizal plants. Despite these biases, two points are clear. First, a predictable shift from arbuscular to ectomycorrhizal dominance is not common across long-term chronosequences (Table 1). Second, a single mycorrhizal type can dominate ecosystems on geological timescales despite major shifts in plant species composition, changes in soil physical properties such as pH and organic matter content, and strong shifts in soil nutrient status.

An additional complication is that some plants form more than one type of mycorrhiza. For example, dual arbuscular mycorrhizal and ectomycorrhizal infection is reported in many tree genera including Eucalyptus (Bellei et al. 1992), Salix (Dhillion 1994), Uapaca and Leptospermum (Moyersoen and Fitter 1999) and more limited arbuscular mycorrhizal infection in predominately ectomycorrhizal genera such as Quercus (Dickie et al. 2001) and Pinus (Horton et al. 1998). Because some dual-mycorrhizal-status plants appear common in early succession, Read (1993) suggested that dual-mycorrhizal-status plants might establish early in succession with arbuscular mycorrhizas and then facilitate a transition to ectomycorrhizal dominance. This mechanism can be important in secondary succession, where arbuscular mycorrhizal inoculum is likely to be present (e.g., Arveby and Granhall 1998). However, evidence that dual-mycorrhizal-status plants switch from being arbuscular mycorrhizal to ectomycorrhizal in primary succession is weak. Rather, it appears that arbuscular mycorrhizal fungi are often slower to arrive than ectomycorrhizas (e.g., Cázares et al. 2005), and a number of studies now suggest that Salix and other dual-mycorrhizal plants establish relying primarily on ectomycorrhizal symbionts and potentially dark-septate endophytes (discussed below), with very limited arbuscular mycorrhizal infection (Trowbridge and Jumpponen 2004; Hobbie et al. 2005; Nara 2006a; Obase et al. 2007). As an alternative explanation for the prevalence of dual-mycorrhizal plants in early succession, van der Heijden and Vosatka (1999) suggested that relatively low levels of arbuscular mycorrhizal infection (<10 % of root length) can increase uptake of soil mineral P by plants in early succession. Thus dual-mycorrhizal plants may establish as ectomycorrhizal in early succession and subsequently acquire arbuscular-mycorrhizas to maximise mineral P uptake in young soils. A third, myco-centric explanation for dual infection in early succession is also possible. Ectomycorrhizal development can supress arbuscular mycorrhizas in individual root segments. If ectomycorrhizal inoculum is limiting, or if soil nutrients are sufficient to suppress ectomycorrhizal development, this may reduce competitive exclusion of arbuscular mycorrhizal infection by ectomycorrhizal fungi.

Shifts in composition within mycorrhizal types

Given that a single mycorrhizal type can dominate over long periods of ecosystem development, it becomes important to consider shifts in fungal community composition within mycorrhizal type. Each mycorrhizal type contains considerable variation in function among fungal species (Table 2). Arbuscular mycorrhizas evolved once in fungi, but are an ancient lineage with substantial diversification in species and functional traits. Further, a separate group, the Mucoromycotina, can form arbuscular-mycorrhiza-like endomycorrhizal associations with some basal plant clades (Bidartondo et al. 2011). Ectomycorrhizas and ericoid mycorrhizas evolved multiple times in fungi and also display high functional diversity among species within each type (Agerer 2001). Differences among fungal species in their dispersal traits, degree of host specificity and nutrient requirements will strongly determine the order in which species arrive during primary succession and the duration of species through ecosystem development into retrogression. In the following sections we review published studies of fungal community responses to ecosystem development and, where possible, identify how fungal traits underpin those responses.

Table 2 Examples of variability within the three main mycorrhizal types in terms of taxonomic origin, dispersal and soil interactions
Full size table

Ectomycorrhizal fungi along chronosequences

Ectomycorrhizal plants dominate many boreal and temperate forest ecosystems, some tropical forests, and a range of other ecosystems including some alpine and boreal grasslands (Gao and Yang 2010). Although many ectomycorrhizal plants are woody, herbaceous ectomycorrhizal plants are commonly found in early succession (e.g., Kobresia mysuroides, Polygonum viviparum, Salix herbacea). On the fungal side, there is considerable variability in enzymatic capability for decomposition of organic matter (Bruns 1995; Courty et al. 2010), hyphal foraging strategies (Agerer 2001), lifespan and dispersal mechanisms (Fiore–Donno and Martin 2001), all of which may be important for understanding the ectomycorrhizal fungal community during ecosystem development.

Ectomycorrhizal fungi in early succession

Mycorrhizas are horizontally transmitted symbioses: each generation of plants must obtain its symbionts independently of the parent plant. This can be an important limitation on plant establishment in early succession if fungal propagules are absent (Allen 1987; Allen et al. 1992). Despite this, ectomycorrhizal plants are early colonists of primary successional surfaces, and include woody shrubs and trees (e.g., Salix, Pinus, Alnus), and herbaceous plants (e.g., Kobresia and Polygonum viviparum; Muhlmann et al. 2008; Muhlmann and Peintner 2008). Relatively few fungal genera are reported as frequent early in succession (Fig. 1), notably Cenococcum, Cortinarius, Inocybe, Laccaria, Scleroderma, Sebacina, Thelephora and Tomentella, as well as fungi in the Pezizales. Some of these fungi are frequent across studies, but not necessarily abundant, either as sporocarps or belowground. Others produce abundant sporocarps with few ectomycorrhizal root tips or, conversely, are abundant as mycorrhizal root tips but produce few, inconspicuous, or no sporocarps (e.g., Tomentella, Cenococcum).

Fig. 1
figure 1

Frequency of observation of common ectomycorrhizal fungal genera across soil chronosequences and putative trait groupings. Filled circles indicate presence of the genus at that site, open circles indicate that the species was absent at that site. Dotted lines connect observations from the same study. Number of studies (n) for each ecosystem age given at bottom. Based on data from Helm et al. (1996); Jumpponen et al. (1999, 2002); Alfredsen and Hoiland (2001); Nara et al. (2003); Hobbie et al. (2005); Ashkannejhad and Horton (2006); Obase et al. (2007); Trocha et al. (2007); Muhlmann and Peintner (2008); Muhlmann et al. (2008); Reverchon et al. (2010); Fujiyoshi et al. (2011); Blaalid et al. (2012) and Peintner and Dammrich (2012)

Full size image

The ability to disperse may be a strong filter on the establishment of ectomycorrhizal fungal species early in primary succession. Many ectomycorrhizal fungi produce copious wind-dispersed spores. For example, Suillus bovinus may produce up to 1.2 × 109 spores per sporocarp (Dahlberg and Stenlid 1994). Although >95 % of spores fall within 0.5 m of sporocarps (Li 2005; Galante et al. 2011), the sheer number of spores produced appears sufficient for effective long-distance dispersal (Peay et al. 2012). For example, wind-dispersed spores arriving early in succession were observed at low density at Mt St Helens (Allen 1987), and Jumpponen (2003) detected DNA of many obligate, biotrophic fungal species at a glacial forefront in the absence of plant hosts, most likely representing aerially-deposited spores. Nonetheless, low ectomycorrhizal fungal inoculum potential can limit plant establishment during primary succession (Nara 2006b; Fujiyoshi et al. 2011), and many species detected in spore traps fail to infect seedlings (Peay et al. 2012). The paradox of producing copious wind-dispersed spores yet failing to establish remains largely unresolved (Peay et al. 2010a). In contrast, animal dispersal of fungal spores may be particularly important in primary succession, including dispersal by soil invertebrates (Lilleskov and Bruns 2005), small mammals (Terwilliger and Pastor 1999) and deer (Ashkannejhad and Horton 2006). These fungal species may also receive an initial boost via co-deposition of high-nutrient organic matter.

Dispersal ability is likely a major filter determining the composition of early-successional fungal communities. Dispersal traits of fungi (sporocarp height and palatability, spore volume and capacity to survive passage through animal guts) vary widely among species (Ashkannejhad and Horton 2006; Galante et al. 2011). Similarly, the ability to persist as dormant spores may be a strong filter on fungal communities, with some early-successional fungi such as Wilcoxina, Rhizopogon or Suillus having a spore longevity of at least several years and potentially much higher (Bruns et al. 2009). Equally important is an ability to germinate when a host root is available. Early-successional taxa (Laccaria, Inocybe) have much higher spore infectivity than late-successional taxa (Ishida et al. 2008). Finally, most fungal spores of ectomycorrhizal fungi are monokaryotic (having one haploid nucleus) and the fusion of two compatible monokaryotic mycelium – forming a dikaryon – is necessary for reproduction and the establishment of a population. Having the ability to form a functional mycorrhiza as a monokaryon (e.g., Laccaria; Kropp and Fortin 1988) or to form even a limited number of binucleate heterokaryotic spores may therefore be a significant advantage in primary succession (Horton 2006). Once established on a single root, an ectomycorrhizal fungus can infect other roots on either the same or other plant hosts through hyphal growth. Low root densities in early succession may thus favour fungal species having relatively long hyphae and rapid growth (Peay et al. 2011).

An additional important factor in the composition of early-successional fungal communities may be plant host identity, with a number of early-successional fungi showing high host specificity (Molina et al. 1992). For example, Rhizopogon and Suillus are common in early successions where Pinus is present (Peay et al. 2012), but otherwise are generally absent (but see Krpata et al. 2007). Similarly, Alnicola (=Naucoria) and Alpova both show relatively high specificity to Alnus (Molina 1981; Moreau et al. 2006) and many Leccinum have a high specificity to various early-successional trees, including Betula (den Bakker et al. 2004). Leccinum, Suillus and Rhizopogon are notable in producing abundant and large sporocarps, while most other early-successional ectomycorrhizal fungi have small or inconspicuous sporocarps (Muhlmann and Peintner 2008). Bruns et al. (2002) suggested that high host-specificity may be related to increased symbiotic benefits received by a fungus. Suillus, for example, forms only a minor component of below-ground fungal communities but produces abundant sporocarps, implying that it is more efficient at converting host carbon into reproductive output (Bruns et al. 2002) or may be accessing other carbon sources (Chapela et al. 2001). The occurrence of high host specificity in early-successional communities may therefore be a mechanism for increasing host C-allocation under relatively high soil nutrient conditions, where host benefit is limited. However, whether high carbon cost implies a less mutualistic interaction depends on nutrient benefit and the cost : benefit ratio (Hobbie et al. 2005), which is poorly understood.

Wilcoxina, which often occurs in early succession, can form ‘ectendomycorrhizas’, in which a Hartig net and thin fungal mantle are formed (as in ectomycorrhizas) but fungal hyphae also penetrate plant root cells, with the degree of root cell penetration related to soil nutrient status (Yu et al. 2001; Navarro-Ródenas et al. 2012). Somewhat similarly, arbutoid mycorrhizas are formed by the basal Ericaceae Arbutoideae (Arbutus, Arctostaphylos). Like ectendomycorrhizas, the fungi penetrate plant cells, but form a mantle and Hartig net. The fungi forming arbutoid mycorrhizas form ectomycorrhizas on other plant species (Horton et al. 1999). The Monotropideae (also basal Ericaceae) form monotropoid mycorrhizas, in which the plant parasitises ectomycorrhizal fungi for carbon. While distinctive in some physiological traits, the strong overlap of fungal species suggests these categories can generally be considered as specialised variants of the ectomycorrhizal association (Brundrett 2004; Finlay 2008).

Ectomycorrhizal fungi in mature ecosystems

As ecosystems develop, communities of ectomycorrhizal fungi generally become more diverse, with early-successional species largely retained and additional species colonising later stages of ecosystem development (Nara et al. 2003; Peay et al. 2011). Genera commonly reported in mature ecosystems, but less frequently earlier in succession, include Amanita, Boletus, and Russula (Fig. 1; also see Peay et al. 2012). Some of these genera are frequently associated with soil organic layers and may depend on soil development to become established. For example, Russula, which increases in frequency in mature ecosystems, has also been shown to be positively correlated with total soil C and depth of the organic layer in mature forest (Dickie et al. 2009). Nonetheless, some genera such as Cortinarius are commonly dominant in soil organic layers in mature forest, but also occur earlier in succession (Muhlmann and Peintner 2008).

Fungi found in mature ecosystems, but absent earlier in succession, include many genera with large sporocarps (e.g., Amanita, Boletus), which may have high carbon demands relative to earlier-successional species (Gibson and Deacon 1990). Many of these genera appear to depend on mature trees in order to successfully infect seedlings, as trenching around seedlings in mature forests reduces the abundance of late-successional species and can increase seedling growth (Fleming 1984; Simard et al. 1997). Unlike large-sporocarp genera found earlier in succession, many of those found later in mature ecosystems can simultaneously infect multiple host species (e.g., Dickie et al. 2004), and the ability to switch plant host may be important in the long-term persistence of fungal taxa in mature ecosystems.

Ectomycorrhizal fungi in retrogression

Because retrogression is a process, determining whether an ecosystem is in retrogression depends on having a comparable sequence of ecosystems on younger soils. Regrettably, few, if any, studies have examined ectomycorrhizal fungal communities along well-documented retrogressive sequences. As an imperfect interim solution, we can infer patterns from the lowland tropics, where much of the soil has characteristics similar to those of retrogressive ecosystems (i.e., strongly P-limited, with most nutrients in organic form). On one very-low-P sandy soil under Malaysian dipterocarp forest, Peay et al. (2010b) found Russula-dominated communities, but also Cortinarius, Tomentella, Amanita, Lactarius, and Gymnomyces. Similarly, in tropical South America, Smith et al. (2011) found Russula to dominate ectomycorrhizal fungal communities along with Clavulina, Tomentella, Tylopilus, Xerocomus, Cortinarius and other genera. The dominance of Russula in these low-P soils may imply a unique role in P uptake. Russula is notable in having few emanating hyphae but abundant surface cystidia with associated calcium-oxalate crystals, which may be related to mining P from soils (Massicotte et al. 2005; Avis 2012). The occurrence of the early-successional genera Tomentella and Cortinarius on old, P-depauperate tropical soils in both studies suggests that at least some fungal genera can be present across most, or all, of pedogenic development and retrogression. Conversely, both Cenococcum and Laccaria are absent from both studies and most tropical ectomycorrhizal fungal communities, despite being common in both the Holarctic and Austral regions (Tedersoo et al. 2010). The absence of these two genera, which are common in early primary succession, could indicate a retrogression-driven loss of species. Further studies of ectomycorrhizal fungal communities in primary succession in the tropics are needed to resolve whether the absence of Cenococcum and Laccaria is driven by pedogenesis as opposed to biogeography.

Ectomycorrhizal summary

The overarching pattern for ectomycorrhizal fungal communities during progressive and retrogressive phases of ecosystem development is an accumulation of diversity, with few host-generalist genera lost. Host-specialist genera, in contrast, appear to be more transient due to their dependence on the presence of compatible hosts. The persistence of early-successional fungal genera through the mature stages of ecosystem development is intriguing, and contrasts markedly with plant communities (e.g., Walker and del Moral 2003; Wardle et al. 2008). Some early-successional taxa may persist in the mineral soil layers later in ecosystem development, and perhaps considering these species to be ‘mineral soil’-adapted is more useful than considering them to be ‘early successional’. Three other factors may also contribute to the persistence of ectomycorrhizal fungal genera from early succession. First, fungi can occur as small genets, which may permit species to utilise small, localised disturbances and persist despite overall ecosystem development. Laccaria, for example, persists in mature ecosystems but appears to be associated with localised disturbance (Dickie et al. 2009), and occurs as very small, short-lived genets (Gherbi et al. 1999). Alternatively, fungi can be extremely long lived and show strong priority effects, such that early-establishing species can persist for extended periods (Kennedy and Bruns 2005; Kennedy et al. 2009). It is possible that a single fungal genet might persist for thousands of years, even if conditions no longer favour its establishment. Finally, there is some evidence that peripheral roots (those far from the host plant stem) have a distinct fungal community of early-successional species (Last et al. 1984; Dickie and Reich 2005). It is possible that resource partitioning by host roots may allow low-carbon-demanding species to persist on peripheral roots of trees within contiguous forest.

Arbuscular mycorrhizal fungi along chronosequences

Arbuscular mycorrhizal fungi enhance plant-uptake of P and, to a lesser extent, N, but have a limited capacity to use insoluble soil nutrients relative to ectomycorrhizal fungi (George et al. 1995). Communities of arbuscular mycorrhizal fungi respond strongly to soil variables at local and regional scales, including soil pH and texture (Lekberg et al. 2007, 2011), and we thus anticipate there will be clear shifts in fungal community composition during ecosystem development.

Most studies of arbuscular mycorrhizal fungi during ecosystem development focus on abundance, particularly the extent of root infection, with few studies characterising community composition. Regrettably, we found no studies of arbuscular mycorrhizal fungal communities during retrogression. Our review is thus confined to compositional shifts during the early and mature phases of ecosystem development, but these cover a diverse range of situations including sand dunes (Jehne and Thompson 1981; Sikes et al. 2012), volcanic systems (Oba et al. 2004), and glacial moraines (Oehl et al. 2011a).

Early taxonomic efforts to characterise arbuscular mycorrhizal fungal communities were based largely on spore morphology, with many species grouped under the genus Glomus. Recent advances in molecular methods have resulted in new classifications of orders, families and genera, and an increase in the number of species from c. 170 to 220. Two recent classifications based on genetic characters have been proposed (Oehl et al. 2011b; Kruger et al. 2012), which presents a hurdle for reviewing studies that have used either the traditional taxonomy, one of the two more recent taxonomies, or various combinations thereof. Here, we have updated data from prior studies following the taxonomy of Oehl et al. (2011b) on the basis of its strength in using both genotypic and phenotypic characteristics.

Despite correcting taxonomy, we found no evidence of consistent shifts in arbuscular mycorrhizal fungal community composition during the early and mature phases of ecosystem development (Fig. 2; Jehne and Thompson 1981; Johnson et al. 1991; Oba et al. 2004; Pezzani et al. 2006; Wu et al. 2007; Oehl et al. 2011a; Sikes et al. 2012). Rather, fungal genera are present across a wide range of ecosystems from early to mature stages. For example, Acaulospora is frequent across studies in early succession, yet remains frequent as ecosystems mature (Fig. 2). There are also no consistent shifts in the diversity of arbuscular mycorrhizal fungi during ecosystem development, with diversity increasing (e.g., Pezzani et al. 2006), decreasing (Sikes et al. 2012; Zangaro et al. 2012), or remaining constant (Johnson et al. 1991). Increasing diversity was associated with the addition of new species and the persistence of pioneer species into the mature phase (Wu et al. 2007), similar to the pattern observed for ectomycorrhizal fungal communities.

Fig. 2
figure 2

Shifts in arbuscular mycorrhizal fungi across soil chronosequences from the earliest phase through to the mature phase (peak biomass). Filled circles indicate presence of the genus at that site, open circles indicate that the species was absent at that site. Dotted lines connect observations from the same study. Number of studies (n) for each ecosystem age given at bottom. Note that retrogressive phases remain entirely in the ‘black-box’ due to a lack of data, and are hence not shown. Based on data from Jehne and Thompson (1981); Oba et al. (2004); Pezzani et al. (2006); Wu et al. (2007); Oehl et al. (2011a); Sikes et al. (2012) and mature stage data from Johnson et al. (1991)

Full size image

Several factors may contribute to a lack of consistent compositional change in arbuscular mycorrhizal fungal communities during ecosystem development. First, communities of arbuscular mycorrhizal fungi may be more strongly structured by competition for plant roots than competition for soil resources (Maherali and Klironomos 2012), with host specificity playing a dominant role (Öpik et al. 2006) relative to soil development. Second, responses of arbuscular mycorrhizal fungi to fine-scale environmental variability and seasonal changes may overwhelm site-scale responses to ecosystem development, as has been observed in plant communities in early succession (del Moral et al. 2010). Third, the large spore size and short dispersal distances of arbuscular mycorrhizal fungi may result in fungal community composition being dominated by random dispersal from proximal landscapes, rather than globally dispersed propagules being filtered by environmental selection. There are also possible methodological explanations. For example, most studies have been based on fungal spores, frequently measured at a single time point, and may not accurately reflect which species infect plant roots (Oba et al. 2004; Sanders 2004; Wu et al. 2007; Martínez-García et al. 2011). It is also possible that the phylogenetic concept of a genus in arbuscular mycorrhiza is sufficiently broad as to obscure ecological patterns.

Ericoid mycorrhizal fungi along chronosequences

Ericoid mycorrhizal fungi are characterised by their ability to form coils in the epidermal cells of the finest roots (‘hair roots’ as distinct from ‘root hairs’ of other plants) of most plants in the Ericaceae (excluding the basal clades, Arbutoideae, Monotropoideae and Enkianthoideae which form arbutoid, monotropoid and arbuscular mycorrhizas, respectively). Ericoid mycorrhizal fungi include taxa in the Ascomycota such as Rhizoscyphus, Meliniomyces, Oidiodendron, Capronia and Cryptosporiopsis, and in the Basidiomycota such as Sebacinales group B (Weiß et al. 2011). The Ericaceae and ericoid mycorrhizas occur worldwide but are absent from large parts of Africa, much of the lowland neotropics, and northern and central Australia. They are characteristic of low soil fertility and high soil acidity sites (Cairney and Meharg 2003). The few species of ericoid mycorrhizal fungi that have been studied in vitro show enhanced capability to degrade enzymes compared with all other mycorrhizal fungi (Smith and Read 2008). The ecological context of most laboratory studies has been that of mature and retrogressive phases of ecosystem development, focusing on how ericoid mycorrhizal fungi mobilise nutrients locked up in complex organic matter. Much less is known of the role of ericoid mycorrhizal fungi in nutrient uptake in high-P early-successional sites, despite the frequent occurrence of Ericaceae early in succession (Table 1).

We did not find any studies describing the community response of ericoid mycorrhizal fungi to long-term ecosystem development, perhaps reflecting difficulties identifying fungi having few distinctive morphological or spore types. Instead, the focus to date has been on the presence of the ericoid mycorrhizal type relative to other mycorrhizal types. Cázares et al. (2005) reported that ericoid mycorrhizal plants were absent from the earliest successional sites of the Lyman glacial chronosequence (USA) but present on moraines that were 35–60 year old, while the intensity of root infection was unresponsive to ecosystem age. Typical coils of ericoid mycorrhizas have been observed at the youngest sites along the Franz Josef chronosequence (Table 1) in roots of Gaultheria macrostigma and Dracophyllum longifolium (G.-A. Grelet, unpublished data). The response by ericoid mycorrhizas to ecosystem development may be driven by the presence and depth of the organic horizon as there is evidence for vertical niche partitioning among soil horizons. For example, Wurzburger et al. (2012) found that communities of ericoid mycorrhizal fungi hosted in the roots of Rhododendron maximum varied strongly among soil horizons, with Basidiomycota taxa occurring more frequently in the organic horizon and Leotiomycetes more frequently in the mineral horizon. Taxa closely related to the archetypal ericoid mycorrhizal Rhizoscyphus ericae and Oidiodendron maius were found at similar frequencies in both organic and mineral horizons. Their results suggest that ericoid mycorrhizal fungal communities might be more diverse later in ecosystem development once a mature organic horizon has developed.

An important factor in early-successional communities is dispersal, and observational evidence suggests that ericoid mycorrhizal fungi can be dispersal limited. For example, Hutton et al. (1997) observed that ericoid mycorrhizal infection in disturbed sites (in which top soil had been removed and subsequently replaced) took about 12 year to return to levels comparable to undisturbed sites. Furthermore these authors showed that infection levels dropped to virtually zero when bait plants were planted in soils > 40 cm away from established ericoid mycorrhizal plants. Dispersal strategies in ericoid mycorrhizal fungi remain poorly understood, hence the mechanisms by which they may colonise soil and infect host plants at the earliest stage of succession are largely unknown. In part, this is confounded by the difficulty of observing and characterising the cryptic sexual structures formed by microfungi, such as most of the ascomyceteous taxa forming ericoid mycorrhiza. We know that some ericoid mycorrhizal fungi produce microscopic ascocarps (e.g., Rhizoscyphus ericae produces apothecia up to 1 mm diameter; Read 1974) or asexual propagules (e.g., arthrospores in Oidiodendron species), but conditions under which those are produced, and how far they disperse, are unknown. Many ericoid mycorrhizal fungi have never been observed to sporulate, either sexually or asexually, and are still considered to be sterile (Hambleton and Sigler 2005). Ashford et al. (1996) documented the formation of thick-walled heavily colonised epidermial cells in hair roots of Western Australian Ericaceae. Based on the fact that these cells are often sloughed off the root, and that the fungus remains alive in sloughed cells, the authors suggested that these specialised cells could act as fungal propagules for year-to-year persistence. This points towards a dissemination strategy dominated by asexual propagules (hyphae from neighbouring plant or thick-walled detached epidermial cells), effective only over very short distances. On the other hand, Bergero et al. (2000) showed that viable ascomyceteous ericoid mycorrhizal fungal structures persisted in woodlands, for possibly 10–20 year, in the absence of suitable ericoid mycorrhizal host. This may indicate the presence of viable persistence structures such as spores, but it may also reflect (1) the ability of some ericoid mycorrhizal fungi to colonise ectomycorrhizal roots, which then act as fungal refugia until the appearance of Ericaceae host plants (see below for further discussion on this point) or (2) the ability of some ericoid mycorrhizal taxa to also function as saprotrophs (Piercey et al. 2002). Hence for many ascomyceteous taxa, the available evidence suggests dispersal is local, short-scale and dominated by asexual propagules, but the physical constraints on propagule production and dissemination are largely unknown. Dispersal strategies of ericoid-forming Sebacinales group B are unknown due to a lack of morphological data (Weiß et al. 2011), no fruiting body descriptions, and an inability to culture these fungi. However, their dispersal strategies are expected to differ markedly from those of other ericoid mycorrhizal fungi given the contrasting fruiting body and morphological characteristics of members of Sebacinales group A. As with arbuscular mycorrhiza, further studies are needed of ericoid mycorrhizal fungal communities across chronosequences, particularly where ericoid mycorrhizal plants occur over long-term ecosystem development (e.g., Cooloola, Australia; Hawaii, USA; Franz Josef, NZ; Arjeplog, Sweden; Table 1).

There are several lines of evidence that fungal taxa traditionally regarded as ectomycorrhizal occur in field-grown ericoid roots and vice versa (e.g., Bergero et al. 2000; Vralstad et al. 2000; Bougoure et al. 2007; Tedersoo et al. 2007) with function demonstrated by structural data (Villarreal-Ruiz et al. 2004, 2012) and reciprocal transfer of resources between fungus and plant (Grelet et al. 2009). Genetic studies also demonstrate that neighbouring ericoid mycorrhizal and ectomycorrhizal roots share the same fungal genet (Grelet et al. 2010). However, the impact of infection by these potentially dual-mycorrhizal taxa on plant host is unclear. In a pot experiment, Kohout et al. (2011) found that the presence of ericoid-mycorrhizal plants suppressed the formation of ectomycorrhizas between neighbouring Pinus spp. and dual-mycorrhizal Melinomyces bicolor. On the other hand, along a primary successional gradient on the dune shores of Bothnian Bay, Finland, Grau et al. (2010) reported a higher proportion of Pinus sylvestris seedlings established in early to mid-successional patches already colonised by the ericoid mycorrhizal shrub Empetrum nigrum. The apparent facilitating role of Empetrum was also noticeable in its impact on the ectomycorrhizal fungal community, as it increased both morphotype diversity and the proportion of ectomycorrhizal root tips with abundant external mycelia. Nonetheless, in an older successional stage, Empetrum had a detrimental effect on both seedling recruitment and ectomycorrhizal symbiosis, possibly due to allelopathic or competitive interactions. Hence whether and how these potentially dual-mycorrhizal taxa play a significant role in shaping plant communities is unknown and requires further investigation.

Dark septate endophytes along chronosequences

In addition to mycorrhizal fungi, there is another common group of root-associated fungi known as dark septate endophytes (DSE). The nature of the interaction between DSE and their host plant is still a matter of debate (Grünig et al. 2008; Newsham 2011). However, several papers have considered their role in early succession (e.g., Jumpponen and Trappe 1998; Cázares et al. 2005; Tejesvi et al. 2010; Day and Currah 2011). Additionally, there have been several reports of negative correlations between mycorrhizal and DSE infection along defined environmental gradients (e.g., Vohník and Albrechtová 2011) and in both polar regions, DSE infection is more prominent than mycorrhizal infection (Newsham et al. 2009). Therefore we have included here a discussion of DSE in ecosystem development.

Dark septate endophytes are fungi with melanised septate hyphae which infect plant roots by growing within and between cells, and along the root surface, but which do not form intracellular coils. They have been found in the roots of over 600 plant species spanning more than 100 plant families (Jumpponen and Trappe 1998) including the Ericaceae (although not restricted to hair roots in the Ericaceae unlike ericoid mycorrhizal fungi) and plants that are typically ectomycorrhizal or arbuscular mycorrhizal. Dark septate endophytes include diverse unrelated genera of Ascomycete fungi (e.g., Cadophora, Microdochium, Trichocladium, Phialophora, Leptodontidium and Phialocephala). Functional differences aside; the structural and genetic distinction between DSE and mycorrhizal fungi is sometimes obscure. For example, some genera include both DSE and ectomycorrhizal taxa (e.g., Cadophora, Acephala), and taxa initially considered as DSE have later been shown to form ericoid mycorrhizal associations in vitro (Usuki and Narisawa 2005; Grelet et al. 2009) or in vivo (Vohník and Albrechtová 2011).

Unlike many other root-associated fungi, dark septate endophytes can precede plants in early succession because they can persist and produce propagules in the absence of a host plant (Day and Currah 2011). Cázares et al. (2005) reported that DSE were present from the very earliest stages of succession of the Lyman glacial chronosequence (USA), albeit at low intensity (< 25 % of root length infected). Tejesvi et al. (2010) investigated the community composition of DSE along an early-successional vegetation gradient (~50–300 years) created by land uplift, using culture-based methods that favoured the growth of DSE taxa such as Phialocephala fortinii and precluded the growth of slow-growing ericoid mycorrhizal taxa (e.g., Rhizoscyphus ericae, Capronia spp.). Their results suggest that DSE fungal species richness increases through ecosystem development, at least during early succession through to peak biomass. The authors reported that taxa related to the Phialocephala fortinii – Acephala applanata species complex (PAC) were ubiquitous along the gradient, but Phialophora spp. were apparently restricted to the earliest successional stage.

DNA of dark septate endophytes has been sampled from the forefront of the Lyman Glacier chronosequence (Jumpponen 2003) despite the absence of suitable host plants. Jumpponen (2003) suggested that this DNA was sourced from dormant spores, highlighting the importance of airborne spore deposition early in succession for fungal communities. Dispersal limitation in DSE communities is poorly understood. DSE taxa typically produce asexual survival propagules (but see Zaffarano et al. 2011) such as microsclerotia and conidia, and some have been shown to do so even in the absence of host roots (Day and Currah 2011), which could also contribute to presence on glacial forefronts. As with ericoid and arbuscular mycorrhizas, further studies across chronosequences, particularly at later stages of ecosystem development, are needed.

Plants, fungi, and pedogenesis

The interaction of plant communities with pedogenesis has long been recognised (Jenny 1958), with plant communities both responding to and influencing soil properties (Northup et al. 1998; Reich et al. 2005). To the extent that mycorrhizal fungi have been considered, it has largely been in the context of their role in modulating the plant–soil interaction, particularly expanding the realised niche of plants (Bever et al. 2010; Rodriguez-Cabal et al. 2012). At the most extreme, mycorrhizal fungi have been considered as part of the plant’s ‘second genome’; effectively treating the fungal community as an extension of the plant’s enzymatic and physiological capabilities (Berendsen et al. 2012). This approach has some merit, but fails to recognise that mycorrhizal fungi are, themselves, free-living macroorganisms, both responding to and driving pedogenesis and plant community composition. In this section we focus on the fungal parts of the interaction (see Fig. 3) considering interactions between mycorrhizal fungal communities and pedogenesis and interactions between mycorrhizal fungal communities and plant communities. In focusing on these two components, it is important to recognise that these interactions occur in a broader context of rhizosphere and soil interactions. In particular, mycorrhiza helper bacteria (bacteria found in the mycorrhizosphere of arbuscular and ectomycorrhizal fungi; Frey-Klett et al. 2007) play important roles in the establishment and infection of the mycorrhizal fungi on the plant root (Garbaye 1994), the acquisition of nutrients from mineral or organic substrates for the plant–fungus symbiosis (Calvaruso et al. 2007; Koele et al. 2009; Uroz et al. 2009) and protection against pathogens (Frey-Klett et al. 2005). Mycorrhizal fungi also play a major role in soil trophic webs, and may be key components of the increasingly fungal dominance of soil food web dynamics as ecosystems age (Williamson et al. 2005).

Fig. 3
figure 3

Ecosystem development and retrogression includes the interaction between not just plant communities and pedogenesis but also the community of mycorrhizal fungi that form a major interface between plants and soils, and that can modulate plant–soil interactions (thick arrows). These latter interactions form the scope of this review. Some of the major theories and concepts for components of the interaction are shown in phrases in grey text. The position of each phrase along arrows reflects the degree to which it focuses on one end or the other of the interaction. For example, in plant–fungal interactions, plant–soil feedbacks have focused primarily on plant community outcomes, whereas host-specificity has tended to focus on fungal community effects, with network theory being more balanced between the two

Full size image

Pedogenesis and mycorrhizal fungi

Soil chemical and physical properties strongly influence community composition of mycorrhizal fungi (Dickie et al. 2009), with fungal species responding to soil variables such as pH, nitrate, base cation status, and P (Fitzsimons et al. 2008). Soil pH emerges from multiple studies as a relatively strong factor in fungal community composition compared to more apparent drivers (Fitzsimons et al. 2008). For example, in a study of geothermal grassland soils in Yellowstone National Park, USA, Lekberg et al. (2011) found that pH was a stronger determinant of arbuscular mycorrhizal fungal communities than either temperature or plant communities. Nitrogen also has strong effects on fungal communities, with numerous studies of atmospheric N deposition showing a shift in composition, change in function, and loss of diversity following N deposition (Lilleskov et al. 2002; Wright et al. 2009; Cox et al. 2010). The influence of P on fungal communities has received less attention, yet it may have stronger and more consistent effects than N on the degree of mycorrhizal infection of plants (Treseder 2004) and fungal community composition (Twieg et al. 2007; Dickie et al. 2009).

Young soils are characteristically highly inorganic, dominated by bare rock or mineral surfaces with little or no vegetation and limited soil microbial activity. As ecosystems develop, complex soils become dominated by organic materials, with nutrients increasingly sequestered within biomass or in recalcitrant soil-organic-matter complexes (Richardson et al. 2004; Wardle et al. 2004; Peltzer et al. 2010; Turner et al. 2012a, b). Through time, soils become vertically stratified with the development of an organic horizon and acidification of upper mineral horizons. This soil horizon development is an important axis of fungal niche differentiation (Taylor and Bruns 1999; Dickie et al. 2002b; Rosling et al. 2003; Genney et al. 2006; Lindahl et al. 2007; Wurzburger et al. 2012). The accumulation of organic materials creates a novel environment from the largely inorganic mineral soil, and mycorrhizal fungi can dominate the functioning of organic nutrient cycles. As discussed above, changes in mycorrhizal community structure with ecosystem development may be more closely related to the development of different organic and mineral soil horizons than with ecosystem stage. In support of this view, Dickie et al. (2002b) and Lindahl et al. (2007) have shown niche differentiation in different organic horizons and the top mineral soil, and Rosling et al. (2003) and Scattolin et al. (2008) found differences in ectomycorrhizal community structures between organic and mineral soil horizons. Arbuscular mycorrhizal fungal and ericoid mycorrhizal fungal communities are both also responsive to soil depth (Oehl et al. 2005; Wurzburger et al. 2012). The degree to which vertical niche differentiation is driven by nutrient resources versus microsite conditions (e.g., absolute levels of, or variability in, pH, soil moisture, temperature, CO2, and other factors) or other microsite differences remains largely unexplored.

Mycorrhizal fungal communities are not simply passengers in the process of pedogenesis, but can also be important drivers. Ectomycorrhizal and ericoid mycorrhizal fungi are commonly thought to decompose organic matter in soil (Read and Perez-Moreno 2003), including some recalcitrant forms. More recently it has been suggested that arbuscular mycorrhizal fungi can also enhance the uptake of organic N and P (Tarafdar and Marschner 1994; Cheng et al. 2012; Veresoglou et al. 2012) although Hodge and Fitter (2010) suggest that arbuscular mycorrhizal fungi use organic N predominantly for their own nutrition and do not transport it to the host plant, unlike ectomycorrhizal and ericoid mycorrhizal fungi. Because most mycorrhizal fungi obtain carbon from hosts, nutrient uptake by mycorrhizas from organic matter can influence soil nutrient stoichiometry by removing N, P or other nutrients decoupled from C. Using a modelling approach, Orwin et al. (2011) suggested that mycorrhizal fungal uptake of nutrients depletes the residual organic matter and generates a poorer-quality substrate for decomposition by saprotrophs and other free-living microorganisms. This is in line with observations of decreased litter decomposition when ectomycorrhizal roots are present (Gadgil and Gadgil 1975), although Koide and Wu (2003) note that these field observations may be driven by plant water uptake as well. Furthermore, the accumulation of mycorrhizal exudates, biomass, and necromass over time can form a major part of the organic carbon and nutrient pools in retrogressive ecosystems (Rillig et al. 2001; Wallander et al. 2009; Turner et al. 2012a, b). This accumulation is potentially augmented by the relative recalcitrance of fungal tissue to decomposition (Wallander et al. 2009). Better incorporation of mycorrhizal stoichiometric relationships, particularly the decoupling of C and soil nutrients, may be critical in understanding the role of mycorrhizas in soil organic matter development (Orwin et al. 2011).

In mineral soil, fungal hyphae and exudates bind soil particles and play a key role in soil aggregation (Jehne and Thompson 1981; Rillig et al. 2003; Rillig and Mummey 2006). Mycorrhizal fungi, together with plant roots, soil microorganisms and decaying organic matter, release organic acids, acidifying the soil and increasing mineral weathering (Bormann et al. 1998; Chadwick and Chorover 2001). As soils develop, some mineral grains develop tunnel-like dissolution pits over a period of several thousand years (Jongmans et al. 1997; Smits et al. 2005; van Schöll et al. 2006). These tunnels have been attributed to the hyphae of ectomycorrhizal fungi, and the excretion of organic acids and dissolution of the mineral structure (Hoffland et al. 2004, 2005; van Schöll et al. 2008; Gazzè et al. 2012). Ectomycorrhizal fungal weathering of minerals has the potential to be an important source of plant nutrients (e.g., Paris et al. 1996; Landeweert et al. 2001; Blum et al. 2002; Glowa et al. 2003; Hoffland et al. 2004, 2005; Smits et al. 2012), although tunnels are likely to be quantitatively much less important than mineral surface weathering (Smits et al. 2005). Most mycorrhizal weathering experiments are lab-based (e.g., Paris et al. 1996; Adeleke et al. 2012; Smits et al. 2012) and give contrasting results. Field-based experiments, particularly those involving substrate bag incubations, show that fungal biomass increases in the presence of apatite particularly when plants are nutrient limited, but suggest that fungal responses are not sensitive to fungal species (Hagerberg et al. 2003; Wallander and Thelin 2008; Berner et al. 2012). Most studies have assumed that fungal tunnelling only occurs in ectomycorrhizal fungi; nonetheless, similar tunnels have been reported in areas where ectomycorrhizal fungi are absent, suggesting non-ectomycorrhizal or even non-fungal explanations in some instances (Sverdrup 2009). Similarly, whether localised weathering by mycorrhizal fungi contributes substantively to the total weathering budget induced by plant growth, organic matter decomposition and associated acidification of the the soil is not known. It is also worth noting that mycorrhizal helper bacteria, which have high mineral weathering capability (Koele et al. 2009; Uroz et al. 2009), may be particularly important in mineral tunnelling.

Tunnelling of primary minerals is a transient phenomenon in ecosystem development. The first tunnels appear in mineral grains coinciding with the depletion of the initial pool of available mineral nutrients, at around 2000 years but vary with parent material, climate, hydrology and vegetation (Hoffland et al. 2002; Smits et al. 2005). After this initial period, mineral tunnelling could increase mineral nutrient availability. However, eventually all primary minerals will have been weathered from the soil leaving no substrate for further weathering. The implications of this for long-term ecosystem development are unclear. By increasing the speed of weathering, mineral tunnelling could increase peak plant biomass, but could also ultimately increase the rate of nutrient loss from the ecosystem. Alternatively, the active use of mineral nutrients by ectomycorrhizal plants could inhibit leaching of these nutrients, thus decreasing the depletion of mineral nutrients and delaying retrogression. Curiously, two of the best-studied chronosequences in terms of P limitation, Franz Josef and Hawaii, both lack ectomyorrhizal fungi. Comparison of the dynamics of these sequences with sequences where ectomycorrhizal fungi are present could be informative, although any direct comparison is confounded by climate and parent material differences. A modelling approach may be better suited to address this question. For instance the model of Orwin et al. (2011) could be used to incorporate ectomycorrhizal weathering to predict shifts in mineral, recalcitrant and organic nutrient pools. A further advantage of the modelling approach would be to allow differentiation of mineral weathering effects from differences in organic nutrient cycling between mycorrhizal types.

Apart from the role of mycorrhizal fungi in nutrient cycling at the surface, many of the important mineral-nutrient pedogenic effects of mycorrhizal fungi occur deep in the soil profile. For example, in a temperate grassland chronosequence, Moore et al. (2010) reported that biological and geochemical cycling of nutrients was spatially separated: biological cycling occurred at the soil surface in a tightly coupled cycle, whereas geochemical nutrient cycles (weathering) occurred at >1 m depth. Unfortunately, most studies of mycorrhizal ecology do not sample below c. 30 cm depth (e.g., Dickie et al. 2002b), leaving deeper soil processes obscure. However, we know that ectomycorrhizal and arbuscular mycorrhizal fungi occur to at least 2–4 m depth in soils (Egerton-Warburton et al. 2003; Bornyasz et al. 2005) and that mycorrhizal effects on soil organic C may extend to at least 1 m depth (Chapela et al. 2001). Unravelling the role of mycorrhizal weathering deep in the soil profile should be a key priority for chronosequence research.

Mycorrhizal fungi – plant community interactions

Mycorrhizal fungal communities can be important determinants of plant community composition and drivers of plant–soil biota feedbacks (Bever et al. 2010; Klironomos et al. 2011). During early succession or transitions between mycorrhizal types (e.g., ectomycorrhizal tree succession into arbuscular mycorrhizal grasslands), positive plant–soil biota feedback appears common (Dickie et al. 2002a). Most of the evidence for positive biotic feedback involves established plants increasing the local density of compatible symbionts in a site where symbionts would otherwise be limiting (e.g., Dickie et al. 2002a; Ashkannejhad and Horton 2006; Nara 2006a, 2006b; Spence et al. 2011). Nonetheless, Mangan et al. (2010) found positive feedback in arbuscular mycorrhizal forest trees on Barro Colorado Island, Panama, suggesting that positive feedback can also occur within mature, relatively stable ecosystems. Further positive feedback may occur if mycorrhizal fungi serve as conduits for the transfer of nutrients, water or carbon between plants through hyphal linkages. Inter-plant transfer is clearly important in some achlorophyllous plants, but evidence for the ecological significance of hyphal-network resource transfers between photosynthetic plants remains equivocal (reviewed in Bever et al. 2010), perhaps due to the very small size of such transfers (Teste et al. 2010) relative to larger effects of seedling genetics and environment (Bingham and Simard 2013). Regardless of mechanism, where positive feedback occurs it has the potential to contribute to reduced plant diversity (Bever et al. 1997, 2010).

Negative plant–soil biota feedback occurs where plants modify soil communities such that other plant species have increased relative competitiveness. Negative plant–soil biota feedback can occur through increased populations of pathogens. Somewhat counter-intuitively, negative plant–soil feedback can also be driven by beneficial mycorrhizal symbionts (Bever 1999, 2002). This reflects the species-specific interaction between fungi and plants, with plant response to the symbiosis depending on the identity of both plant species and fungal species. Negative feedback does not necessarily require a negative symbiotic interaction, as feedbacks are due to the effect of soil on the relative growth of competing plant species. Negative plant–soil feedbacks can be important for maintaining plant diversity, but can also contribute to species turnover and successional change at larger spatial and temporal scales, particularly in early-successional communities (Kardol et al. 2006).

The extent to which specific plant–fungal interactions can be considered a mutualism rather than parasitism depends on environmental context (Hoeksema et al. 2010). Although no studies to date have examined the nature of this interaction along a chronosequence, previous work on soil nutrient availability suggests that interactions may become less parasitic and more mutualistic as nutrients (particularly P) decline along the sequence and greater nutrient limitation increases the relative benefits to the plant (Graham and Eissenstat 1994). Plants can respond to nutrient conditions by limiting carbon allocation to non-beneficial arbuscular mycorrhizal fungi (Bever et al. 2009; Kiers et al. 2011), thereby reducing the benefit of the association to both partners. However, plant species differ in their ability to adjust allocation in this way, probably related to differences in their root structures and associated capacity to access different nutrient pools (Grman 2012). Host-specific fungi may also be able to demand more resources from their hosts (Bruns et al. 2002), potentially explaining their frequency in early- to mid-succession.

A relatively new development in understanding interactions between mycorrhizal fungal communities and plant communities has been the application of network theory (Bascompte 2009). Network theory is a framework for the study of relationships (links) between objects (nodes). It is applied to a variety of network types, including neural, computer, social, and ecological networks, with the latter normally describing the architecture of interactions among species irrespective of their identity (Bascompte 2009). In ecology, this has expanded from earlier work on food webs to include a wider range of interactions, such as mutualism (Bascompte and Jordano 2007) and parasitism (Tylianakis et al. 2007). Community-wide plant–mycorrhizal fungus association networks have only recently begun to be quantified (Chagnon et al. 2012; Montesinos-Navarro et al. 2012), though practical constraints on identification have meant that these have not yet been studied in a replicated way (e.g., along chronosequences). Despite some potential caveats about the interpretation of these networks (Caruso et al. 2012), they provide a potentially valuable tool in understanding how the structure of mycorrhizal interactions between plant and fungal communities changes through ecosystem development.

Networks of interactions among mutualists tend to have a nested structure, whereby specialists interact with symbionts that also interact with generalists (Bascompte and Jordano 2007). Not only is this pattern widespread across regions and types of organisms, but it is known to be important for maintaining network stability (Bastolla et al. 2009; Thebault and Fontaine 2010). Therefore, association networks between plants and mycorrhizal mutualists would be predicted to have a nested structure, and this is supported by recent empirical work on plant–arbuscular mycorrhizal networks (Chagnon et al. 2012; Montesinos-Navarro et al. 2012).

Both mycorrhizal fungi and mycorrhizal plants vary strongly in their specificity, with fungi ranging from ultra-generalist species such as Cenococcum geophilum (Dickie 2007), to fungi specific to particular plant genera (e.g., Alnicola), and plants ranging from a few with extremely broad receptivity to fungi (e.g., arbutoid mycorrhizal Arctostaphylos uva-ursi; Krpata et al. 2007) to plants that associate with only a very few species of fungi (e.g., Pisonia spp.; Suvi et al. 2010; Hayward and Horton 2012). Smith and Read (2008) suggested that plant–mycorrhizal fungi associations become increasingly specialised through time, in keeping with Odum’s (1969) suggestion that niches become narrower as ecosystems develop. However, examples of ectomycorrhizal fungi specialising on early-successional species (see Ectomycorrhizal fungi along chronosequences and Fig. 1), and increasing evidence of host specificity in some arbuscular mycorrhizal fungi, suggest that plant compositional changes may drive turnover in the presence or absence of certain mycorrhizal fungi, rather than a consistent shift in specialisation along the chronosequence.

We suggest that application of network theory to mycorrhizal fungi across resource gradients has strong potential to enhance both our understanding of network theory and our understanding of mycorrhizal ecology. The outcome of network interactions can be highly dependent on the nature of the interaction (Thebault and Fontaine 2010). By grading from mutualism to parasitism depending on environmental context (Hoeksema et al. 2010), mycorrhizas provide a unique system that can help us to understand whether different interaction types among the same organisms can produce different network structures, as well as whether mutualistic network structure theory developed largely in the context of motile organisms (e.g., pollinators) can also be applied to non-motile mutualists. Recent empirical plant–mycorrhizal fungus networks (Chagnon et al. 2012; Montesinos-Navarro et al. 2012) suggest patterns consistent with those of mutualist networks involving mobile animals, though simulation studies have demonstrated that declining mobility of species leads to lower connectedness of interaction networks, higher asymmetries of interaction strengths between partners, and less predictable interaction patterns (Morales and Vázquez 2008). Therefore, as more plant–fungal networks are studied, important differences between their structure and those of other mutualist networks may emerge.

Conclusions

Despite the critical role of mycorrhizas in nutrient acquisition by plants, mycorrhizal fungi remain poorly integrated within the ecosystem development and retrogression paradigm. In part, this may reflect a widespread view of predictable shifts in mycorrhizal status over ecosystem development, despite evidence such shifts are not predictable (Table 1). Instead we find that single mycorrhizal types can dominate over very long periods and strong gradients of soil N and P. Chronosequences therefore present an ideal system for testing hypotheses about the role of mycorrhizas and mycorrhizal fungal community composition in ecosystems (Klironomos et al. 2011) and, conversely, the role of soil nutrients in fungal community composition and mutualistic interactions.

Across different mycorrhizal types there are striking differences in the types of ecosystems that have been studied. For example, an extensive literature has developed on ectomycorrhizal fungi and dark septate endophytes in early primary succession, yet virtually nothing is known of ericoid mycorrhizal fungi in primary succession. Conversely, ericoid mycorrhizal fungi are frequently studied in high soil organic matter, potentially retrogressive sites, where relatively little work has been done on arbuscular mycorrhiza or ectomycorrhizal fungi. Indeed, across most mycorrhizal types there is a lack of information on retrogressive sequences. Despite these limitations, some overarching conclusions emerge:

  1. 1.

    Existing models of predictable trajectories of mycorrhizal types over ecosystem development do not adequately describe the observed trajectories on well-described retrogressive sequences. Instead, we find that single mycorrhizal types can dominate over most or all of ecosystem development.

  2. 2.

    In early succession, dispersal limitation is an important factor shaping fungal communities. For ectomycorrhizal fungi, which have effective long-distance dispersal, this results in a predictable suite of fungi adapted to early succession. Arbuscular mycorrhizal fungi appear less predictable in early succession, perhaps due to less effective long-distance dispersal leading to a greater stochasticity and larger role of the regional species pool.

  3. 3.

    Pedogenesis is an important factor in shaping fungal communities through development of organic horizons, changing pH and altered nutrient availability. Although a literature has developed around vertical niche differentiation by ectomycorrhizal fungi, little is known of how soil pedogenesis influences arbuscular or ericoid mycorrhizal fungal communities. Even within ectomycorrhizal fungal ecology there has been no integration of vertical niche studies with pedogenesis, which is surprising given how closely pedogenesis and vertical profile development are linked.

  4. 4.

    Mycorrhizal fungi are important drivers of pedogenic processes, through physical binding of soil particles, exudation of organic compounds, shifting soil nutrient stoichiometry, and weathering of minerals. This suggests that the absence of major mycorrhizal types along some chronosequences (e.g., ectomycorrhizas at Franz Josef, NZ, and Hawaii, USA) could have important consequences for the way that ecosystems develop, and potentially limit generalisations across chronosequences, if differences in mycorrhizal type are not considered.

  5. 5.

    The interactions between fungal communities and plant species have been studied primarily in the context of plant–soil feedbacks and mainly for arbuscular mycorrhizal plants. Recent application of network theory to these complex community interactions shows promise as a technique for elucidating emergent properties, as well as for testing network theory itself.

Retrogressive sequences provide a unique opportunity to improve our understanding of how mycorrhizal communities respond to gradients of soil nutrient availability and physical structure. We hope that one outcome of this review may be to encourage mycorrhizal researchers and chronosequence researchers to look for ways to unify and integrate findings between the two fields of research. In particular, we suggest that chronosequences represent an underutilised natural experiment for mycorrhizal ecology. This is especially important for those mycorrhizal symbioses that are difficult to manipulate ex situ, including most arbuscular mycorrhizal fungi, some ectomycorrhizal fungi (e.g., Cortinarius, Russula), and orchid mycorrhizas. Better utilisation of these sequences will advance both mycorrhizal science and our understanding of ecosystem development and retrogression.