Trophic mediation by a fungus in an ant–plant mutualism


  • Céline Leroy,

    1. CNRS, UMR Ecologie des Forêts de Guyane, Campus Agronomique, 97379 Kourou Cedex, France
    Search for more papers by this author
  • Nathalie Séjalon-Delmas,

    1. Université Paul Sabatier, Laboratoire Surfaces Cellulaires et Signalisation chez les Végétaux, Pôle de Biotechnologie Végétale, 24 Chemin de Borde Rouge, B.P. 42 617 Auzeville, 31326 Castanet-Tolosan, France
    Search for more papers by this author
  • Alain Jauneau,

    1. CNRS, IFR40 Pôle de Biotechnologie Végétale, Plateforme d’imagerie cellulaire de Toulouse, 24 Chemin de Borde Rouge, B.P. 42 617 Auzeville, 31326 Castanet-Tolosan, France
    Search for more papers by this author
  • Mario-Xavier Ruiz-González,

    1. CNRS, EDB (Laboratoire Evolution et Diversité Biologique), 118 route de Narbonne, 31062 Toulouse, France
    2. Université de Toulouse, EDB, 118 route de Narbonne, 31062 Toulouse, France
    Search for more papers by this author
    • Present address: Instituto de Biología Molecular y Celular de Plantas, Consejo Superior de Investigaciones Científicas–Universidad Politécnica de Valencia. C/Ingeniero Fausto Elio s/n. 46022 València, Spain.

  • Hervé Gryta,

    1. CNRS, EDB (Laboratoire Evolution et Diversité Biologique), 118 route de Narbonne, 31062 Toulouse, France
    2. Université de Toulouse, EDB, 118 route de Narbonne, 31062 Toulouse, France
    Search for more papers by this author
  • Patricia Jargeat,

    1. CNRS, EDB (Laboratoire Evolution et Diversité Biologique), 118 route de Narbonne, 31062 Toulouse, France
    2. Université de Toulouse, EDB, 118 route de Narbonne, 31062 Toulouse, France
    Search for more papers by this author
  • Bruno Corbara,

    1. Clermont Université, Université Blaise Pascal, LMGE, B.P. 10448, 63000 Clermont-Ferrand, France
    2. CNRS, UMR 6023, LMGE, 63177 Aubière, France
    Search for more papers by this author
  • Alain Dejean,

    1. CNRS, UMR Ecologie des Forêts de Guyane, Campus Agronomique, 97379 Kourou Cedex, France
    Search for more papers by this author
  • Jérôme Orivel

    Corresponding author
    1. CNRS, UMR Ecologie des Forêts de Guyane, Campus Agronomique, 97379 Kourou Cedex, France
    2. CNRS, EDB (Laboratoire Evolution et Diversité Biologique), 118 route de Narbonne, 31062 Toulouse, France
    3. Université de Toulouse, EDB, 118 route de Narbonne, 31062 Toulouse, France
      Correspondence author. E-mail:
    Search for more papers by this author

Correspondence author. E-mail:


1. Plants often rely on external, mutualistic partners to survive and reproduce in resource-limited environments or for protection from enemies. Such interactions, including mycorrhizal symbioses and ant–plant associations, are widespread and play an important role at the ecosystem and community levels. In ant–plant mutualisms, the plants may benefit from both the protection provided by the presence of ants and from the nutrients absorbed from insect debris. However, the role of third partners in plant nutrition, particularly ant-associated fungi, has never before been demonstrated.

2. We investigate this issue in the ant–plant Hirtella physophora. In this model system, Allomerus decemarticulatus ants are involved in two, highly specific interactions: first, with their host plant, and, secondly, with a fungus that they actively manipulate. Moreover, the ants combine both plant trichomes and fungal hyphae to make a trap to capture prey.

3. We empirically demonstrate the existence of a third type of interaction between the fungus and the plant through the use of both experimental enrichments with stable isotopes (15N) and histological approaches. The fungus growing in the galleries plays a role in providing nutrients to the host plant, in addition to the structural role it plays for the ants. Fungus-facilitated nitrogen uptake occurs mainly in old domatia, where abundant hyphae are in close contact with the plant cells. Whether the fungi inside the domatia and those in the galleries are the same is still uncertain.

4.Synthesis.Together, our results show that a fungal partner in an ant–plant mutualism can benefit the plant by improving its nutrient uptake, and they demonstrate the existence of a true tripartite mutualism in this system. Our results add further evidence to the notion that interpretations of some ant–plant symbioses as purely protective mutualisms have overlooked these nutritional aspects.


Light and nutrient availability, as well as herbivory, are among the main factors limiting plant growth (Coley, Bryant & Chapin 1985; Baraloto, Bonal & Goldberg 2006; Lopez-Toledo et al. 2008). As fixed organisms, plants have developed a variety of strategies to cope with environmental pressures. Such strategies include the development of life-history traits that enable the growth, survival and reproduction of individuals, as well as phenotypic plasticity and mutually beneficial interactions with other organisms (Sultan 2000; Poorter et al. 2005). Mutualisms are ubiquitous in natural communities, and, as the primary source of carbon, plants are involved in a great diversity of such interspecific interactions. In particular, symbioses with mycorrhizal fungi, known to concern 90% of all terrestrial vascular plants, play a major role in enhancing plant productivity by supplying limited nutrients to the plants, and thus contributing to the structure of plant communities (Smith & Read 2008). It has been shown that mycorrhiza-infected plants are more resistant to the root infections caused by pathogens (Azcón-Aguilar & Barea 1996; Selosse, Baudoin & Vandenkoornhuyse 2004). Many plant species also rely on arthropods, mostly ants and Acari, to protect their foliage from herbivores (Gaume & McKey 1998; Heil & McKey 2003), and it has been widely demonstrated that mutualistic ant–plant associations play an important role in tropical ecosystems (Heil & McKey 2003; McKey et al. 2005). For myrmecophytic species, such protective interactions range from facultative and diffuse associations to strict and obligate mutualisms (Hölldobler & Wilson 1990). Myrmecophytes typically provide their guest ants with nesting space in hollow structures called domatia, and also, in certain cases, food resources (Beattie 1985; Fonseca 1999; Beattie & Hughes 2002). Besides the protective aspect of this mutualism, ants may also provide nutrients to their host plant (Beattie 1989; Benzing 1991; Sagers, Ginger & Evans 2000; Fischer et al. 2003). These nutrients are supplied to the plant by the ants through their debris or faeces that accumulate in the domatia, and can be absorbed directly through the rhizomes, roots or protuberances, or through the walls of the domatia (Janzen 1974; Rickson 1979; Treseder, Davidson & Ehleringer 1995; Sagers, Ginger & Evans 2000; Fischer et al. 2003; Solano & Dejean 2004; Watkins, Cardelus & Mack 2008).

Here we focus on the mutualistic interaction between the neotropical ant–plant Hirtella physophora Mart. & Zucc. (Chrysobalanaceae), its specific plant–ant associate Allomerus decemarticulatus Mayr (Myrmicinae), and a third fungal partner, a sooty-mould that the ants use to build galleries under the stems of their host plant (Dejean et al. 2005). To build these galleries, the A. decemarticulatus workers first cut the plant’s trichomes that they then assemble together to form a frame. Finally, they manipulate the mycelium of the sooty-mould to consolidate the frame, and the galleries are then used as a trap to ambush prey. Recent studies on this ant–fungus interaction have highlighted its high specificity (Ruiz-González et al. 2010). Only a few ant–fungus associations have been reported to date in which the fungi have a primarily structural role in nest and gallery construction (Dejean et al. 2005; Schlick-Steiner et al. 2008; Mayer & Voglmayr 2009), although the cultivation of fungal strains for nutritional purposes is an obligate habit in Attini ants (Mueller, Rehner & Schultz 1998; Currie et al. 1999). Recently, Defossez et al. (2009) also demonstrated the occurrence of Ascomycete fungi inside the domatia of African ant–plants when the plants are associated with mutualistic ant species. These results suggest that fungi playing no such structural role might be widespread associates of ant–plant mutualisms. Defossez et al. (2009) suggested that this additional partner could lead to radical shifts in our understanding of ant–myrmecophyte mutualisms. However, so far, no study has demonstrated the trophic implication of a fungus in an ant–plant mutualism. We address this issue by demonstrating that the fungus manipulated by A. decemarticulatus can be considered a trophic mediator in the H. physophoraA. decemarticulatus interaction. First, we investigate whether, and to what extent, H. physophora acquires ant-supplied nitrogen. Then, through the use of experimental enrichment with 15N isotopes, we examine the role of the fungus in facilitating plant nutrition. Finally, we identify where the nutrients transferred by the fungus are absorbed by the plant tissue.

Materials and methods

Study site and model

This study was conducted in 2008 and 2009 in the pristine forest situated around the field station at Petit Saut, Sinnamary, French Guiana (05°03′30.0″N; 52°58′34.6″W). Hirtella physophora is restricted to the understory of pristine Amazonian forests. These treelets have long-lived leaves that bear a pair of leaf-domatia at the base of each lamina. The domatia differ both morphologically and anatomically from the lamina (Leroy et al. 2008). In the study area, H. physophora is strictly associated with the plant–ant A. decemarticulatus (Myrmicinae).

Morphological isolation and molecular identification of the fungus

The occurrence of a specific fungus or of a variety of fungi in the galleries built by the ants was assessed through the determination of the morphological types of the hyphae on the different parts of the galleries. Then, samples from areas of the trap under construction were collected from 26 plant individuals from one population for molecular identification. We singled out this method to maximize the successful isolation of the mutualistic fungi and to minimize the risk of taking opportunistic fungi into account. Pieces of traps were immersed into sterile water and thoroughly washed to rid them of as many exogenous and highly prevalent spores as possible. Then, we pre-cultured these small pieces of loose mycelium by placing them in wet cotton where fast-growing contaminants were easily detected. We removed the new, expanding hyphae with forceps, and then washed them in a fresh drop of sterile water. Finally, between 15 and 30 single hyphae were collected and cultured in solid YMG medium (4 g yeast extract, 10 g malt extract, 4 g glucose, and 15 g of agar in 1 L of distilled water) including a cocktail of five antibiotics (100 mg L−1 Na–ampicillin, 120 mg L−1 SO4–streptomycin, 15 mg L−1 tetracycline, 30 mg L−1 chloramphenicol, and 30 mg L−1 SO4–kanamycin) at 25 °C for 6–20 days.

Total DNA from these pure cultures was extracted using a 10% Chelex (BioRad, Marnes-la-Coquette, France) solution. We amplified the entire internal transcribed spacer, including ITS1, 5S and ITS2 regions, for all of the samples with primers ITS1-F (Gardes & Bruns 1993) and ITS4 (White et al. 1990). The products were sequenced with the same primers by Genoscreen (Lille, France), and they ranged between 558 and 579 nucleotides.

Over-feeding the ants

We investigated the role of ants in the nitrogen provisioning of their host plant by providing surplus prey to the ant colonies. The prey, mostly phytophagous insects, were captured through the use of a light trap, and then provided to the ants by holding them close to the galleries where they were immediately seized and captured as demonstrated in Dejean et al. (2005). A total of 32 A. decemarticulatus colonies were thus supplied twice a week during 8 months and 42 other H. physophora individuals of a similar size were chosen as control plants. At the end of the experiment, a 4 cm2 piece was harvested from the youngest leaf of each host plant to estimate the δ15N values.

Removal of the fungal and ant partners

A total of 80 plant individuals were randomly assigned into five treatment groups (16 plants per treatment): (i) negative control, i.e. natural abundance of 15N; (ii) positive control, i.e. provisioning the ants with a 15N-enriched solution (see below) without manipulating the system; (iii) insecticidal treatment, i.e. ant removal and provisioning the fungus with a 15N-enriched solution; (iv) fungicidal treatment, i.e. fungus removal and provisioning the ants with a 15N-enriched solution and (v) removal of both the ants and the fungus and provisioning the plant with a 15N-enriched solution injected directly into the domatia.

The ants and/or the fungus were eliminated from the plants one week before the beginning of the experiment. The ants were eliminated by injecting an aqueous solution of 0.5‰ Pyrethrum inside all of the domatia. The rapid knockdown effect of the Pyrethrum immediately killed all of the resident ants, but had no effect on either the plant or the fungal galleries. The fungus was eliminated by spraying an aqueous solution of thiabendazole [2-(1,3-thiazol-4-yl)benzimidazole] at 1 g L−1 twice with a 2-day interval between treatments. Thiabendazole is widely used as a post-harvest fungicide on fruits before packing and transportation. No effect on the plant or the ant colony was observed after the treatment. The isolation trials of the fungus 1 week after the fungicidal treatment did not succeed, demonstrating its efficiency. Nevertheless, the galleries were still present even after the death of the fungus as the cell walls of the hyphae remained in place, at least over the course of the experiment.

Depending on the treatment, either the ants, the fungus or the plant were provisioned every day during nine consecutive days (see the details of the five treatments above). Two days after the final time the plants were supplied with the 15N-enriched solution, the two youngest leaves were harvested from each plant (including the control plants). For each leaf, a 4-cm2 piece of the lamina was dissected and pooled into a single sample per individual to estimate the δ15N values (and δ13C for the control plants). Then, the ants sheltering in the domatia of these leaves and/or the fungus present in the section of the gallery connecting the two leaves were collected.

Location on Hirtella physophora where 15N is absorbed

Fifteen H. physophora individuals having at least one branch with eight leaves were selected, and the ants inhabiting each branch were eliminated in the same way as described above. To avoid the recolonization of the branch by the rest of the resident ant colony, a ring of sticky Tanglefoot® (The Tanglefoot Co., Grand Rapids, MI, USA) was placed at the base of the branch. Then, the fungus present in the trap between the fourth and the fifth leaves was provisioned with a 15N-enriched solution during nine consecutive days as described below. Two days after the final time the plants were supplied with the 15N-enriched solution, the oldest and the youngest leaves of the branch were harvested from each plant by separating the domatia from the lamina. The inner part of each domatium was cleaned and all of the dead ants and eggs were removed.

15N provisioning

15N was provided to the ants, plants and fungus through 20 μL of a solution containing 2 mg ammonium nitrate (NH415NO3, 10 atom%15N, Isotec) and 2 mg of urea (H215NCO15NH2, 10 atom%15N, Isotec). For both the ants and the fungus, the substances were provided in an aqueous honey solution 50% w/v, while only water was used to provision the plants. The solution was provided to the ants on a small, plastic disc pinned close to a domatium entrance to avoid any contact with the plant tissues. To provision the fungus, the solution was carefully deposited with a micropipette on the gallery situated under the stems so that it would not reach the plant surface. To feed the plant, the solution was injected into a domatium using a micropipette. All of the feeding sites were the same for each individual over the course of the experiment and they were always located at least six leaves away from the youngest leaf on the branch chosen.

Isotopic analysis

All of the samples collected were vacuum-dried and ground into a homogeneous powder using a mixer mill. Around 170 mg of ant and 1 g of plant or fungal samples were analysed for their δ15N and δ13C content. Stable isotope analyses were conducted at the Stable Isotopes in Nature Laboratory (University of New Brunswick, Canada) using a Finnigan DeltaPlus gas isotope-ratio mass spectrometer (ThermoFisher Scientific, Waltham, MA, USA) interfaced with a Carlo Erba NC2500 elemental analyser or at the Colorado Plateau Stable Isotope Laboratory (Northern Arizona University, USA) using a Thermo-Finnigan DeltaPlus Advantage gas isotope-ratio mass spectrometer (ThermoFisher Scientific) interfaced with a Costech Analytical ECS4010 elemental analyzer. The natural abundances of 15N and 13C were calculated as follows:


where X is the element of interest, and Rsample and Rstandard are the molar ratios (i.e. 13C/12C or 15N/14N) of the sample and the standard, respectively (DeNiro & Epstein 1978a).

Morphological and anatomical analyses

Samples of old and young stems and domatia were collected from different plant individuals and then fixed in paraformaldehyde to study whether or not the fungal mycelium interacts with the plant tissues.

For the analyses using a scanning electron microscope (SEM), the dehydrated materials were critical-point-dried with liquid CO2. The dried materials were attached with double-sided tape to metal stubs, grounded with conductive silver paint, and sputter coated with goldpalladium. Observations were conducted using a SEM (Hitachi C450, Tokyo, Japan) operated at 15 kV. Photographs of the specimens were taken using Ilford 125 ISO film (Essex, UK).

Cross-sections of stems and domatia, 60-μm thick, were obtained using a vibrating microtome (Leica VT 1000S, Mannheim, Germany). As the hyphal cell wall could not be stained using WGA-FITC labelling, we used the ELF®97 Endogenous Phosphatase Detection Kit (E6601, Molecular Probes, Eugene, OR, USA) to detect the endogenous phosphatase activity of the fungus. We followed the protocol developed by Cox & Singer (1999). The ELF 97 has a maximum excitation at 345 nm and a maximum emission at 530 nm. Images of epifluorescence microscopy were acquired using an inverted microscope (Leica DMIRBE, Rueil-Malmaison, France) equipped with a typical DAPI longpass filter set and a CCD camera (Color Coolview; Photonic Science, Robertsbridge, UK). Confocal images were taken using a laser scanning confocal microscope (Leica LSM-SP2, Mannheim, Germany) with a 40 × water immersion lens (NA 0.75). Emitted fluorescence was collected in the range 520–550 nm, taking advantage of the spectral detection of the scanning head, using the 405 nm ray line of the diode laser. Images were computed from a maximum projection of 20–30 plan-confocal images acquired in z dimension.


Identification of the fungus

Only two morphological types of hyphae were consistently found throughout the different parts of the galleries, including the entrances to the domatia. Type 1 is dark, elongated hyphae with short rectangular articles, 6–9 μm in diameter. Type 2 is similar to type 1 but less melanized and slightly thinner at 4.5–6.5 μm in diameter. Cultures and molecular identification showed that these two types represent ontogenetic stages of the same fungal species. This species is the only one that has been found in all of the samples analysed so far and is the one growing in the galleries. In addition, a third type of hyphae is rarely found in the older parts of the trap, and it has strongly melanized and non-branching hyphae 9.7–17.2 μm in diameter terminating in globose conidia with ornamented walls.

A BLAST search and the phylogenetic analyses of the ITS region revealed that a clade with four monophyletic taxa was the closest (85–91% identity) to our sequences of hyphal types 1 and 2. Three out of the four taxa are uncultured, environmental samples (GenBank accessions: FJ820738, AJ582964, AY969659); the fourth is an Ascomycete species from the order Chaetothyriales described as Trimmatostroma cordae N.D. Sharma & S.R. Singh [as ‘cordaicis’] (GenBank accession: AJ244263) (Sharma & Singh 1976; de Hoog et al. 1999). It should be noted that we have not, however, yet been able to demonstrate that this particular fungus sequenced from the galleries is the same taxon as the one found inside the domatia.

Natural abundance of stable isotopes

The natural abundance of stable isotopes, shown in Fig. 1, allows the three partners to be distinguished on the basis of their different levels of δ15N and δ13C. Generally, 15N and 13C have distinct patterns within food webs with higher trophic levels being commonly enriched in 15N, but less so in 13C (DeNiro & Epstein 1978b; Peterson & Fry 1987). Thus, δ15N indicates the trophic position with a 15N enrichment of c. 3‰ per trophic level, whereas δ13C indicates the type of diet as it is little altered (c. 1‰) with an increase in trophic level.

Figure 1.

 Natural abundance of δ15N and δ13C within the tripartite ant–plant–fungus mutualism. Mean δ15N (‰) vs.δ13C (‰) values for Allomerus decemarticulatus, Hirtella physophora and the fungus present in the galleries. Error bars represent ± 1 SE.

The ant partner, A. decemarticulatus, had the highest value of δ15N (mean ± SD; + 4.76 ± 0.32‰). This value is significantly different from the values for the plant (+ 2.66 ± 0.94‰) as well as for the fungus in the trap (+ 0.99 ± 1.01‰) (ants vs. plant: Mann–Whitney U test, U = 3.26, P = 0.001 and ants vs. fungus: U = 3.81, P < 0.0005). The natural level of δ15N for A. decemarticulatus at c. 5‰ also reflects its omnivorous diet that includes both prey and the extrafloral nectar of its host plant (Davidson et al. 2003).

Moreover, H. physophora exhibited a significantly higher amount of the nitrogen isotope than did the fungus (U = −2.71, P = 0.0067). On the contrary, the fungus displayed the highest value of δ13C (−31.12 ± 0.35‰), which differed significantly from that of the plant (δ13C = −34.71 ± 0.35‰, U = 4.50, P < 0.0005), but not the ants (δ13C = −31.51 ± 0.40‰, U = −1.37, P = 0.168); this indicates that ant and fungus may be obtaining carbon from similar sources. The C isotopic shift of c. 3.5‰ observed in the fungus might be related to fractionation during the biochemical process involved in the transfer of C from the plant detritus to the fungus (Kohzu et al. 1999). The amount of carbon isotope recorded in the ants (−31.51 ± 0.40‰) was also significantly higher than in the plant (−34.71 ± 0.35‰; U = 4.50, P < 0.0005). The depletion of 13C in the H. physophora leaves compared with others C3 plant species is probably related to: (i) the low level of available light in the understorey (as intercellular CO2 concentrations in leaves increase with decreasing light availability) (Ehleringer & Werk 1986), and (ii) a lower proportion of 13C in the CO2 released through soil respiration (Schleser & Jayasekera 1985).

Ants’ influence on the natural abundance of 15N in their host plant

The over-feeding of the ant colonies translated into a significant increase in the δ15N in their host plants compared with the control plants (one-way anova, F1,72 = 22.74, P < 0.0005; Fig. 2). In the latter case, the ant colonies fed only on the prey that they captured themselves. Such an increase (c. 1‰) in δ15N (mean ± SD; 2.47 ± 0.99‰ vs. 1.50 ± 0.71‰) in plants whose ant colonies were over-provisioned compared with the control plants is due to the absorption of nitrogen from the prey and highlights the ability of H. physophora to absorb nutrients from ant waste via its above-ground organs.

Figure 2.

 Effect on the δ15N content in the leaf of over-feeding the ants. δ15N values (‰) for the young leaf on Hirtella physophora plants for which the resident Allomerus decematiculatus colonies were over-provisioned (experimental plants) or not (control plants). Error bars above and below the boxes indicate the 90th and 10th percentiles, the ends of the boxes indicate the 25th and 75th percentiles and solid circles indicate outliers.

Experimental investigation of nitrogen fluxes through the addition of a 15N tracer

Provisioning the ants with the 15N-enriched solution (positive control) resulted in a significant increase in δ15N in all of the associated partners (Mann–Whitney U test, Uants = −3.97, Pants < 0.0005; Ufungi = −4.28, Pfungi < 0.0005; Ulamina = −3.41, Plamina < 0.0005; Fig. 3). The removal of the fungus also resulted in a significant increase in δ15N in the lamina compared with the negative control group (U = −3.69, P < 0.0005), and the level of enrichment was of the same magnitude as in the positive control group (U = −0.79, P = 0.428). The latter demonstrates that the nitrogen uptake by the plant can come directly from ant faeces and/or ant-deposited wastes.

Figure 3.

 Effect of 15N provisioning after the removal of the fungus, ants, or both partners. δ15N values (log ‰) for the three partners (Allomerus decemarticulatus ants, Hirtella physophora leaves and the fungus) for each of the five treatment groups. Either the ants, the fungus in the galleries or the plant were provisioned every day during nine consecutive days with an aqueous solution of 15N-urea and 15N-ammonium nitrate. Error bars above and below the boxes indicate the 90th and 10th percentiles, and the ends of the boxes indicate the 25th and 75th percentiles and solid circles indicate outliers.

Surprisingly, the removal of the ants and the provisioning of the fungus led to a significant increase of several orders of magnitude in the δ15N in the plant tissue (873.95 ± 179.45‰) compared with the positive control group (δ15N = 102.31 ± 120.13‰, U = −3.99, P < 0.0005) and the fungicidal treatment (δ15N = 113.52 ± 94.60‰, U = −4.25, P < 0.0005). A non-significant increase in 15N was observed when the 15N-enriched solution was injected directly inside the domatia on plants from which both the ants and the fungus had been removed compared with the insecticide treatment (δ15N = 742.41 ± 230.54‰ vs. 873.95 ± 179.45‰, respectively, U = −1.24, P = 0.213).

The amount of nitrogen isotope in the domatia and the lamina, after provisioning the fungus, differed according to their ages (Fig. 4). Old domatia, bearing fungus, had higher values of δ15N than old lamina (δ15N = 423.51 ± 291.18‰ vs. 93.35 ± 41.91‰, respectively, Wilcoxon paired test = 2.49, P = 0.0124), while for the young leaves, the lamina had higher δ15N values than the domatia (devoid of fungus) (δ15N = 419.69 ± 100.80‰ vs. 168.68 ± 29.81‰, respectively, Wilcoxon paired test = 3.41, P = 0.0006). There were no statistical differences between the old domatia and either the lamina or the domatia of young leaves (old domatia vs. young lamina: Mann–Whitney U test, U = −1.64, P = 0.101 and old domatia vs. young domatia: U = −0.228, P = 0.819).

Figure 4.

 Location on Hirtella physophora where 15N is absorbed. δ15N values (‰) for lamina and domatia taken separately from young and old Hirtella physophora leaves. The fungus in the galleries was provisioned every day during nine consecutive days with an aqueous solution of 15N-urea and 15N-ammonium nitrate. Error bars above and below the boxes represent the 90th and 10th percentiles, the ends of the boxes indicate the 25th and 75th percentiles, and solid circles indicate outliers.

Site of the nutrient transfer

The galleries built by A. decemarticulatus, and on which the fungal mycelium grows, run along most of the stems of the host plant. A histological investigation of the stems did not, however, reveal any interaction between the fungal hyphae and the plant tissue (Fig. 5a,b). In contrast, microscopical observations of old domatia showed that their inner surfaces were covered with hyphae (Fig. 5c–e). Moreover, the ants scarify the inner epidermis of the domatia so that broken cells cover most of that surface. Clusters of hyphae were observed in close contact with these cells and anatomical observations using ELF 97 staining revealed the presence of hyphae inside intact cells in the mesophyll of the domatia, adjacent to broken cells (Fig. 5f,g). Fungal hyphae, however, were not observed inside young domatia, even if their inner epidermis was already scarified by the ants.

Figure 5.

 Morphological and anatomical study of the plant–fungus interaction. (a) Portion of a Hirtella physophora stem showing the trap built by the Allomerus decemarticulatus ants using the plant trichomes and the fungus. (b) Cross-section of the stem showing intact tissue and fungal hyphae stained with ELF 97 and observed through an epifluorescence microscope (yellow fluorescence), around the trichomes on the stem. (c,d) scanning electron micrograph of the inside of old domatia with running hyphae (c), and broken cells with clusters of hyphae (d). (e–g) Confocal micrograph of cross-sections of old domatia showing clusters of hyphae (arrows) inside the domatia; (e) transmitted light and (f,g) confocal image showing ELF 97-stained hyphae inside some scarified epidermal cells, but also in some of the cells of the mesophyll.


Together our results highlight the key role of a third, ‘hidden’ partner that can benefit the plant, revealing the existence of a true tripartite mutualism in this system. Allomerus decemarticulatus has developed a new type of highly specific interaction with a fungus, which has primarily a structural role (Dejean et al. 2005), but here we demonstrate that it is also involved in transferring nutrients to the plant.

The 15N provisioning of the fungus in the galleries resulted in sharp increases in the 15N signature of the plant’s leaves, and such results prove for the first time that there is an indirect transfer of nitrogen mediated by a fungus in an ant–plant mutualism. Moreover, a high level of enrichment was observed when the plants were fed in the domatia, thus highlighting the possibility that nutrients are absorbed through the domatia walls. Such nutrient acquisition from ant faeces or waste accumulated in the domatia has already been demonstrated for other ant–plant interactions (Fischer et al. 2003; Solano & Dejean 2004; Watkins, Cardelus & Mack 2008).

The total 15N uptake by the plant was much greater after the resident ants were removed, whether the plant was provisioned directly or through the fungus. On the contrary, this shows the importance of the fungus in the transfer of nitrogen to the plant. That the same level of 15N enrichment was observed in the plant’s leaves when the fungus or the plants themselves were fed leads us to believe that this fungus retains very little nitrogen and transfers the majority of it to the plant. Under natural conditions, A. decemarticulatus integrates the remains of prey into the galleries built from plant material (i.e. the trichomes and plant pellets) (Solano & Dejean 2004). Such remains are the only external source of nitrogen for the plant, aside from the nutrients absorbed by the roots. Consequently, the absorption of these remains by the fungus can also benefit the plant. On the contrary, the higher level of 15N uptake in the absence of ants reveals the role that the ants play in filtering the nitrogen available to the plant. Most of the nitrogen provided to the ants is kept for their own nutritional requirements, and only a small fraction will reach the plant and/or the fungus through their faeces. This can be easily explained, as nitrogen is one of the principal factors limiting the growth and reproduction of ant colonies in an arboreal environment (Davidson 1997; Davidson et al. 2003). Worker ants require a mainly carbohydrate-rich diet, whereas larvae depend on nitrogen-rich sources for growth (Dussutour & Simpson 2009).

Based on our results and observations, it appears that nitrogen is mainly transferred from the fungus in the galleries to the plant in old domatia. The level of 15N was significantly higher in the old domatia than in the corresponding lamina, and it was similar to the levels found in the young and developing leaves (which are the main nitrogen sink in a plant). A thorough investigation of the morphological structure of the interior of old domatia revealed that fungal hyphae were numerous running along the epidermis as well as in areas with broken cells and further inside the mesophyll. On the contrary, the stem tissues were always intact. The abundant hyphae in close contact with the domatia cells increase the area of plant surfaces with a potentially nutrient-absorptive role, thereby enhancing the transfer of nutrients (at least of nitrogen) to the plant. We cannot, however, assert that the hyphae present inside the domatia are from the same taxon as the one found in the galleries. Consequently, further studies are needed to better understand the way nutrients are transferred to the plant tissues.

The presence of fungi inside domatia has been reported in several myrmecophytic species, suggesting that it could be a common and widespread phenomenon that has been overlooked (Blatrix et al. 2009; Defossez et al. 2009). Such trophic interactions could also be viewed as a striking convergence with mycorrhizal symbiosis, in which plant roots and fungi exchange nutrients (Smith & Read 2008); in our study, however, nutrient transfer occurs in the leaves and through interactions with an ant species. Nonetheless, further research on both the cellular interactions between the plant and the fungus and on the precise characterization of the transfers using the SIMS imaging technique are needed. Moreover, the potential reciprocal benefits for the fungus still need to be both qualified and quantified to better understand the evolution of associations in which multiple partners are involved.

In conclusion, we demonstrate here that fungal mediation can improve nutrient uptake by the plant, suggesting, as outlined by Defossez et al. (2009), the need for a reinterpretation of the mutualistic outcomes of such associations. The commonness and the implication of mutualistic micro-organisms in myrmecophytic systems can indeed induce a shift in the interpretation of ant–plant symbioses from purely protective mutualisms to more nutritionally oriented ones. Several studies have already provided evidence of the artificial nature of the dichotomy between protective and nutritional mutualisms, as myrmecotrophy is widespread in myrmecophytic plants (Beattie 1989; Benzing 1991; Sagers, Ginger & Evans 2000; Fischer et al. 2003; Heil & McKey 2003). Nevertheless, such interactions are not purely nutritionally oriented and thus the beneficial outcomes might be better understood within the framework of mutualisms rather than food webs. Nevertheless, trophic mediation by fungi adds to our understanding of the processes whereby the flow of nutrients occurs between organisms. Allomerus decemarticulatus, for example, is known to partially castrate its host plant (Orivel et al. 2010), but this cost is balanced by the protection the ants also provide. The nutritional benefit provided by the fungus could thus positively affect the evolutionary persistence of the ant–plant interaction, especially as this fungus is strictly associated with the presence of the ants.


We would like to thank the Laboratoire Environnement de Petit Saut for furnishing logistical help. We are grateful to Veronika Mayer as well as anonymous reviewers for helpful comments on an earlier version of the manuscript, and also to Andrea Dejean for proofreading the manuscript. Financial support for this study was provided through a research programme of the French Agence Nationale de la Recherche (research agreement n°ANR-06-JCJC-0109-01), by the ESF-EUROCORES/TECT/BIOCONTRACT program, by the Programme Amazonie II (project 2ID) of the French Centre National de la Recherche Scientifique, by the Programme Convergence 2007-2013 Région Guyane (project DEGA) from the European Community, and by a research grant from the IFR 40 (Toulouse).