Interactions between animals and fungi are increasingly recognized as being crucial to ecosystem functioning. Although the role of individual species in structuring communities has long been recognized (Paine, 1966), few examples of belowground cross-phyla animal, plant, and fungal interactions have been described. Leaf-cutter ants (LCAs) in particular are widespread and can be found in Patagonia through the neotropics and into the deserts of Arizona and California and the warm subtropical savannahs of the southeastern United States (Mueller et al., 2017). In addition, they create large belowground structures that alter local carbon flows (e.g., Fernandez-Bou et al., 2019; Swanson et al., 2019; Fernandez-Bou et al., 2020b). Mycorrhizal fungus–plant mutualistic symbioses are even more widespread. They are found almost everywhere globally and play crucial roles in structuring communities and regulating ecosystem processes (Allen, 2022). Is there a direct interaction between mycorrhizal fungi and LCAs affecting community composition and ecosystem functioning? Indirect mycorrhizal fungus–ant interactions have been described, but direct interactions have rarely been documented (Gehring and Bennett, 2009). For example, herbivory by insects can reduce photosynthate allocation to roots and mycorrhizae, thereby reducing mycorrhizal activity and shifting fungal species composition (Gehring and Whitham, 1994). Mycorrhizae can also alter the abundance of both leaf parasites and their predators (Wooley and Paine, 2011), and of aphid populations (Chitty and Gange, 2022), so we would infer that mycorrhizae indirectly interact with ants by affecting their farming of aphids. In meadows and clearcuts within coniferous forests, ants moved mycorrhizal fungal inoculum, thereby facilitating plant recovery following a volcanic eruption (Allen et al., 1984). In semiarid shrublands, harvester ants constructed seed chambers using fine roots containing mycorrhizal fungi (Friese and Allen, 1993), resulting in the rapid formation of arbuscular mycorrhizae (AM) on abandoned nests. By contrast, succession from non-mycotrophic to obligately mycorrhizal plants is slow following other disturbances (Janos, 1980; Allen, 2022). In this study, we pose a novel question: are there LCA– mycorrhizal fungal interactions that have the potential to alter plant community composition?

In lowland neotropical rainforest, the vast majority of studies describing the functioning of mycorrhizae focus on AM, from succession to carbon fluxes (Janse, 1897; Went and Stark, 1968; Janos, 1980; Swanson et al., 2019). We designed instrumentation to study carbon cycling in a neotropical rainforest, which included an automated soil microscopic observation platform coupled to a soil sensor network (Rundel et al., 2009; Swanson et al., 2019). Previously, we demonstrated that LCAs alter gas fluxes between the soil and atmosphere (Fernandez-Bou et al., 2019; Fernandez-Bou et al., 2020a), and, as part of the study of methane production and consumption, we looked at the impacts of LCAs and El Niño phenomena on soil microorganisms (Aronson et al., 2019). From early studies in neotropical forests, an additional issue has always been of interest. We know that ectomycorrhizae (EM), orchid mycorrhizae (ORCM), and ericoid mycorrhizae (ERCM) are widely dispersed across these forests and that EM fungi are common locally in tropical rainforests (Alexander and Lee, 2005). However, we know little about the mechanisms of colonization at the community scale (Helbert et al., 2019; Delevich et al., 2021). Do the impacts of LCAs go beyond localized biogeochemical impacts to affecting the surrounding community? Do LCAs create patches that facilitate the establishment of mycorrhizal communities? Specifically, in this study, using our observatory systems, we asked the following questions: do LCAs import into the nest inocula of mycorrhizal fungi in the form of sporocarpic material, in addition to desirable leaves? In addition, could the physical alterations in nest soils created by the LCAs facilitate new colonization of these communities when nests are abandoned? Our focus was on the LCA nests located within alluvial soils of the La Selva Biological Station. If the mechanisms we propose exist, the importance of interactions between LCAs and mycorrhizal fungi becomes a crucial element in the dynamic stability of these complex communities.

Lowland neotropical rainforest is composed of a hyperdiverse array of plants, fungi, and animals. The extensive forests in the Central American and Amazon lowlands range from saturated and often standing water conditions, to occasional drier periods tied to El Niño–Southern Oscillation (ENSO) conditions. However, because of the climatic stability, rainfall occurs nearly daily, even during drier periods, and the temperatures often vary as much diurnally as they do seasonally, subject to periodic hot moments of climate events (Fernandez-Bou et al., 2020a). With generally stable environmental conditions, plant diversity can be extraordinarily high. In such situations, mycorrhizal fungal diversity does not necessarily track plant diversity. There exists a lower richness of AM fungi, a monophyletic, ancient group of fungi, the Glomeromycotina, which may predominate because they tend to be less host-specific and functionally more generalist (Allen et al., 1995; Steidinger et al., 2019). Among the vast variety of late-seral plants, AM fungi are well adapted to rainforest conditions (Janos, 1980). Across these hyperdiverse plant communities, communities of AM fungi form extended, interstitial hyphal networks nearly everywhere horizontally, and to several meters deep in the soil (Allen, 2022). In mature rainforests, AM fungi are found in every soil core examined, colonizing both large trees and understory herbs and forming a network occupying nearly every cubic mm ( Figure 1 ) of soil observed.

By contrast, there are also low-diversity tropical forest types, such as the high-elevation Dicymbe and Quercus patches of South America, which are dominated by EM (Nouhra et al., 2019; Delevich et al., 2021), and EM Dipterocarpaceae forests are widespread in Southeast Asia where environmental conditions appear to facilitate a high diversity of EM partners (Helbert et al., 2019). The establishment phase in these plant species-poor patches of EM forests is likely very different from that in high-plant-diversity AM forests. AM fungi show host preferences, but are not known to be species specific (Allen, 2022). EM are often more specific and have repeatedly and independently evolved from many taxa of fungi, ranging from Endogone in the Zygomycota to Ascomycota and Basidiomycota. EM plants in the Myrtaceae, Nyctaginaceae, Fagaceae, and some in the Caesalpinioideae form both AM and EM associations. Trees that form dual mycorrhizae, such as Populus, Quercus, and Eucalyptus, will initially form AM on extending roots into an AM network, then transition to EM as individual trees mature (Egerton-Warburton and Allen, 2001; Querejeta et al., 2007; Teste et al., 2020). Other plants form ERCM (primarily Ascomycota), and ORM (Ascomycota and Basidiomycota). Additional detail can be found elsewhere (Allen, 2022). Outside of tropical environments, where the dieback of both plant roots and fungi occurs seasonally, gaps open (Hernandez and Allen, 2013). With drought followed by rainfall events (Taniguchi et al., 2018), or following freeze–thaw events (Allen and Kitajima, 2014), new roots form new mycorrhizae. In seasonal tropical forests, AM predominate, but large disturbances, such as hurricanes, create tree blowdowns and tip-up mounds and thereby open patches that facilitate the establishment of EM (Hasselquist et al., 2011). At smaller scales, animals create disturbances (Aguirre et al., 2021) that facilitate the establishment of new plant–mycorrhizal symbioses. (In this case, EM pines invading into AM shrub/grasslands). However, plants such as the Gnetaceae appear only to form EM, and only with a low richness of fungal partners (Bechem and Alexander, 2011). How do these obligately EM and ERCM plants colonize new patches?

In a lowland neotropical field station, we observed Gnetum leyboldii forming widely dispersed patches across the forest. The large seeds are dispersed by water, birds, rodents, monkeys, and even fish (Won and Renner, 2006), and the seedlings form EM. One particular EM fungus observed fruiting in the G. leyboldii patch was Scleroderma sinnamariense. We do not know how this fungus is dispersed, but no rodent or peccary activity was observed: only that of ants. Wind turbulence in this forest floor is inadequate to disperse spores (Allen et al., 1993; Fernandez-Bou et al., 2020b). From a biodiversity perspective, opportunities to sustain EM and ERCM taxa among patches that shift around the landscape may become important. We know that EM and other mycorrhizae invade and establish themselves in patches within the highly networked AM matrix (Allen, 2022), but we do not how they do so. At extremely small scales, grazing by mesofauna can clip or even graze hyphae from roots, offering some opportunity for colonization (e.g., Fitter and Sanders, 1992), but the existing network is more likely to recolonize than are invading fungi. Disturbance is well known as a critical process in creating patches that allow other species to migrate and establish (Levine and Paine, 1974), thereby increasing community diversity. In tropical rainforests, by opening the canopy, LCA nests create patches of higher temperatures, lower relative humidity (Meyer et al., 2011), and lower nutrient availability (Meyer et al., 2013). These gaps could play a critical role in the ability of plants to reproduce and move by opening the canopy, and facilitating seed germination and photosynthesis (Denslow and Hartshorn, 1994). However, newly established seedlings need both a belowground and aboveground canopy gap. Large, infrequent disturbances (LIDs), such as hurricanes or volcanoes (Dale et al., 1998), can create opportunities for shifting mycorrhizal fungal communities (Hasselquist et al., 2011). However, in addition to these catastrophes, other small-scale disturbances must occur.

Why is this important? Community and ecosystem hierarchies interface where the community composition drives ecosystem processes (Rillig and Allen, 1999). The diverse evolutionary breadth of the fungi forming EM enables them to undertake many biogeochemical processes that AM cannot (Allen, 2022). Some of these EM may be critical for these communities to persist. Small patches of EM plants within an AM matrix are often an important element of these hyperdiverse forests. A key question then emerges: do gaps in the AM fungal network exist? Or does the establishment of EM, ORM, or ERCM, or even the new taxa of AMF, require a large-scale disturbance?

In old-growth neotropical forests, one organism in particular creates soil gaps and canopy gaps. The LCAs in the genus Atta create nests that can extend for many meters, opening patches both above and below ground. In the Costa Rican rainforest, nests of the LCA Atta cephalotes average 67 m², with 1.8 nests per ha. There are an average of 32 vents per nest, covering a total of 1.2% of the soil surface (Fernandez-Bou et al., 2019). The nests generally have a depth of between 40 cm and 60 cm at the La Selva Biological Station (Swanson et al., 2019). They create gaps that extend up through the canopy, creating chimneys large enough for gas exchange and sunflecks for seedling photosynthesis to occur ( Figure 1 ). Internally, AM fungi and root standing crops within the nest is not significantly different from that in the surrounding forest, but the net primary production of roots and mycorrhizal fungi is higher in the nests as turnover is significantly faster (Swanson et al., 2019). The nests turn over approximately on a decadal time scale (Perfecto and Vandermeer, 1993). This very turnover could also provide opportunities for new root–mycorrhizal fungal symbioses if the appropriate inoculum is present.

Our study was undertaken at the La Selva Biological Station (Organization of Tropical Studies) in the lowlands of the Caribbean basin of northeastern Costa Rica. The site is in a lowland tropical wet forest, with an average annual rainfall of 4,260 mm (1986–2015) (McDade et al., 1994; Fernandez-Bou et al., 2020a). The average atmospheric temperatures are relatively constant (≈ 25°C). Our site was in an old-growth forest (i.e., older than 1,000 years). The soils at La Selva are derived from volcanic activity, particularly lava flows. At the higher elevations, the volcanic parental material weathered into residual soils. At lower elevations, including in our sites, soils are of alluvial origin, formed in river terraces during the Pleistocene and are classified as Oxisols.

The long-term persistence of communities (animals, plants, fungi, and other microbes) depends in part on an ability to be dynamic in both composition and activity. The tropical rainforest, similar to all other ecosystems, has persisted through variable climates, geological transformations, and biological invasions. The importance of LIDs cannot be underestimated, as hurricanes, volcanoes, earthquakes, and human tillage all transform soils, associated microbes, and plant communities (Lodge and McDowell, 1991). With LIDs, symbionts also invade and change, in part due to soil animal activity (Allen et al., 2018). However, subtle shifts may also occur with small perturbations, facilitating the adjustment of existing communities to less-disruptive change (Morris et al., 2016). For example, global environmental shifts are resulting in elevated CO₂ and N deposition, along with projected increased temperatures and decreased water availability (Green et al., 2019). For ancient forests to persist, biological symbionts also need to be dynamic in activity and composition. Animals such as LCAs that perturb soils may be especially crucial in maintaining forests and associated microbial species composition.

Similar to human farmers, LCAs create a unique environment that supports the LCA community. They construct “cities” that facilitate the importation of resources (leaves) for an agricultural (in this case, fungal) base. These cities/nests are a network of tunnels that also promote the disposal of wastes, and in the process, facilitate aeration. What other materials might the ants be exploiting? High-protein sources (e.g., fungal sporocarp biomass), unknown vitamins, or trace elements could also be limiting and facilitated by importation. One means of ensuring their self-sufficiency involves LCAs creating disturbances that not only provide for their own survival in a complex ecosystem, but also provide opportunities for increasing the establishment of diverse elements across that community, creating opportunities for shifting other taxa beyond the need for a LID to initiate succession.

Although LCAs select the leaves of particular plant species to move into nests, the criteria for selection remain unknown and other materials may well be utilized (Rockwood, 1976; Masiulionis et al., 2013). We have observed that not all individuals are carrying leaves; ants also carry other materials into the nest. Could some of those other materials be fungal biomass for improving the growth of the Leucoagaricus spp. or used as food for the ants themselves? Two materials that have not been addressed are the peridium and the gleba of the sporocarp of fungal fruiting bodies. Masiulionis and colleagues (Masiulionis et al., 2013) observed basidiocarps of Psilocybe coprophila being collected by Acromyrmex lobicornis LCAs in Argentina, and Allen (2022) observed that peridium and sporocarp top material of Russula sp. was cut and transported by ants in a temperate forest. LCA individuals were observed on the S. sinnamariense sporocarp, and the peridium and gleba of many of the S. sinnamariense sporocarps were observed being clipped by ants. The gleba and peridium are composed of living hyphae. Fragments of these tissues, along with interspersed basidiospores, could serve as local inoculum. This removal and transport could account for the S. sinnamariense DNA material within the nests. Several questions emerge from these observations. Why are LCAs moving sporocarp material into nests? Are they looking for added N due to the high C-to-N ratio of the leaf material? Are there specific nutrients, vitamins, or even antibiotics that outside fungal material provides? Once invertebrates are in the nest, their populations could readily disperse inoculum such as S. sinnamariense. Collembola, symphylans, and midge larvae, among the many invertebrates found within the nests, could have readily consumed much of the sporocarpic material while simultaneously redistributing any spores from the sporocarp.

Tropical rainforest communities are predominantly AM (e.g., Allen et al., 1995). The existing AM mycorrhizal fungal networks appear to occupy nearly all of the soil space available, an example of a concept known as community coalescence (Rillig et al., 2016). In EM-monodominant forests, seedlings may acquire the majority of fungi from networks extending from neighboring mature trees (Delevich et al., 2021), but in hyperdiverse AM communities, opportunities for the establishment of EM plants are likely to be limited. In these cases, disturbance gaps become critical to creating opportunities for patches of EM plants, for the invasion of new AM fungi, or even for terrestrial orchid and ericoid mycorrhizal plant and fungal taxa. Therefore, LCA nests may well provide some of these opportunities.

In this study, we have shown that the opportunity exists for LCAs to facilitate transitions in mycorrhizal types and fungi. The ants themselves transport materials, and the chambers and tunnels connect, forming highways for invertebrates, fungi, and roots to navigate the dense clay soil and find nutrients, and enable symbionts to find each other and facilitate gas exchange (Fernandez-Bou et al., 2019; Swanson et al., 2019). The AMR observations suggest that after LCAs abandoned nests, new roots rapidly regrew into any gaps. Roots are positively geotrophic and grow rapidly downward into the reforming soil. In contrast, mycorrhizal fungi are not geotropic, growing in all directions, but are sensitive to the volatile compounds produced by growing root tips (Koske and Gemma, 1992). This means that there is a window of opportunity for any viable fungi brought into contact with newly growing roots, thereby affecting the establishment of a community structure.

The vertical architecture of soil structure and organisms may well be crucial for new mycorrhizal formations. In a sagebrush shrub steppe, roots rapidly grew down before surface inocula could infect them. Newly wind-dispersed inoculum was of limited effectiveness for initiating new mycorrhizae (Friese and Allen, 1991). Therefore, wind-dispersed fungal spores from other locations are unlikely to initiate new mycorrhizal symbioses, as these are deposited on surfaces, where infectable root tips are rarely located. The LCAs and opportunist invertebrates, by bringing EM and other mycorrhizal fungal material deeper into the nest where new roots grow, provide the spatial structure facilitating new mycorrhizal infections. G. leyboldii, S. sinnamariense, and likely many other EM plant and fungal taxa, may well have narrow associations (Allen et al., 1995; Tedersoo and Põlme, 2012), and two symbionts’ finding each other could lead to the reduced expansion or even survival through stress periods.

The observations of numerous invertebrates within the LCA nests along with the mycorrhizal fungi in comparison with the adjacent non-nest soils suggests that LCA nests are hotspots for biodiversity (Folgarait, 1998; Lavelle et al., 2022). The interactions between ants and mycorrhizal fungi likely go well beyond just LCAs, given the myriad of other taxa burrowing into soil in a similar manner, albeit on a smaller scale. Far more research into both community and ecosystem functioning and biological interacting networks is an opening into a far more complex world than even ecologists have acknowledged.

In scaling up, we need a better spatial understanding of the roles played by EM, ERCM, and ORM in tropical rainforest ecosystems, which are dominated by AM plants, and a better sense of the time scales on which these dynamic ecosystems function. A better understanding of these processes will allow us to better predict and manage these crucial ecosystems as our global environment changes. Using cores allows us to gain insight into ecosystem processes. However, only an infinitesimally small amount of the forest can be sampled from cores at infrequent intervals. For example, in the most comprehensive survey to date of fine-root standing crop (Espeleta and Clark, 2007), conducted across 18 plots that were 0.5 ha in size using a 47.4-mm diameter corer, only approximately 0.001% of the surface area was actually analyzed. Our SEO increased the temporal resolution but at a limited number of locations. Coupling campaigns and SEO observations increased our spatial and temporal resolution, but the number of data to be processed and understood also dramatically increased. Newer approaches are needed to merge sensing–observation–sequencing datasets. In addition, applying these newer approaches to critical processes such as CO₂ sequestration and fluxes in complex environments ranging from rainforest to alpine ecosystems is becoming not just an interesting research direction, but crucial to our own survival under global change. As biodiversity declines in response to human-driven change, what organisms, symbioses, and interactions are we losing? How are these shifts altering global biodiversity and ecosystem processes? The LCAs–mycorrhizal fungus–mycorrhizal plant model might be one such model system for further study.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA749330.

Ethical review and approval was not required for the study on animals in accordance with the local legislation and institutional requirements.

The idea for this research came collectively from MA, PR, TH, EA, and HS. The first draft of the manuscript was written by MA, and reviewed and edited by PR, TH, HS, and EA. EA undertook the soil sampling, and HS undertook the sequencing. HS and EA completed the sequence analysis. All authors contributed to the article and approved the submitted version.