Introduction

In the absence of major soil-rejuvenating processes (e.g., glaciations, volcanic eruptions), concentrations of soil phosphorus (P), one of the key plant macronutrients, decline over time, due to erosion and leaching (Walker and Syers 1976; Turner and Laliberté 2015), especially after fires (Lambers et al. 2014; Wittkuhn et al. 2017). By contrast, concentrations of nitrogen (N) first increase during primary succession, and then decline over time (Walker and Syers 1976; Laliberté et al. 2012; Turner and Condron 2013). In seasonally-dry climates, loss of N through fires plays a major role (Wittkuhn et al. 2017). Compounding the decline of soil P stocks, much of what remains of the inorganic P is tightly bound (sorbed or occluded) to soil particles, with the remainder of the P being locked away in relatively inaccessible organic forms (Walker and Syers 1976; Turner and Condron 2013; Turner and Laliberté 2015). As a result, old soils tend to be severely P-impoverished, and, consequently, P is the major macronutrient limiting plant productivity (e.g., Richardson et al. 2004; Laliberté et al. 2012; Hayes et al. 2014).

A general trend, most pronounced in the northern hemisphere, is of increasing plant diversity approaching the tropics (i.e. with decreasing latitude) (Huston 1994). However, soil age tends to increase, and soil fertility decreases towards the equator, so changes in plant diversity may have little to do with latitude per se (Huston 2012). Studies from soil chronosequences, series of soils formed from the same parent material, but of contrasting soil ages, reduce many of the confounding factors that affect broad-scale latitudinal gradients (e.g., changes in climate) and have demonstrated consistent increases of plant diversity with soil age, suggesting that declining P availability is associated with increasing plant diversity (Wardle et al. 2008; Laliberté et al. 2013; Zemunik et al. 2016).

Several global biodiversity hotspots (Myers et al. 2000) contain P-impoverished soils, e.g., Southwest Australia, Cape Floristic Province in South Africa (McArthur 1991; Witkowski and Mitchell 1987; Cowling et al. 1996b), and campos rupestres in Central Brazil (Oliveira et al. 2015; Silveira et al. 2016). In these biodiversity hotspots, soil produced in old, climatically-buffered landscapes (OCBILs) has undergone little rejuvenation over geological time scales (Hopper 2009), and the plant species in the communities have evolved a suite of adaptations to those P-impoverished conditions (Lambers et al. 2014). In south-western Australia, for example, the general relationship of increasing plant diversity with decreasing soil P concentrations can be observed at the regional scale (Lambers et al. 2010), but it is most pronounced for species with roots specialised for the acquisition of P, with a variety of non-mycorrhizal species using a carboxylate-releasing P-mining strategy (Figs. 1a, b). Along a coastal dune chronosequence within the Southwest Australian hotspot, the same pattern of greater diversity of non-mycorrhizal species with decreasing P concentrations occurs (Zemunik et al. 2015, 2016) (Fig. 1c). This is a clear illustration of the point that soils in OCBILs are not uniformly old and nutrient-impoverished, and this juxtaposition of different soil ages/geologies has been linked to species turnover (Cowling et al. 2014).

Fig. 1
figure 1

(a) Species richness of mycorrhizal (open circles) and non-mycorrhizal (NM) plant species (filled circles) in relation to total soil phosphorus (P) concentrations, from a comprehensive floristic survey of over 1000 quadrats in the wheatbelt of Western Australia by Gibson et al. (2004). (b) Species richness of three main NM species groups from the same floristic survey as in (a). (c) Relative canopy cover of NM species in relation to total soil P concentrations, from floristic survey plots within the Jurien Bay chronosequence (Zemunik et al. 2015). The log curve fits for the NM monocot and NM Proteaceae species were better by AIC than linear fits (AIC values: 64.3 versus 91.0 for the NM monocot species; 90.5 versus 121.5 for the NM Proteaceae species). Solid lines indicate significant (P < 0.05) regression relationships: (a) P = 0.006; (b) P = 0.0007 for monocot NM species, P = 0.0001 for Proteaceae species; (c) P = 0.006

Full size image

The majority (>80%) of vascular plant species form mycorrhizal symbioses (Brundrett 2009). Mycorrhizas are structures produced by associations of soil fungi and roots, and these specialised structures transfer poorly-available nutrients, especially P, but also N, that are acquired by mycorrhizal hyphae in soil to the plant; in return, the plant supplies the fungus with carbon compounds derived from photosynthesis (Brundrett 2009; Smith and Read 2008). Mycorrhizas allow plants to scavenge P that is in the soil solution beyond the zone that roots and root hairs can deplete. Many non-mycorrhizal species, in contrast, have evolved root specialisations that combine morphological and physiological adaptations (e.g., cluster roots and dauciform roots that release large amounts of carboxylates) capable of extracting all of their required nutrients from the soil without the use of mycorrhizas (Lambers et al. 2015d; 2008; Shane and Lambers 2005). Cluster-rooted plants (e.g., Lupinus and Hakea species) efficiently mine P from P-impoverished soils by quickly producing clusters of fine, densely packed rootlets that release carboxylates in an ‘exudative burst’ to solubilise P that is tightly sorbed to soil particles (Lambers et al. 2008; Shane et al. 2004; Watt and Evans 1999). Mycorrhizas are an effective scavenging strategy for moderately infertile soils, but a carboxylate-releasing P-mining strategy is superior in severely P-impoverished soils (Lambers et al. 2008, 2015a) (Fig. 2). Yet, mycorrhizal species coexist with non-mycorrhizal species that release carboxylates on severely P-impoverished soils. In this review, we develop conceptual models explaining the coexistence of mycorrhizal and non-mycorrhizal species, with a focus on the roles of (i) belowground facilitation and (ii) interactions with root-associated soil biota, particularly mycorrhizal fungi and soil-borne pathogens.

Fig. 2
figure 2

Effects of carboxylates on mobilisation of phosphorus (P); other exudates may have similar effects, e.g., polygalacturonate (Nagarajah et al. 1970). Carboxylates (organic anions) are released via a carboxylate channel. The manner in which phosphatases are released is unknown. Carboxylates mobilise both inorganic (Pi) and organic (Po) P, which are both sorbed onto soil particles. At acid pH, Pi and Po bind to oxides and hydroxides of iron (Fe) and aluminium (Al); at alkaline pH, these compounds are precipitated by calcium (Ca). The carboxylates effectively take the place of Pi or Po, thus pushing this into solution. The released phosphatase enzymes hydrolyse Po compounds, after they have been mobilised by carboxylates. For further explanation, see text; modified after Lambers et al. (2015c)

Full size image

Facilitation is a positive interaction that can occur between plants and results in an increase in species performance of at least one of the interacting species (Callaway 1995). Recent reviews of the available data show that facilitative mechanisms can contribute to the maintenance of biodiversity patterns globally (McIntire and Fajardo 2014; Peay 2016; Stachowicz 2001) and can also help better explain biodiversity-ecosystem function relationships across ecosystems (Wright et al. 2017). Both arbuscular mycorrhizas and ectomycorrhizas not only play a role in nutrient acquisition, but also in defence against pathogens (Marx 1972; Azcón-Aguilar and Barea 1997 Wehner et al. 2010). This has been known for a long time (Marx 1969), but its ecological implications are now receiving increasing attention (Maherali and Klironomos 2007; Sikes et al. 2009). The role of mycorrhizas in defence is not restricted to pathogens (Bennett et al. 2017; Cameron et al. 2013), but also includes protection against nematodes (Vos et al. 2013) and herbivores (Minton et al. 2016).

Several forms of mycorrhizal symbioses (e.g., arbuscular mycorrhizal (AM), ectomycorrhizal (ECM), ericoid and orchidaceous mycorrhizal) and non-mycorrhizal specialisations (e.g., cluster roots and dauciform roots) exist, with many of these occurring side-by-side, both within the same plant (Reddell et al. 1997; de Campos et al. 2013) and in neighbouring plants (Zemunik et al. 2015). The diversity of these P-acquisition strategies, like that of species diversity, often increases with decreasing P availability (Lambers et al. 2008, 2014; Zemunik et al. 2015). This is puzzling, since the P-mining strategy is superior in severely P-impoverished soils (Lambers et al. 2008). Despite a trend of decreasing richness and abundance of mycorrhizal plant species with declining soil P concentrations, mycorrhizal species do persist at relatively high abundances, even on the most severely P-impoverished soils (e.g., Lambers et al. 2014; Zemunik et al. 2015). Arbuscular mycorrhizal plant species, for example, remain abundant on the poorest soils (Zemunik et al. 2015), even though species of AM fungi investigated to date appear to have limited capabilities of accessing all pools of soil P (Yao et al. 2001; Smith et al. 2015). Is the P acquisition of mycorrhizal plants facilitated by growing in close proximity with specialised non-mycorrhizal neighbours? Is the success of non-mycorrhizal species compromised by greater susceptibility to pathogens, thereby preventing competitive exclusion? These pivotal questions are addressed in this review to provide answers to why high terrestrial plant diversity occurs on severely P-impoverished soils (Lambers et al. 2010). We develop an argument that a protective role of mycorrhizal fungi accounts for the very high plant species diversity in severely P-impoverished landscapes (Laliberté et al. 2015; Teste et al. 2017). We explore this particular aspect throughout this review.

Mycorrhizal fungi

The vast majority of empirical studies on mycorrhizal symbioses provide unequivocal evidence that these root symbiotic structures may enhance plant acquisition of poorly-mobile nutrients, especially P (Smith et al. 2015). The mycorrhizal symbioses achieve this primarily via their extraradical hyphae (ERH) that can scavenge poorly-available nutrients beyond root depletion zones (Li et al. 1991a, b; Owusu-Bennoah and Wild 1979; Yao et al. 2001). Therefore, a vital component of mycorrhizal fungi are the ERH for both the plants and fungi, since it is the component that physically explores the soil matrix (Smith and Read 2008). When soil P availability is low, plants will allocate more resources to mycorrhizal fungi and their ERH, and alleviate soil nutrient limitations of plant growth (Smith and Read 2008).

Recently, Teste et al. (2016) reported on field experiments with fungal in-growth cores along a retrogressive 2 million-year old coastal dune chronosequence (Hayes et al. 2014; Laliberté et al. 2012), with the oldest soils along this sequence being some of the most severely P-impoverished soils ever described (Turner and Laliberté 2015; Turner et al. 2017). Interestingly, these severely P-impoverished soils show very high local plant species diversity, with a large proportion of non-mycorrhizal plant species (Zemunik et al. 2015, 2016). This chronosequence stretches inland over 10 km and represents one of the strongest natural soil fertility gradients that have been characterised. It exhibits a ~ 60-fold decline in total soil P concentration, and strong and clear shifts from N to P limitation of plant growth with increasing soil age (Laliberté et al. 2012; Turner and Laliberté 2015; Turner et al. 2017).

Teste et al. (2016) measured fungal biomass with key lipid biomarkers (NLFA 16:1ω5, and PLFA 18:2ω6,9) at three contrasting stages of soil development along this chronosequence. Sampling was done using closed, mesh and open (two no-core treatments: control and undisturbed) fungal in-growth cores during five months, including the wet winter months, which made it possible to study the dynamics of mycorrhizal fungal scavenging. Given the relatively high abundance of mycorrhizal species, surprisingly, there was consistently less AM and ECM fungal biomass in the oldest, most severely P-impoverished soils compared with that in the younger soils (Fig. 3a). This trend was found across all winter months (June, July and August) and at all levels of the core treatment except for the closed cores (Fig. 3a). Extraradical hyphal scavenging in these different-aged soils partly supported the trends found with the mycorrhizal fungal biomass data. In particular, it was found that there was significantly less AM fungal scavenging in the old soils compared with that in the young soils (Fig. 3b).

Fig. 3
figure 3

(a) Mycorrhizal fungal biomass in soils of different ages along the Jurien Bay chronosequence as measured with the neutral lipid fatty acid (NLFA) biomarker (16:1ω5) a proxy for arbuscular mycorrhizal (AM) fungi, and the phospholipid fatty acid (PLFA) biomarker (18:2ω6,9) a proxy for ectomycorrhizal (ECM) fungi. Fungal biomass was measured with the fungal in-growth core treatments allowing root and hyphal infiltration (i.e. all except the PVC treatment following Teste et al. (2016) during the winter months (June, July, August) at Jurien Bay. (b) Mycorrhizal fungal scavenging as expressed as biomass differences between the mesh and PVC core treatments in August at the end of the winter (Teste et al. 2016). This is a reanalysis of data from Teste et al. (2016). Percent root colonisation by AM showed a significant hump-shaped pattern along soil age, where colonisation levels were low in the oldest soils (Teste et al. 2016) partly supporting the biomass responses in (a) and (b)

Full size image

The findings of Teste et al. (2016) are intriguing. For the first time in a natural field study, mycorrhizal fungal biomass and the biomass of their ERH was found to be considerably less in the most severely P-impoverished soils, despite extensive root colonisation, high root biomass (Laliberté et al. 2017) and high abundance and diversity of mycorrhizal fungi (Krüger et al. 2015) and host plants (Zemunik et al. 2016). Teste et al. (2016) suggested that extremely low P availability in soil constrains ERH biomass, similarly to what Abbott et al. (1984) and Jones et al. (1992) found in detailed glasshouse studies with a range of soil P-levels. These findings also suggest that the productivity of mycorrhizal fungal ERH is limited by declining soil P availability, as well as declining soil pH (AM only). The results from this study support the conceptual framework proposed by Lambers et al. (2008) which argues that mycorrhizal species are less successful than carboxylate-releasing P-mining species in severely P-impoverished soils. However, that does lead to the question: why are there so many mycorrhizal plants, in terms of both species numbers and their cover on severely P-impoverished soils (Lambers et al. 2014; Zemunik et al. 2015, 2016)? We explore this question in the following sections, focusing on two possible, not mutually exclusive explanations: facilitation of mycorrhizal species by non-mycorrhizal species, and greater susceptibility of non-mycorrhizal plants to soil-borne pathogens.

Facilitation

Three broad facilitative mechanisms have been proposed: (i) indirect biotic facilitation; (ii) abiotic facilitation leading to nutrient enrichment (belowground facilitation for soil N and P, and nutrient exchanges between plants); and (iii) classic abiotic facilitation leading to microclimate amelioration (Wright et al. 2017). Here, we briefly focus on reviewing the role of abiotic facilitative mechanisms, explaining how mycorrhizal species can benefit from the carboxylate-releasing P-mobilising non-mycorrhizal plants. Then we discuss biotic mechanisms (see sections on pathogens below) for maintaining biodiversity.

Soil nutrient enrichment that results from belowground facilitation has been well documented in ecosystems inhabited by plants whose productivity is more limited by N than by any other nutrient (Wright et al. 2017). In particular, positive effects of legumes and their associated N2-fixing biota on local soil nutrient enrichment and the growth of neighbouring plants have been demonstrated (Vitousek et al. 2013). However, non-legume nutrient enrichment such as enhanced P mobilisation, due to the production of nutrient-mobilising root exudates (Lambers et al. 2013; Li et al. 2007), has not received as much attention, and few studies have assessed if these enrichments are ecologically important.

We propose a number of hypotheses that could be tested in model systems such as hyperdiverse shrublands in South Africa (Cowling and Richardson 1995; Cowling et al. 1996a) and Australia (Lamont et al. 1977; Lambers et al. 2014) that host both N- and P-impoverished soils (i.e. where either N or P limit plant productivity), and also in other old, climatically-buffered, infertile landscapes, such as the campos rupestres of Brazil (Oliveira et al. 2015; Silveira et al. 2016), South America’s Pantepui (Hopper 2009), and parts of the pampas in southern South America (Sainz Rozas et al. 2012; Hopper et al. 2016).

Nutrient mobilisation-based facilitation

Nutrient mobilisation-based facilitation in crop plants is not only relevant for soil macronutrients such as P, but also for micronutrients such as iron (Fe), zinc (Zn) and manganese (Mn), since these nutrients are also mobilised by cluster-rooted plants (Li et al. 2014). Therefore, this process can improve uptake of these nutrients by neighbouring non-mobilising species provided their root systems are sufficiently close at the time the nutrients have been made available. Plant-based mechanisms underpinning this type of facilitation may involve the release of large quantities of acid phosphatases, protons, carboxylates and chelating substances from roots which increases the concentration of P and other nutrients in the soil solution (Li et al. 2007; Lambers et al. 2013). Recent microcosm studies support this contention, since it was found that nutrient-mobilisation facilitation occurs and improves plant growth (Muler et al. 2014; Teste et al. 2014). We surmise that nutrient-mobilisation facilitation in natural ecosystems may be more common than previously thought (Li et al. 2014; Wright et al. 2017); this hypothesis requires more formal testing as proposed in this review (Fig. 4).

Fig. 4
figure 4

Hypothesised nutrient exchanges between neighbouring ‘nutrient-donor’ plants; (a) shortly after fire and (b) ~10 years after fire. Relative abundance and importance of the exchanges and pathways are shown with increasing font and arrow width. *Importance varies with distance to neighbour and mass flow is pervasive, but not considerable in most soils. Shortly after a fire, N2-fixing plants ultimately benefit more than P-mining plants, since they would grow even faster in the presence of the interactions outlined in (a). Some time after fire (>10 years), a different set of interactions is expected to lead to more benefits and faster growth by the P-mining plants, as outlined in (b)

Full size image

Since soil N and P supply changes drastically with time after fire (Certini 2005), we expect that changing fire regimes due to climate change will alter plant belowground strategies for N and P acquisition in species-rich ecosystems (Fig. 5). Belowground facilitation is also predicted to play an increasingly important role in mediating the negative impacts of harsher environmental conditions due to global climate change (Brooker et al. 2008). Since more frequent fires and less annual precipitation (Williams and Bradstock 2008; Indian Ocean Climate Initiative 2012) are predicted to occur in south-western Australia, we therefore expect that belowground facilitation, for N and P acquisition in particular, will play an increasingly important role in the maintenance of biodiversity in these ecosystems.

Fig. 5
figure 5

Source-sink relationships and soil nitrogen (N) and phosphorus (P) levels as drivers of nutrient exchanges between plants with time after fire. Belowground vectors of transfer are mechanisms that may operate and vary in importance with time since fire. Mass flow (mostly for N, but also for P in very sandy soils and high-rainfall environments (Huang et al. 2017)) is likely prominent soon after fire following significant rain. Root intermingling plays a role throughout, especially for species growing in close proximity. Mycorrhizal fungi and the formation of mycorrhizal networks become pivotal for exchanges later when the fungal communities have completely re-established. In general, arbuscular mycorrhizal (AM) and ectomycorrhizal (EM) fungal abundance is reduced shortly after fire, and subsequent recovery with time and litter accumulation occurs without any long-term effects on colonisation or spore abundance (McMullan-Fisher et al. 2011). The availability of N and P generally follow the pattern shown here, but is mainly dependent on fire intensity (Burrows and Wardell-Johnson 2003)

Full size image

Nutrient exchange-based facilitation

Plants have evolved diverse strategies to gather soil nutrients and the high functional diversity in severely-impoverished soils was suggested to be linked to high species richness in shrublands of Australia and South Africa (Lamont 1982). These belowground nutrient-acquisition strategies interact when different plant species grow close to each other (Cu et al. 2005; Gardner and Boundy 1983; Horst and Waschkies 1987; Orians and Milewski 2007). Indeed, plants are capable of exchanging (i.e. ‘sharing’) nutrients with their neighbouring plants via root intermingling or mycorrhizal networks (Teste et al. 2009, 2014, 2015; Jakobsen and Hammer 2015; Simard et al. 2015). While microcosm studies show clear evidence for facilitation (Muler et al. 2014; Teste et al. 2014), further work is required in natural ecosystems. So far, we do not yet know whether ecologically relevant amounts of P or micronutrients exchange between non-mycorrhizal and mycorrhizal plants during the cluster-root exudative ‘burst’, where neighbouring mycorrhizal plants could take advantage of this sudden mobilisation of nutrients (Gardner et al. 1983; Lambers et al. 2013; Muler et al. 2014; Teste et al. 2014).

Belowground nutrient exchanges are likely initiated early after fire, and we hypothesise that plants acting as so-called ‘nutrient-donor plants’ (Fig. 4) are determined by soil nutrient availabilities that vary with time since fire (Certini 2005). Thus, the relative importance of N and P exchanges is also expected to vary with time after fire (Fig. 5), because N levels in the soil build up as a result of an abundance of N2-fixing plants (i.e. legumes, actinorhizal species and cycads) that effectively fix atmospheric N2 and release some of that N to the soil (DeLuca et al. 2008; Halliday and Pate 1976; Hansen and Pate 1987; Wall 2000). For example, within the first two years after fire, at least 1 kg ha−1 yr−1 of N can be fixed by N2-fixing plants such as Acacia species (Hingston et al. 1982), and ~10% of this fixed N can be transferred between plants based on interplant N transfer studies (He et al. 2009). However, the ability to fix N2 strongly depends on P availability, since N2 fixation is a process requiring substantial amounts of P (Raven 2012; Valentine et al. 2017; Vitousek et al. 2002). A few years (>5) after fire, soil P becomes less available, and the abundance of N2-fixing plants declines (Fig. 5); however, they are not excluded from the system (Pate and Beard 1984), and they may benefit from and be sustained by interactions with P-mining plants (Fig. 5). As proposed by Power et al. (2010), a P-availability threshold likely exists for the N2-fixing plants. Therefore, we suggest that when the soil has a low N:P ratio, N2-fixing plants are most abundant with high species richness (Fig. 5), and propose that a P availability threshold exists for the N2-fixing plants to successfully remain competitive over other species (Pekin et al. 2011). Legume species that co-occur with P-mining species must obtain enough P to avoid competitive exclusion, but the mechanisms allowing coexistence with surrounding plants are not fully understood and require further testing (Figs. 4, 6).

Fig. 6
figure 6

(a) Acacia pulchella (Fabaceae, producing nitrogen-fixing root nodules) with yellow flowers, together with Banksia species (Proteaceae, producing phosphorus-mobilising cluster roots), B. attenuata (foreground) and B. prionotes (background) or (b) Banksia sessilis growing in very close proximity shortly after fire in south-western Australia. (Photos: (a) Graham Zemunik, and (b) François P. Teste)

Full size image

In the case of P-mining plants, shortly after fire (< 5 years), these plants become enriched in 15N, resembling more closely that of the atmosphere, thus possibly resulting from exchanges of N with N2-fxing plants (Schmidt and Stewart 1997). However, it is not known how relevant belowground nutrient exchanges are in this process and what the N sources are (Schmidt and Stewart 1997). We hypothesise that P-mining plants benefit from N-nutrient exchanges with the abundant N2-fixing plants (Figs. 4, 6), and later (> 5 years after fire) they can contribute to maintaining the N2-fixing plants in the system via P-nutrient exchanges (Fig. 4). Co-occurrence of P-mining and N2-fixing plants on a site is common after fire and some species are often found growing in close proximity where roots and their mycorrhizal fungi can interact (Fig. 6). Furthermore, the soil N concentrations shortly after fire can be spatially variable (Hopmans et al. 2005), where N2-fixing plants may form small N-rich patches for neighbouring plants. Similarly, P-mining plants may form ephemeral ‘available-P’ patches for neighbouring plants.

Various legume species in south-western Australia form both AM and ECM along with N2-fixing nodules (Albornoz et al. 2016a; Brundrett 2009; Png et al. 2017), giving rise to the possibility for an elaborate guild of structures for nutrient exchanges between plants that have never been documented (Fig. 4). We further hypothesise that belowground nutrient exchanges may also promote plant species coexistence via resource sharing that may form the basis for indirect resource partitioning, such as belowground interactions between N2-fixing plants and P-mining plants (Fig. 4). However, to establish how these facilitative mechanisms remain ecologically and evolutionarily stable will require further study.

Source-sink relationships have been used to conceptualise the potential flow of nutrients through mycorrhizal fungi from plants with greater nutrient levels to plants with lower nutrient levels (Simard et al. 2012). Soon after fire, N2-fixing plants have a high demand for P-demanding molecules (e.g., ATP and rRNA for making nitrogenase) for symbiotic N2-fixation (Raven 2012; Paul 2014). These N2-fixing plants are P sinks at this early stage, and may meet their P requirements with the help of scavenging mycorrhizal hyphae and/or the intermingling of roots around cluster roots of neighbouring P-mining species (Teste et al. 2014). Such scavenging may represent an efficient strategy for obtaining P in extremely P-impoverished soils; legumes also show higher root phosphatase activity than co-occurring non-legume species, suggesting that they are particularly good at acquiring organic P (Png et al. 2017). Both AM and EM fungi could access newly mobilised P from cluster root ‘burst’ microsites, but EM may be favoured, since they possess high-affinity P transporters at their mycelia front (Cairney 2011). Phosphorus-mining plants, in contrast to N2-fixing plants, may act as N sinks shortly after fire, due to the N losses during fire (Wittkuhn et al. 2017). These N demands could be offset by transfer of N from N2-fixing plants (Fig. 4). Root intermingling could allow N to move between roots, due to leakage and/or root turnover, ultimately raising the possibility of N transfer from N2-fixing plants to P-mining plants (Fig. 4). There is great potential for discovering an impressive array of unique belowground root interactions geared to promoting plant coexistence and retaining nutrients in these ecosystems.

Concluding remarks on belowground facilitation and the maintenance of biodiversity

Local plant species coexistence and the maintenance of biodiversity in nutrient-poor ecosystems can be promoted by a multitude of interacting mechanisms such as resource partitioning (McKane et al. 2002; Turner 2008), plant-soil feedback (Teste et al. 2017), degree of dependency on mycorrhizal fungi (Hart et al. 2003), or as proposed here, by way of belowground facilitation due to complementary nutrient-acquisition strategies and nutrient exchanges.

The role of mycorrhizas in defence against pathogens

Nutrient uptake is widely accepted as the primary role of mycorrhizal fungi in contributing to plant fitness (Smith et al. 2015). However, increasing evidence over the last few decades has shown that mycorrhizas can also offer their hosts defence against pathogens (Marx 1972; Maherali and Klironomos 2007; Sikes et al. 2009; Wehner et al. 2010; Albornoz et al. 2016b). They can provide physical barriers around roots against soil-borne pathogens (Marx 1972), produce antimicrobial compounds (Duchesne et al. 1988a, b), or neutralise the negative effects of pathogens for seedling survival and growth (Liang et al. 2015). In this section, we summarise the current evidence of pathogen defence of both ECM and AM fungi, and discuss their implications for plant-plant interactions in hyperdiverse ecosystems.

Mechanisms of pathogen defence

There are several mechanisms by which mycorrhizal fungi provide defence against pathogens to their hosts that are not necessarily mutually exclusive (Sikes et al. 2009). They do, however, seem to differ between ECM and AM fungi. Both ECM and AM fungi can competitively exclude pathogenic organisms from roots and the rhizosphere, and induce changes in the root system (Marx 1972; Pozo et al. 2002; Wehner et al. 2010). For example, ECM fungi develop a hyphal mantle around their roots tips that acts as a physical barrier (Marx 1969; Branzanti et al. 1999), while AM fungi can induce production of root callose around infected root cells (Pozo et al. 2002). On the other hand, while ECM fungi can produce a vast number of antibiotic compounds (Marx 1972; Strobel and Sinclair 1991), there is as yet no evidence that AM fungi produce such compounds. However, there is evidence that AM fungi modify the microbial community in the rhizosphere, favouring organisms capable of producing such antibiotics (Wehner et al. 2010).

The extent of the protection offered by mycorrhizas also differs between ECM and AM fungi. The protection provided by ECM fungi to roots against pathogens tends to be localised (Marx 1972). That is, within the same plant, non-mycorrhizal roots can be more susceptible to infection than ECM roots of the same plant (Marx 1972). On the other hand, AM fungi can provide not only localised protection, but they can also induce systemic resistance (Pozo et al. 2002). For example, a split-root experiment conducted by Pozo et al. (2002) evaluated the role of AM fungi in protecting tomato plants against the oomycete Phytophthora parasitica. They found local production of protective enzymes by AM-colonised roots, and these enzymes were also found in non-mycorrhizal roots of the same individual. Additionally, Herre et al. (2007) showed that colonisation by AM fungi can offer systemic protection against the foliar pathogen Phytophthora palmivora. We infer that a small amount of AM root colonisation could be sufficient to offer a systemic defence against pathogens, while the pathogen susceptibility of the ECM hosts is negatively and directly related to the amount of ECM root colonisation.

Since mycorrhizal symbiosis depends on soil fertility (Smith and Read 2008), the protection against pathogens offered by mycorrhizas is potentially context-dependent, and thus, pathogen susceptibility can also depend on soil properties. Several studies have shown the effects of soil nutrients, pH and moisture on both the amount of root colonisation and mycorrhizal community composition (Krüger et al. 2015; Schappe et al. 2017). Neighbouring plants in a community can also have strong effects on root colonisation by mycorrhizal fungi (Dickie et al. 2004; Kennedy et al. 2012; Teste et al. 2014). Furthermore, different mycorrhizal fungal species can produce contrasting antibiotic compounds or mantle structures (Marx 1972), and, hence, offer different levels of root protection. Therefore, altering mycorrhizal fungal communities can potentially have strong effects on the way in which hosts respond to pathogen infections.

Susceptibility to pathogens: the role of suberised endodermis and exodermis

A major role of plant roots is the uptake of nutrients and water from soil (Marschner 1995; Steudle and Peterson 1998). Simultaneously, they must be able to exclude potentially harmful substances and prevent the entry of pathogens into the roots. This selectivity in root function is accompanied by a complex root anatomy (Esau 1977; Schreiber et al. 1999). However, our knowledge of the anatomy of Proteaceae roots is scarce. Here, for the first time, we discuss comparative detailed anatomical structures for three selected Proteaceae species, all endemic to south-western Australia, and all producing roots with specialised P-mining clusters: harsh hakea (Hakea prostrata) and acorn banksia (Banksia prionotes), which both grow in weathered soils and seasonally-dry environments, and river banksia (Banksia seminuda), which also grows in weathered soils, but in high-rainfall zones.

Detailed anatomical and histochemical studies of Proteaceae roots reveal species-specific differences. Soil-grown H. prostrata roots develop an early endodermis (internal sheath) with a complete ring of suberin lamellae as early as 20 mm from the apex; however, at this stage, the root does not have an exodermis (external sheath; Fig. 7a). This strongly suberised internal sheath isolates the stele from the rest of the root, and is likely the main barrier against pathogen invasion into the xylem and spread throughout the plant. At 50 mm from the apex, roots develop a patchy exodermis in which some cells do not develop suberin lamellae, and continue to grow as ‘passage cells’ (Fig. 7a). This patchy exodermis is unlikely to be an efficient barrier against pathogen invasion into the root. At this stage, roots have a strongly-suberised and complete endodermis without passage cells. In the stele, strongly-lignified penta-poler xylem vessels are apparent, which is different from typical tetra-poler dicotyledonous xylem vessels (Esau 1977). Extra xylem poles likely provide an extra path for transport of water and minerals, which radially move from the soil into the stele (Fig. 7a).

Fig. 7
figure 7

Deposition of suberin (a) and lignin (b) in roots of Hakea prostrata, Banksia prionotes growing on the Jurien Bay sand dunes on the west coast of south-western Australia, and B. seminuda growing on the Warren sand dunes on the south coast of south-western Australia. Roots of H. prostrata and B. prionotes were collected at the end of the dry season (April), seedlings of B. seminuda were collected in December and grown in pots in a garden till February. Cross-sections were taken at 20 mm and 50 mm from the apex of a non-cluster root and stained with fluorol yellow 088 or phloroglucinol. The presence of suberin lamellae was detected by yellow-green fluorescence (white arrowheads); the presence of lignin was detected by orange-red staining (black arrowheads). Bars = 100 μm

Full size image

The structure of B. prionotes and B. seminuda roots differs from that of H. prostrata. While weakly-developed, an incomplete endodermis is apparent in B. prionotes at 20 mm from the apex; brighter stains in the cell walls of inter-xylem poles reveal a stronger endodermis in B. seminuda (Fig. 7a). At 50 mm, the endodermis of B. prionotes has relatively strong suberin depositions with bright stains compared with the younger zone, yet it is not complete. Some endodermal cells near the xylem poles remain as passage cells without depositing suberin lamellae. These cells may provide an easy path that has low resistance for water and nutrient transport into the xylem. However, as roots age, more passage cells deposit suberin lamellae, and thus their numbers decline (Enstone et al. 2003). In contrast to B. prionotes, B. seminuda roots develop a strongly suberised, complete ring of endodermis, which is similar to that in H. prostrata (Fig. 7a). Both Banksia species do not develop a suberised exodermis up to 50 mm behind the root tip, which is different from H. prostrata. Absence of a suberised exodermis indicates (1) lack of filtration of ions from the soil solution, (2) low resistance for radial water transport from the soil solution into the root, (3) rapid carboxylate exudation (citrate and malate) from roots into the rhizosphere to first solubilise and then absorb sparingly-soluble mineral nutrients, especially P (Lambers et al. 2015b; Shane and Lambers 2005).

Not only Proteaceae, but also Fabaceae such as Lupinus angustifolius, L. luteus, Cicer arietinum and Glycine max, lack a suberised exodermis in their roots, even under stressful culture conditions (Bramley et al. 2009; Hartung et al. 2002; Perumalla et al. 1990; Ranathunge et al. 2008). Most of these Fabaceae, as well as many other Fabaceae, are known to release large amounts of carboxylates (Pearse et al. 2006; Veneklaas et al. 2003). On the other hand, monocots such as Oryza sativa, Saccharum officinarum, Triticum aestivum and Zea mays do produce a suberised exodermis (Clark and Harris 1981; Perumalla and Peterson 1986; Perumalla et al. 1990). Monocots that have been studied often release some specific exudates, but do not release large amounts of carboxylates, unlike many Proteaceae and Fabaceae (Delhaize et al. 1993; Li et al. 2013; Ma et al. 2003; Oburger et al. 2014; Pearse et al. 2006; Sun et al. 2016). There is good information on angiosperm species in specific families that do or do not produce an exodermis; the vast majority of angiosperms does produce an exodermis (Perumalla et al. 1990). However, there is a distinct lack of knowledge on the carboxylate exudation or leaf manganese concentrations, which can be used as a proxy for carboxylate concentrations in the rhizosphere (Lambers et al. 2015c) of angiosperms of which we know the root structure. We surmise that species that release large amounts of exudates have evolved to strategically modify their root structure for rapid carboxylate exudation in order to efficiently take up P from nutrient-poor soils. This hypothesis requires further testing involving a wider range of plant families.

Absence of a suberised exodermis may have some negative consequences, such as (1) provide easy access for pathogens to enter into the root, and (2) loss of water and nutrient ions from the roots to the dry soil by back-flow (Hose et al. 2001; Ranathunge et al. 2008; Thomas et al. 2007). In some plant species, such as Glycine max (soybean), ‘diffuse suberin’ in the epidermal cell walls fulfils the requirement of an exodermis, which is lacking in soybean. Diffuse suberin in the epidermis acts as a physical and chemical barrier for the penetration of Phytophthora sojae, an oomycete causing soybean root rot disease (Ranathunge et al. 2008). However, there is no histochemical evidence indicating the presence of ‘diffuse suberin’ in the epidermal cell walls of Banksia species studied so far. Instead, they exhibit intense deposition of phenolic compounds in the cell walls of the entire cortex, as indicated by bright autofluorescence (Fig. 7a). Such soluble phenolic compounds, which are associated with the suberin polymer, are known to act as antifungal agents (Biggs and Miles 1988; Kolattukudy 1984; Lulai and Corsini 1998; Thomas et al. 2007). The presence of a suberised endodermis, on the other hand, serves as the last line of defence before pathogens invade the vascular cylinder and spread throughout the plant (Enkerli et al. 1997; Enstone et al. 2003; Huitema et al. 2004; Kolattukudy and Espelie 1989; Thomas et al. 2007).

Not only suberin but also lignin, another complex aromatic polymer, which is deposited in the secondary cell walls of all vascular plants, plays a major role in defence against abiotic and biotic stresses, especially against pathogens and insects (Bonawitz and Chapple 2010; Moura et al. 2010). Lack of suberin in the exodermis of B. prionotes is associated with depositing more lignin into the exodermal cell walls (sub-epidermal cell layer) than in the two other species (Fig. 7b). This modification would be expected to have positive adaptive functions, especially against pathogens. Lignin acts as a non-degradable ‘physical barrier’ with inter-monomeric C-C linkages that inhibit fungal pathogen penetration into the root (Lygin et al. 2009; Moura et al. 2010; Richter 1996). Thus, lignification of the exodermis would certainly reduce the risk of being vulnerable to pathogen attack. In contrast, H. prostrata and B. seminuda, which contain a strongly developed and completely suberised endodermis show weak lignification in the exodermis (Fig. 7). However, certain exodermal cell walls of H. prostrata also show mild staining for lignin, perhaps by developing tertiary cells walls, but not as pronounced as in B. prionotes (Fig. 7b).

The anatomical studies of roots of three Proteaceae species show that, remarkably, none of them forms a complete suberised exodermis. Similar observations have been made for Fabaceae. Lack of this important feature in Proteaceae and Fabaceae roots may be an adaptive trait enabling rapid carboxylate exudation into the rhizosphere and acquisition of scarcely available nutrients from severely-impoverished soils in south-western Australia. However, the absence of an exodermis makes roots more vulnerable and susceptible to soil-borne pathogens. Taken together, these observations suggest that root anatomical adaptations for efficient nutrient acquisition from severely P-impoverished soils trade-off against root defence against soil-borne pathogens.

Soil-borne pathogens and negative density dependence

Soil-borne pathogens are often considered as a threat to native plant diversity in natural ecosystems. This view is based mostly on research that has studied the ecological impacts following the introduction of exotic, virulent soil-borne pathogens into new ecosystems. For example, the introduced oomycete Phytophthora cinnamomi has major negative impacts on plant biodiversity in south-western Australia, because a large number of native plant species are highly susceptible to it, leading to ‘dieback’ (Coates et al. 2014; Shearer et al. 2004). However, the traditional view of soil-borne pathogens as ‘threats’ to biodiversity has been gradually changing, as it is now recognised that native, less virulent pathogens that have co-evolved with a given flora can actually play a positive role in the maintenance of local plant species diversity (Bever et al. 2015).

How can soil-borne pathogens possibly contribute to local plant species coexistence? One explanation is a process termed ‘negative conspecific density dependence’, whereby plants suffer disproportionately from being exposed to a high density of conspecific neighbours (Terborgh 2012) (Fig. 8). This could be due to the build-up of root pathogens, when conspecific plant density increases, with these pathogens having either host-specific effects or stronger negative effects on conspecific individuals than individuals of other plant species. Although this idea has recently seen a ‘renaissance’ in plant ecology (Bever et al. 2015), it dates back to the work of Gillett (1962) on the ecological importance of pests. Later, the idea that pests could promote local plant diversity through negative conspecific density dependence was formalised by Janzen (1970) and Connell (1971) as an explanation for high tree species diversity in tropical rainforests.

Fig. 8
figure 8

How belowground interactions with soil biota can lead to (a) negative density dependence or (b) mono dominance, depending on a plant nutrient-acquisition strategy. In (a), inoculum of soil-borne pathogens builds up in the root zone of mature non-mycorrhizal plants, leading to poor conspecific seedling survival and growth. Such negative density dependence promotes local plant species coexistence by preventing dominance by one species. In (b), seedlings of an ectomycorrhizal (ECM) plant species show greater survival and growth near a conspecific mature plant because they recruit into an established ECM fungal network that provides resistance against soil-borne pathogens, while also enhancing P acquisition. Moreover, ECM fungi can directly suppress the activity of soil-borne pathogens. This can lead to monodominance by the ECM species. Figure taken from Laliberté et al. (2015) with permission from Wiley

Full size image

More recently, pathogen-mediated negative density dependence was taken out of its initial tropical rainforest context, and touted as a possible explanation for the high local plant diversity in south-western Australian shrublands (Laliberté et al. 2015). At first glance, these seasonally-dry shrublands are quite different from species-rich tropical rainforests, but both tend to occur on old, strongly-weathered soils where soil P availability is very low. Laliberté et al. (2015) hypothesised that such strong P limitation of plant growth in these ecosystems influences plant susceptibility to root pathogens, because many root traits that enable efficient P acquisition from P-poor soils (e.g., high specific root length, high root density, lack of suberised exodermis) trade-off against root longevity and defence to soil-borne pathogens (Laliberté et al. 2015). Consequently, it was suggested that non-mycorrhizal, cluster-rooted Proteaceae in P-impoverished shrublands from south-western Australia might be particularly efficient at mobilising P, yet be quite susceptible to soil-borne pathogens, whereas mycorrhizal species would be better defended against pathogens (Laliberté et al. 2015).

The interaction between pathogens and mycorrhizal fungi can promote plant diversity by several mechanisms (Laliberté et al. 2015; Marx 1972; Newsham et al. 1995a), such as negative density dependence, or by reducing the competitive strength of dominant plant species which allows subordinate species to coexist (Terborgh 2012). For example, Albornoz et al. (2017) grew non-mycorrhizal Proteaceae and ECM Myrtaceae from P-impoverished south-western Australian shrublands, and found that ECM species are not affected by various native Phytophthora species isolated from these ecosystems, whereas non-mycorrhizal cluster-rooted species are susceptible. When non-mycorrhizal and ECM plants are grown in competition with each other, inoculation with Phytophthora reduces the competitive superiority of non-mycorrhizal plants to the benefit of ECM species (Albornoz et al. 2017). Interestingly, there is a positive correlation between ECM root colonisation and ECM plant growth, but only in the presence of Phytophthora (Fig. 9), suggesting the ECM fungi play a stronger role in root defence against pathogens than in P acquisition in these severely P-impoverished habitats (Albornoz et al. 2017). Further support for this concept is provided by Teste et al. (2017), who grew 16 plant species with different nutrient-acquisition strategies in a glasshouse. Plants were exposed to a sterilised soil inoculum, either collected from the rooting zones of conspecific mature plants, or from the rooting zones of species sharing the same or a different nutrient-acquisition strategy. Consistent with the hypothesis proposed by Laliberté et al. (2015), non-mycorrhizal plants showed reduced survival when exposed to a conspecific soil inoculum relative to a sterile inoculum, suggesting a role of pathogens. By contrast, the opposite was found for ECM species, which grew better when exposed to a conspecific inoculum relative to a sterile one (Teste et al. 2017). Other studies have also shown the defensive properties of mycorrhizal fungi against root pathogens (Branzanti and Zambonelli 1994; Branzanti et al. 1999; Marx 1972; Wehner et al. 2010). These studies, taken together with results from Albornoz et al. (2017), suggest that in old and severely-weathered soils, plant diversity may be enhanced by a trade-off between highly-efficient P-acquisition strategies and susceptibility to soil-borne pathogens (Laliberté et al. 2015). This mechanism could be an important factor in promoting plant coexistence in the south-western Australian biodiversity hotspot and possibly elsewhere where Proteaceae play a prominent role, e.g., fynbos in South Africa (Allsopp et al. 2014).

Fig. 9
figure 9

Conceptual figure modified after Albornoz et al. (2017), which includes experimental data for three species. It shows the relationships between ectomycorrhizal root colonisation and seedling biomass in the presence or absence of Phytophthora species. The solid green line indicates the presence of Phytophthora and a significant relationship (P ≤ 0.05), while the red dashed line indicates the absence of Phytophthora and a non-significant relationship (P ≥ 0.05)

Full size image

Albornoz et al. (2016b) concluded that a decline in ECM fungal richness and an increase in ECM root colonisation during ecosystem development along the Jurien Bay chronosequence was in favour of a small number of ECM fungal species that were the most efficient at acquiring P. Based on further studies, Albornoz et al. (2017) inferred that in the same old soils, one of the ECM fungi’s main roles could be defence against pathogens. Hence, mycorrhizal fungi may not only be driven by their nutrient-acquisition efficiency, but, in severely P-impoverished soils, more so by their ability to protect their hosts from pathogens. These results show a dual role of nutrient uptake and pathogen defence of ECM fungi. Other studies have also shown the facilitative effects that plants can have on species with different nutrient-acquisition strategies (Li et al. 2014; Teste et al. 2014). We surmise that species that are less efficient at acquiring nutrients coexist with species with superior nutrient-acquisition strategies by possessing better defence against pathogens, possibly tapping into nutrients made available in the rhizosphere of their less well-defended competitors (Fig. 10).

Fig. 10
figure 10

Diagram showing the net outcome of interaction between a mycorrhizal (plant on the left) and a non-mycorrhizal Proteaceae (plant on the right) in the absence (left panel) and presence (right panel) of native soil-borne pathogens. Drawing produced by Javier F. Tabima

Full size image

In the preceding sections, we argued that cluster-rooted species are particularly susceptible to pathogens. It has been shown, however, that cluster roots of Lupinus albus release a range of compounds that inhibit the activity of both bacteria and fungi (Tomasi et al. 2008; Weisskopf et al. 2006; Weisskopf et al. 2005). There is some evidence that cluster roots of Embothrium coccineum (Proteaceae) function in a similar manner (Delgado et al. 2015). Interestingly, the chitinases released by cluster roots of L. albus (Weisskopf et al. 2006) are ineffective against the cell walls of oomycetes (water moulds) such as Phytophthora and Pythium, because of their different cell-wall structure (Badreddine et al. 2008; Wessels 1994). Therefore, oomycetes, rather than fungi or bacteria, likely are the major pathogens affecting Proteaceae in hyperdiverse systems (Albornoz et al. 2017; Laliberté et al. 2015).

Concluding remarks on protection against pathogens

Two decades ago, a key question was about the paradoxical abundance of non-mycorrhizal species in severely P-impoverished landscapes (Lambers et al. 2006). A series of studies showed that these non-mycorrhizal P-mining species have a superior P-acquisition strategy based on releasing large amounts of carboxylates (Lambers et al. 2008). Then, a new key question emerged: why are there so many P-scavenging mycorrhizal species in severely P-impoverished environments, when this strategy is presumably inferior for acquiring the growth-limiting resource (Lambers et al. 2014)? Now, we conclude that an important role of mycorrhizal fungi in severely P-impoverished megadiverse landscapes is protection against pathogens, rather than nutrient uptake. Johnson et al. (1997) proposed that mycorrhizal associations fall into a mutualism-parasitism continuum based on several studies finding that mycorrhizal root colonisation can have positive to non-existent to negative effects on plant growth. We argue that this proposed continuum does not fully capture recent observations by focusing only on nutrient uptake and discarding other potential benefits that mycorrhizal fungi offer to their hosts, such as pathogen defence (Newsham et al. 1995b). Findings by Teste et al. (2016) showed overall low external mycorrhizal hyphal biomass in a severely P-impoverished habitat, indicating relatively little nutrient scavenging compared with that in other ecosystems, where the P availability is greater. Based on the evidence provided here, we propose that mycorrhizal fungi may not necessarily act as ‘parasites’ when they do not increase host nutrient uptake (Albornoz et al. 2017). Rather, their main effect on plants may change, depending on the environmental context (Sikes et al. 2010).

Perspectives

We focused on recent studies conducted along the Jurien Bay chronosequence located in a biodiversity hotspot (Laliberté et al. 2012), but placed these in as broad a context as possible. These studies have their limitations. Albornoz et al. (2017) only evaluated the role of pathogens in coexistence of plants with two nutrient-acquisition strategies. There are more than 350 plant species along the Jurien Bay chronosequence with contrasting nutrient-acquisition strategies (Zemunik et al. 2015). Hence, results from Albornoz et al. (2017) do not necessarily explain the persistence of other nutrient-acquisition strategies, such as species with dauciform roots (Cyperaceae) or sand-binding roots (Haemodoraceae). The situation for these may be very different, because Cyperaceae (Enstone et al. 2003; Perumalla et al. 1990) and Haemodoraceae (Layla Aerne-Hains and Simpson 2017) do have an exodermis. A trade-off between P-acquisition efficiency and pathogen defence in combination with facilitation of nutrient acquisition (Muler et al. 2014) may not be the only mechanism that maintains hyperdiverse ecosystems. This trade-off may explain why some species exhibit both a cluster-root strategy and mycorrhizas, e.g., Casuarinaceae, including Casuarina cunninghamiana (Reddell et al. 1997) and Allocasuarina humilis (Png et al. 2017), and some Fabaceae, e.g., Viminaria juncea (de Campos et al. 2013). It may not be a case of ‘belt and braces’, i.e. using two complementary strategies, but a matter of using cluster roots for effective nutrient acquisition and mycorrhizas for defence against pathogens.

Future studies should test the generality of the new hypotheses proposed in this review by using other nutrient-acquisition strategies and other nutrient-poor, highly diverse ecosystems, such as the fynbos in South Africa, tropical rainforests in South America and Asia (Myers et al. 2000), campos rupestres in Brazil (Oliveira et al. 2015; Silveira et al. 2016) or the pampas grasslands in southern South America (Hopper et al. 2016; Sainz Rozas et al. 2012). These future studies will not only enhance our understanding of the functioning of biodiversity hotspots, but also underpin management strategies to conserve or restore megadiverse ecosystems. Also, future research should aim to quantify the relative importance of different factors allowing plant coexistence, such as competitive and facilitative interactions through soil microbiota, and their implications for plant diversity. Furthermore, since cluster roots are ephemeral and likely fully deplete the scarce nutrients from their rhizosphere and proficiently remobilise the nutrients from their cluster roots (Shane et al. 2004), facilitation would be effective only if the roots of facilitated plants intermingled with cluster roots at the right time. This would require sensing where the cluster roots are going to be produced at a very early stage of cluster-root development. Such sensing would involve signalling molecules, likely volatiles that are released by roots or decaying microorganisms (Peñuelas et al. 2014; Waters et al. 2016; Weidenhamer 2016), which are affected by antimicrobial compounds released from developing cluster roots (Cesco et al. 2012; Delgado et al. 2015; Weisskopf et al. 2006).

Finally, disentangling the relative importance of the complementary roles of mycorrhizal fungi in nutrient uptake and pathogen defence is difficult, because these occur simultaneously. Future studies should aim at quantifying the relative contribution of the different benefits mycorrhizas can have for plant health by conducting multifactorial experiments using pathogen and mycorrhizal inoculum, and modifying soil nutrient availability (Maherali and Klironomos 2007; Sikes et al. 2009).