Species and study site

Azteca brevis Forel, 1899 (Formicidae, Dolichoderinae) is a reddish-brown ant, c. 4 mm long, known from wet forests of the southern Pacific lowlands of Costa Rica [19]. Colonies have been found on Tetrathylacium macrophyllum (Salicaceae), Licania sp. (Chrysobalanaceae), Grias sp. (Lecythidaceae), Myriocarpa sp. (Urticaceae), Ocotea nicaraguensis (Lauraceae) [19] and Lonchocarpus sp. (Fabaceae). The nesting chambers inside the stems are connected externally by runway galleries dotted with small, circular holes (Figure 1A, B).

The most common host plant for Azteca brevis is Tetrathylacium macrophyllum Poepp. (Salicaceae), a small tree (c. 8 m) that grows on the Pacific slopes of Central and South America in areas characterised by high annual rainfall (>5000 mm). It is found chiefly on steep slopes near rivers and streams in primary forest [20]. About 30% of T. macrophyllum trees are occupied by Azteca brevis [21], [22]. The ants start by colonising hollow chambers in the branches that the plant forms through pith degeneration. As the colony grows, the ants excavate the remaining pith between adjacent naturally formed chambers, and build large nest sites inside the branches.

We sampled within a 5-km circle around the Tropical Research Station La Gamba, Costa Rica (8° 42′ 03″ N, 83° 12′ 06″ W) along the Waterfall Trail, Bird Trail, Río Gamba, Río Bolsa and Río Sardinal. Carton samples, ants and plant parts for herbarium specimens were collected from 18 T. macrophyllum trees and one Lonchocarpus tree colonised by Azteca brevis, under permission from SINAC – Sistema Nacional de Areas de Conservación de Costa Rica of the Ministry of Environment and Energy (MINAE) to M.N. and V.E.M. (No. 182-2010-SINAC). In recent years, trees colonised by Azteca brevis have become inexplicably rare at the study site (VEM, pers. obs.), limiting the sample size to 19 trees colonised by Azteca brevis. At least three carton pieces per tree and colony, each c. 1 cm long, were taken from runway galleries and stored in 1.5-mL reaction tubes sealed with air-permeable cotton wool. The reaction tubes were kept in a sealed plastic bag with silica gel for three weeks before culturing the fungi; the bags were transported and stored at room temperature.

Fungal cultures and DNA-extraction

At the University of Vienna, a c. 10-mm² piece of carton was placed into a droplet (c. 20 µL) of sterile water and fragmented with sterile forceps to make a mycelial suspension. An aliquot of mycelial suspension was then diluted with 1 mL sterile water and spread over each of two 2% malt extract agar plates (MEA) containing 0.5% penicillin and 0.5% streptomycin. This was carried out on average for three samples per colonised tree. The plates were stored at room temperature and visually checked under a dissecting microscope at least once a day for 11 days. Fast-growing “weeds” (Aspergillus, Cladosporium, Fusarium) were excised to prevent overgrowth of the slower-growing carton fungi. The thick, darkly pigmented hyphae that are typical for the carton usually started to grow after 2–4 days and were then transferred to new 2% malt extract agar plates. Several carton samples from each ant colony and tree were processed in this way to minimise any cultivation bias.

Sections of approximately 25 mm² were cut out from mycelia on pure-culture agar plates and stored in 2-mL reaction tubes at −20°C. The frozen samples were subsequently freeze-dried overnight and ground with five glass beads (3 mm diameter) for 10 min at 30 Hz in an MM 400 mixer mill (Retsch, Germany), after which DNA was extracted using NucleoSpin 96 Plant II kits (Macherey-Nagel, Düren, Germany).

PCR and cleanup

A 1.5–3.5-kb nuclear ribosomal DNA (rDNA) fragment comprising partial small subunit (SSU), complete ITS1–5.8S–ITS2 (ITS) and partial long subunit (LSU) sequences was amplified with the fungal primers V9G [23] and LR5 [24] using Thermo Scientific 2.0× ReddyMix Extensor PCR Master Mix and 1.1× ReddyMix PCR Master Mix (ABgene, Epsom, UK) (for primer sequences and detailed PCR protocol, see Table S1 in Appendix S1). The PCR products were purified with 6 U exonuclease I and 0.6 U FastAP thermosensitive alkaline phosphatase (Fermentas, St. Leon-Rot, Germany) [25]; the PCR product was then incubated for 30 min at 37°C, followed by enzyme deactivation for 15 min at 85°C.

Sequencing

DNA was cycle-sequenced with ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction v. 3.1 (Applied Biosystems, Warrington, UK) using the PCR primers and primers LR3 [24] and ITS4 [26]. For sequences with large indels, the additional primers LR2R-A, LR2-A [27], F5.8Sr, F5.8Sf [28] and LR3-CH (5′-GGT ATA GGG GCG AAA GAC TAA TC-3′) were necessary to obtain full-length sequences (see Appendix S1 for detailed sequencing protocol). Sequencing was performed on an ABI 3730xl Genetic Analyzer automated DNA sequencer (Applied Biosystems). One sequence of each genotype was deposited in GenBank. The complete list of accession numbers for the SSU–ITS–LSU locus can be found in Table S2 in Appendix S1.

Analysis of sequence data

After a BLAST search (Basic Local Alignment Search Tool in GenBank, http://blast.ncbi.nlm.nih.gov/Blast.cgi) of the nuITS1–5.8S–ITS2–LSUr DNA sequences obtained from the ant carton (423 in total), 381 sequences were identified as Chaetothyriales and used for further analyses. For phylogenetic analyses, identical sequences from carton samples were reduced to a single sequence per genotype. Cases where two sequences differed only in homopolymer regions were also merged to a single genotype.

For alignment, the chaetothyrialean sequences used in Voglmayr et al. [7] and the closest sequences to our isolates from GenBank were added. Verrucaria denudata, V. csernaensis and V. andesiatica (Verrucariales) were included as outgroups. Ambiguously aligned regions in ITS1/ITS2 and leading gap regions were excluded. The matrix of 258 sequences contained 7767 alignment positions, with the longest sequence comprising 3425 nucleotides. Alignments were produced with Muscle 3.8.31 [29] and revised in BioEdit 7.1.3.0 [30]. (GenBank accession numbers of the sequences included in the phylogenetic analyses are listed in Table S2 in Appendix S1.)

For Bayesian analyses, MrBayes 3.2.1 [31] was run through the Bioportal web service of the University of Oslo [32]. The six-parameter general time-reversible substitution model was used, with a proportion of invariant sites and a gamma distribution for the remaining sites (GTR + I + G), as determined by Modeltest 3.7 [33]. Three parallel runs of four chains were performed over 30 million generations, sampling 30 000 trees in each run. For 90% majority-rule consensus trees, the first 2000 trees of each run were discarded as burn-in.

Maximum-parsimony (MP) bootstrap analyses were performed with PAUP* 4.0b10 [34], using 1000 replicates of a heuristic search with 10 rounds of random sequence addition during each bootstrap replicate and a limit of 100 000 rearrangements per replicate. TBR branch swapping was used, allowing multitrees, and steepest descent was set to ‘no’. Gaps were treated as missing data, and no weighting of nucleotides was applied.

The maximum-likelihood (ML) analyses used RAxML [35], as implemented in the programme raxmlGUI 0.95 [36]. The GTRGAMMAI nucleotide substitution model was applied for the ML heuristic search and the ML rapid bootstrap analysis.

Fungal distribution at the genotype and species levels

To evaluate the specificity of the fungi to Azteca brevis, the frequency of occurrence on the sampled trees was analyzed. For each tree, a matrix containing genotypes and sampled trees was compiled (Table S3 in Appendix S1) and Bray–Curtis similarity indices among sampled trees were plotted by non-metric multidimensional scaling (NMDS) with 1000 restarts. In addition, an ANOVA of similarity (ANOSIM) was conducted with up to 999 permutations, based on different geographical study sites (Table S2 in Appendix S1). All analyses were carried out with Primer 5 5.2.9 (PRIMER-E, 2002).

Slightly different genotypes can, however, represent the same species: the maximum number of mutations between two individuals of the same species (mutation limit) varies with the DNA region and type of organism. For fungi, the ITS region is more variable than SSU or LSU [37]. Because ITS is the main DNA fragment in the present study, the mutation limit was determined by multiplying the maximum intraspecific variation of ITS (0.58%) [37] by the average sequence length. We define an OTU as a species with a maximum genotype variation of 12.02 mutations. The abundance of OTUs was also analyzed using a modified presence–absence matrix (Table S4 in Appendix S1).

Light microscopy and SEM analysis of carton material and fungal hyphae

Pieces of carton from 16 trees and aerial hyphae of the pure cultures were investigated using a Zeiss AxioImager A1 compound microscope with a Zeiss AxioCam ICc3 digital camera. SEM investigations were made with a Jeol JSM-T 300 scanning electron microscope (SEM) at 10 kV.

Cultures

Pure cultures of carton fungi were obtained from carton material of host trees colonised by Azteca brevis. Because different species could not be distinguished morphologically, all the pigmented hyphae that germinated were transferred to MEA plates to obtain pure cultures. In total, 423 pure cultures were sequenced, resulting in 128 different genotypes after identical sequences were removed.

Molecular phylogenetic analyses

After adding relevant sequences from GenBank, the final alignment, including outgroups, consisted of 258 sequences and 7767 alignment positions. In the best tree from the ML search (shown as a phylogram in Figure 2–3), backbone support is mostly low or absent, but most of the subclades are well-supported. Consensus trees across the three analyses differed topologically only in one point: one clade of five sequences containing a large indel of c. 1000 bp within the LSU region is located more basally in the topology resulting from the MP analysis than in the topologies from BA and ML analyses (marked with an asterisk and an arrow in Figure 3). Bootstrap support for this node in the MP analysis was only 73%, so this difference was not considered further.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Phylogram of Chaetothyriales, top part.

The maximum-likelihood tree is shown, based on partial SSU, complete ITS and 5.8S, and partial LSU rDNA regions. Bold branches are supported in all three analyses: BA probabilities higher than 0.9, ML and MP bootstrap support above 70%. Red labels denote fungal genotypes isolated in this study from ant-built carton structures on Tetrathylacium macrophyllum and Lonchocarpus sp. trees; orange and blue mark domatia fungi and carton fungi, respectively, from Voglmayr et al. [7]; GenBank accession numbers follow taxon names; solid red, violet and orange vertical lines indicate clade definitions and captions from Voglmayr et al. [7]; dotted lines mark clade extensions from this study. Blue dots point out three sequences from other continents (2× Cameroon; 1× Thailand) differing by only three mutations from our Costa Rican genotypes. Note the high diversity of isolated genotypes (the large clade extensions compared to Voglmayr et al. [7] are due to a greater number of samples and the new monophyletic carton-fungi cluster (new carton clade 3).The tree is continued in Figure 3.

https://doi.org/10.1371/journal.pone.0112756.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Phylogram of Chaetothyriales, bottom part.

Continuation of Figure 2 with arrowheads indicating the connection. For label and colour descriptions see legend to Figure 2. The clade labelled with an asterisk (*) is placed more basally (arrow) in the MP analysis than in ML and BA analyses. Note the domatia-symbiont clade, which remains distinct from carton fungi.

https://doi.org/10.1371/journal.pone.0112756.g003

The phylogenetic reconstruction revealed three main clades containing 73% (100 out of 128) of the isolated genotypes (Figure 2–3). Carton clades 1 and 2 were described by Voglmayr et al. [7], but carton clade 3, which was represented by a single sequence in previous analyses, is new. The remaining 28 genotypes are distributed across the phylogenetic tree of Chaetothyriales, except for the “domatia-symbiont clade”. Three sequences from carton material from Cameroon and Thailand are nearly identical to some fungi sequenced in this study. Each of those three fungal genotypes (marked by blue dots in Figure 2) differs by only three mutations from fungi grown from carton structures collected in Costa Rica.

Fungal distribution at the genotype and species levels

The 128 different genotypes isolated from carton material of 19 Azteca-inhabited trees were analyzed with a presence–absence matrix (Table S3 in Appendix S1). The matrix showed that no genotype was found on all trees, with the most common one isolated from nine out of 19 trees. On average, 10.5 different genotypes occurred on each tree, of which 46% were unique to single trees. In the carton sample of one Lonchocarpus tree inhabited by Azteca brevis, 11 out of 12 genotypes (92%) were unique.

Bray–Curtis similarity indices between sampled trees were calculated to investigate the correlation between genotype composition and collection site. The non-metric multidimensional scaling (NMDS) plot showed no clustering of trees from the same collection site (Figure 4), and the analysis of similarity (ANOSIM) showed a significance level of only P = 0.32. A correlation between genotype composition and collection site can therefore be ruled out. One sampled tree (Tm10) had to be excluded because only a single, unique fungus could be isolated, and the Bray–Curtis distance to the other sampled trees was too great for Tm10 to be displayed without clustering all remaining trees too tightly together.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Correlation analysis of genotype sets and collection sites.

Non-metric multidimensional scaling (NMDS) plot based on Bray–Curtis similarities between fungal genotype sets occurring on carton material of 18 sampled Tetrathylacium macrophyllum trees, and one Lonchocarpus sp., colonised by Azteca brevis. Different symbols represent trees from different collection sites (squares: Waterfall Trail; triangles: Bird Trail; circle: Río Gamba; diamonds: Río Bolsa; inverted triangle: Río Sardinal).

https://doi.org/10.1371/journal.pone.0112756.g004

The results were similar at the approximated species level. The 128 genotypes were reduced to 62 OTUs, and the most common OTU in the modified presence–absence matrix, represented by nine genotypes, was found on 12 out of 19 trees (63%) (Table S4; Figure S1 in Appendix S1). Three other OTUs were found on a total of 9 out of 19 trees (47%). The correlation of fungal community and collection site at the species level is weak (NMDS plot, Figure S2 in Appendix S1) and not significant (ANOSIM: P = 0.16).

Light microscopy of carton material

Dark, melanised moniliform hyphae with thick cell walls (Figure 1F) appeared to be dominant in each sample. The cell width of these thick-walled hyphae ranged from about 6 to 9 µm. Hyaline hyphae with a cell width less than 5 µm were also present, but appeared less abundant (Figure 1E). Aerial hyphae of pure cultures representing the three carton clades were also examined and, surprisingly, darkly pigmented, thick-walled hyphae are largely restricted to carton clade 1, whereas thin-walled hyaline hyphae are found in carton clades 2 and 3 (Figure 2–3).

Molecular phylogenetic analyses and fungal diversity

There was no ubiquitous fungus (genotype or OTU) among the 128 genotypes found in this study; 78% of the sequenced carton fungi clustered into three clades. Two of those clades were already established [7], and we have discovered a third (Figure 2). The 28 genotypes that were not assigned to any of those clades were scattered across the whole phylogenetic tree (Figure 2–3), but none of the 128 fungal genotypes isolated from the carton samples arose in the “domatia-symbiont clade” [7]. The fungi belonging to this special clade are highly distinct from carton fungi in terms of their hyphal morphology and growth form (hyaline or less pigmented and tending to produce spores in domatia fungi) [7] probably due to their different functions. Domatia fungi are used as food for the larvae [8], whereas carton fungi probably improve the stability of carton walls in ant nests or runway galleries. The specificity and coevolutionary dynamics between domatia and carton symbionts may differ.

Fungus specificity at the genotype and species level

In most known insect–fungus symbioses (e.g. termites [43], [44] or leaf-cutter ants [45]), the associated fungus is cultivated for food, whereas Azteca brevis is not likely to eat the fungi, but cultivates them for their nest architecture. In this association, the ants as a single host use a group of multiple fungus species for this purpose. The interpretation of the degree of specificity depends on the taxonomic level and differs between the fungi involved. At higher taxonomic levels, the interaction specificity between Azteca brevis and Chaetothyriales is high: chaetothyrialean fungi were found in every carton sample analyzed. At the genotype level, a much more modest degree of interaction specificity was seen (Table S3 in Appendix S1): the most common genotype (T121_Tm12) was associated with 9 out of 19 ant colonies (c. 47%). No correlation between fungal community and collection site was found, and the fungal community seems not to be habitat-specific (Figure 4). Merging genotypes into OTUs based on the maximum intraspecific variation of ITS [37] increased the occurrence of the most common fungus (represented by T66 in Table S4 in Appendix S1) to 63% of the ant colonies and the three next most common OTUs to 47% (Table S4 in Appendix S1). Surprisingly, a mean of eight different OTUs were found per carton sample, but no specialist fungal partner obligate to all Azteca brevis/Tetrathylacium macrophyllum associations was found. Environmental samples for the most common OTU could not be analyzed, because no primers could be developed which were specific enough for reliable separation of the OTUs, and only culturing was possible. It might be argued that culturing introduces a bias, in that a genotype could have been missed. Because we sampled the carton galleries of every colony and tree several times each, and also performed the isolation procedure on average three times per sample, it is unlikely that we missed any genotype. A heterogeneous spatial distribution of the genotypes along the branches can also be excluded, because we always took samples at positions from the branching point to the tip of a branch, and would therefore have included any heterogeneity.

Moreover, some fungal sequences from Azteca brevis carton are nearly identical at the SSU–ITS–LSU region with those of fungi from cartons of other ant species and from samples collected on other continents. Two sequences from Crematogaster carton nests from Cameroon (Crematogaster HQ634612 and Crematogaster HQ634621), and one found in Thailand (Crematogaster HQ634632) [7] differ by only three mutations from the Costa Rican sequences (blue dots in Figure 2). This indicates that at least some ant-associated chaetothyrialean fungi have a transcontinental distribution and are associated with more than one ant genus rather than being specific to Azteca brevis hosts.

Fungus selection and origin of the fungi

The carton runway galleries of Azteca brevis are used as a defence against intruders and as ambush traps for capturing prey (V. Mayer, pers. observ.), analogous to the carton galleries described for Allomerus decemarticulatus [13]. The fungal symbionts may play a particularly important functional role by increasing the stability of the carton produced by the ants [6], [13]. The hypothesis that fungi are used for reinforcement was first raised by Lagerheim for the carton constructions of Lasius fuliginosus [10] and it has been accepted by many authors [6], [11], [13], [46]. Trimming and grooming of the carton fungi has been observed and, due to the reinforcement demand, it may be expected that Azteca brevis would select fungal symbionts with a particular phenotype. Because we conducted a pure-culture method, we were able to examine the hyphae microscopically and correlate genotype or OTU with hyphal type. Two types of hyphae were regularly found: thin-walled, hyaline hyphae (Figure 1E), which represent two of our three major carton clades. The other type, thick-walled, melanised hyphae (Figure 1F), are less abundant in the species in our phylogenetic tree. When investigating pieces of carton, however, the thick-walled hyphae visibly dominated the biomass. This indicates that the second hyphae type is favoured on the carton. This may be either due to the substrate or due to the ants' preference. Fungi with thick-walled, melanised hyphae are likely to be better for the carton's stability than thin-walled hyphae.

Unfortunately, no alate queens were found, and interpretations of the transmission of the fungi must therefore be made without experimental evidence. The lack of any strong specificity indicates that the fungi growing on the Azteca brevis carton are either transmitted horizontally or environmentally acquired, although the degree of host–symbiont specificity is not always correlated with the transmission mode [47]. De novo acquisition of fungal symbionts from the environment in each ant generation would, however, explain the high number of genotypes and OTUs in the carton samples. Azteca brevis workers may collect spores or hyphal fragments from the environment, as termites do [48], but detailed field observations of the worker ants' behaviour are needed to prove this. Also, the plant material used for construction (typically bark and epiphylls) may already be infected with chaetothyrialean spores. The branches and stems of trees inhabited by Azteca brevis appear to be completely cleaned of epiphylls (algae, bryophytes and lichens) and epiphytes. In fact, the newly described Chaetothyriales family Trichomeriaceae was found to grow on the surface of living leaves [49]. Investigation of the host plants' surface (stem, branches and leaves) and the epiphylls are needed to clarify whether Chaetothyriales are already present. The chaetothyrialean fungus community of the carton galleries might be a subset of the fungal community found on the host plant's surface.

Two analogous systems: Allomerus sp. and Azteca brevis

A plant–ant–fungus association with runway galleries that are structurally similar to those in the present study is seen in Amazonian Allomerus ants living on Hirtella physophora (Chrysobalanaceae) and Cordia nodosa (Boraginaceae) [14]. In contrast to the Azteca brevis carton, with its guild of numerous species of Chaetothyriales, the Allomerus ants are described as cultivating one specific fungal symbiont on the carton galleries (see upper part of carton clade 1; Figure 3). Allomerus foundress queens apparently store a pellet with the specific carton fungus represented by a monophyletic group of haplotypes on the domatium wall. While building their runway galleries, Allomerus workers were observed to glue pellets from scraped epidermis and mesophyll of the inner domatia walls onto the gallery frame built from trichomes of the host plant [14]. In Azteca brevis colonies, no inoculation pellet was found, an observation that supports the hypothesis that fungal symbionts are acquired de novo from the environment.

Although the sampling methods were different (only newly produced carton material was collected from the Allomerus galleries, whereas it was mainly “mature” black material that was sampled from the Azteca brevis galleries), the frequency of the most common fungus or OTU was more or less the same: on 57% (138 out of 240) of the Allomerus galleries, and on 63% (12 out of 19) of the Azteca brevis galleries. A fungal species found in a little over half of the analyzed carton samples should not, however, be regarded as specific.