•The ectomycorrhizal (ECM) symbiosis was historically considered restricted to the temperate zones, but recent studies have shown the importance of this symbiosis across the tropics. We examined ECM fungal diversity, host plant phylogeny and ECM host preferences in a rainforest dominated by the leguminous host plants Dicymbe corymbosa, Dicymbe altsonii and Aldina insignis.
•Ectomycorrhizal fungi were identified by internal transcribed spacer rDNA sequencing and host species were verified with chloroplast trnL sequencing. To test whether Dicymbe and Aldina represent independent gains of the ECM symbiosis, we constructed a Fabaceae phylogeny using MatK and trnL. We identified four independent ECM lineages within the Fabaceae.
•We detected a diverse community of 118 ECM species dominated by the /clavulina, /russula-lactarius, /boletus, and /tomentella-thelephora lineages. Ectomycorrhizal species in Agaricales, Atheliales and Polyporales may represent previously unrecognized tropical-endemic ECM lineages. Previous studies suggested that ECM fungi did not diversify in the tropics, but the /clavulina lineage appears to have a center of diversity in tropical South America.
•Dicymbe and Aldina represent independent gains of the ECM symbiosis in Fabaceae but their fungal symbionts showed no host preferences. Spatial factors are more important than hosts in structuring the ECM fungal community in this ecosystem.
Despite the ecological importance of tropical ECM symbioses, the fungal symbionts remain poorly studied and relatively little is known about their biodiversity. The general latitudinal gradient of diversity (LGD) posits that biodiversity should increase with decreasing latitude and thus reach peak values in the tropics (Hillebrand, 2004). This pattern has been well documented in many groups of organisms, such as amphibians (Wiens et al., 2006), mollusks (Jablonski et al., 2006), angiosperms (Wright, 2002) and endophytic fungi (Arnold & Lutzoni, 2007). However, latitudinal diversity patterns for most fungal groups remain unresolved (Amend et al., 2010; Tedersoo & Nara, 2010). Recent work from Amazonia suggests that lowland Neotropical ECM fungi exhibit low diversity and high host preference for dispersed, low-density hosts (Tedersoo et al., 2010c). By contrast, other studies have documented high ECM fungal diversity in tropical cloud forests of Mexico (Morris et al., 2009), mixed dipterocarp forests of Borneo (Peay et al., 2010b) and West African forests dominated by leguminous trees (Diédhiou et al., 2010; Tedersoo et al., 2010b). These contrasting patterns indicate that ECM fungal diversity relationships are complex in the tropics and unlikely to be governed purely by latitudinal gradients (Tedersoo & Nara, 2010).
The factors that influence ECM fungal diversity patterns in different plant communities are not fully understood. Phylogenetically distant host plants occurring in sympatry usually host distinct ECM fungal communities (Ishida et al., 2007; Tedersoo et al., 2008; Smith et al., 2009). As many ECM fungi show some degree of host preference, host plant diversity could be a key driver of both local and global ECM fungal diversity (Dickie, 2007). Host preferences appear to be mediated by host plant identity (Tedersoo et al., 2008) but environmental influences such as litter quality and timing of litter deposition are also important (Aponte et al., 2010). Host phylogeny and physiognomy also influence ECM fungal preferences (Ishida et al., 2007; Morris et al., 2008).
Diversity and community structure of ECM fungi may also be influenced by the size of host plants and their distribution on the landscape (Morris et al., 2008; Tedersoo et al., 2010c). Small, widely dispersed sympatric host plants should host fewer ECM species because of limited root resources, reduced inoculum potential, and stochastic limits on fungal colonization (Peay et al., 2007, 2010a). These factors may interact to influence ECM fungal preferences and community composition among co-occurring host plants, but such interactions have not been fully explored.
Leguminous ECM trees are locally and regionally abundant in northeastern South America (Henkel, 2003; ter Steege et al., 2006). Sporocarp surveys indicate high ECM fungal diversity at some sites (e.g. Singer & Aguiar, 1986; Henkel et al., 2002), but there have been no comprehensive molecular studies of ECM mycobionts on leguminous hosts in this region. In this study we examined ECM fungal diversity and host preferences on the roots of three ECM tree species that occur sympatrically at high densities in the central Guiana Shield region of South America: Dicymbe corymbosa, Dicymbe altsonii and Aldina insignis. These are large, late-successional trees that reach the canopy in primary rainforests. The congeneric D. corymbosa and D. altsonii are presumably close relatives whereas Aldina is phylogenetically distant from Dicymbe (Doyle et al., 1997; Pennington et al., 2001). These two genera likely represent independent gains of the ECM habit from AM-forming ancestors in Fabaceae, but this hypothesis has not been explicitly tested (Alexander, 1989).
The high density of ECM trees and diversity of putative ECM fungi in combination with the sympatric occurrence of two putatively phylogenetically distant host genera prompted us to address the following questions: What is the ECM fungal diversity on the roots of the three host tree species?; What are the phylogenetic relationships among species of ECM Fabaceae and do Dicymbe and Aldina represent independent gains of the ECM habit?; Do the ECM fungal symbionts show preferences among the three host species?; and Do ECM fungi in this ecosystem show evidence of tropical origin or diversification?
Materials and Methods
The study site is located in primary rainforest in the Upper Potaro River Basin in the Pakaraima Mountains of Guyana and is part of one of the world’s largest intact forest ecosystems (Fig. 1; Potapov et al., 2008). The area receives c. 3500–4000 mm rainfall per year with peaks during May–July and December–January. Mean daily maximum temperatures range from 25 to 29°C and minima from 19 to 21°C. A base camp was established in an upland forest identified by Degagne et al. (2008) as co-dominated by the ECM canopy trees D. corymbosa Spruce ex Benth. and D. altsonii Sandw. (Fabaceae subfamily Caesalpinioideae) with scattered canopy trees of A. insignis (Benth.) Endl. (Fabaceae subfamily Papilionoideae). Dicymbe altsonii and A. insignis have typical arboreal architecture with a single bole and horizontal crown, although A. insignis is more strongly buttressed. By contrast, D. corymbosa has a reiterative physiognomy including a vertically layered crown, multiple stems which vary in size, and adventitious roots, resulting in a large litter-trapping capacity (Woolley et al., 2008). The base camp was located at 5°16′14.5′′ N; 59°50′39.1′′ W and the surrounding terrain is characterized by low, rolling hills with red loam lateritic soils, ranging in elevation from 650 to 800 m above sea level (asl). Details on the climate, geology, soils, forest structure and distribution of forest types have been published previously (Henkel et al., 2002, 2005a; Henkel, 2003; Degagne et al., 2008).
Ectomycorrhizal root sampling
We identified 19 sites (‘trios’) where individual trees > 20 cm diameter at breast height (dbh) of each of the three host plant species occurred at ≤ 25 m distance from each other. This yielded 57 total trees (19 per species) with a mean dbh of 94.5 cm. The trio sampling sites were distributed along several adjacent ridges with uniform soils and forest composition within 1 km of camp. Distances between site pairs ranged from 28 to 1320 m (mean = 689 m). Four lateral roots from each sampled tree were traced 1–5 m from the base to the fine roots, where roots, soil and litter were excavated and pooled until c. 1000 cm3 of material rich in ECM roots was obtained. Roots were harvested only when unequivocably traced back to the sample tree. These composite root samples were rinsed in water and inspected under a dissecting microscope. Twenty ECM roots were selected from each of the 57 trees. We attempted to maximize ECM diversity for each sample by choosing as many different morphotypes as possible up to a total of 20. A total of 1140 individual ECM roots were preserved in cetyltrimethylammonium bromide (CTAB) buffer and transported to the laboratory. Vouchers of representative trees were obtained and healthy leaves were silica-dried for DNA analysis (see the Supporting Information, Table S2).
Sporocarps of putative ECM fungi were collected within the tree sampling area in December 2009 and at a nearby site on the Upper Potaro River during 2000–2010. For information on site, specimen identification and herbaria see Henkel et al. (2002, 2005b, 2011) and references therein. Sporocarp DNA was extracted using a CTAB protocol (Gardes & Bruns, 1993) or an Extract-N-Amp Plant kit (Sigma-Aldrich, St Louis, MO, USA). Undescribed species were identified to the genus based on morphological data and rDNA sequences. Taxa detected only on roots were identified to genus and their uniqueness at the species level determined using blast comparisons against our in-house ribosomal DNA sporocarp database and GenBank. Internal transcribed spacer (ITS) sequences were considered conspecific if they differed by < 3% across the ITS region (Smith et al., 2007; Hughes et al., 2009). Taxa were assigned to the phylogenetically defined ECM fungal lineages recognized by Tedersoo et al. (2010a). These lineages constitute monophyletic groups that have independently evolved the ability to form ECM symbioses with plants. Any ECM root sequences that did not fit into these groups were putatively considered representatives of new ECM lineages.
Molecular protocols for fungi
The ECM roots were rinsed in water, dried on paper towels, and crushed with forceps in tubes containing 25 μl of extraction buffer from an Extract-N-Amp PCR kit (Sigma-Aldrich). Roots were then incubated and mixed with 25 μl of dilution buffer. Fungal internal transcribed spacer (ITS)-28S rDNA was PCR-amplified with forward primer ITS1F in combination with ITS4. When amplification with ITS4 was unsuccessful, we used either ITS2, ITS4B, or LR3 (Vilgalys & Hester, 1990; White et al., 1990; Gardes & Bruns, 1993). The PCR protocols followed Smith et al. (2007) with modifications in annealing temperature for different primers. Amplicons were visualized on 1.5% agarose gels stained with SYBR Green I (Molecular Probes, Eugene, OR, USA). Amplicons were cleaned with EXO and SAP enzymes (Glenn & Schable, 2005). Sequencing was performed with the same primers using Big Dye Sequencing Kit v.3.1 (Applied Biosystems, Foster City, CA, USA). Sequences were edited with sequencher v.4.1 (Gene Codes Inc., Ann Arbor, MI, USA).
Molecular protocols and phylogenetic analysis of host plants
Plant DNA was extracted from leaves of D. corymbosa, D. altsonii and A. insignis using the DNeasy Plant Kit (Qiagen). Three DNA regions were PCR amplified using published primers and protocols: ITS rDNA and chloroplast Maturase K (MatK) gene and the trnL intron (trnL). The MatK region was amplified and sequenced with primers trnK685F, trnK2R* and matK1100L (Wojciechowski et al., 2004), ITS was amplified and sequenced with NS7 and ITS4 (White et al., 1990), and trnL was amplified and sequenced with trnC and trnD (Taberlet et al., 1991). The trnL region was also PCR-amplified from randomly selected roots of each species to verify host plant identity.
To estimate the number of times the ECM habit has evolved in different lineages of Fabaceae, we compiled a main dataset of MatK and trnL sequences of representative taxa from across the family. This dataset included at least one taxon for each of the known ECM genera of Fabaceae and a diverse range of nonECM species from different subfamilies (Brundrett, 2009). Current phylogenetic concepts in Fabaceae followed Wojciechowski et al. (2004) and Bruneau et al. (2008) (Table S2). Attempts to analyse a more taxon-rich dataset with distant outgroups decreased our ability to infer relationships among the ECM Fabaceae lineages because alignments became progressively more ambiguous; we therefore restricted the number of taxa used in the analysis. We also compiled and analysed a second, three-gene (MatK, trnL, ITS) dataset consisting mostly of Caesalpinioideae and a few Papilionoideae to potentially improve phylogenetic resolution among ECM Caesalpinioideae.
Plant sequences were assembled in mesquite v. 2.73 (Maddison & Maddison, 2010) and aligned with the help of muscle (Edgar, 2004). The main dataset included 73 taxa. Both genes were available for all but the Australian ECM genera in the Mirbelieae for which only trnL was available. The alignment included 1034 bp of trnL and a 2019 bp of MatK. Once ambiguously aligned regions were excluded, 524 bp of trnL and 1679 of MatK were used for analysis. The second, three-gene dataset included 51 species and an additional 1881 bp alignment of ITS rDNA. Once ambiguously aligned regions were excluded, 2720 bp from the three genes were used for analysis. For both datasets Dialium guianense was selected as the outgroup taxon. This caesalpinioid taxon is a member of the basal tribe Dialinae and is not closely related to any of the other species in the analyses (Bruneau et al., 2008).
Preliminary analyses did not reveal strong phylogenetic discordance, so the genes were combined for all analyses. Maximum parsimony analyses (MP) were performed with paup* (Swofford, 2002) and maximum likelihood (ML) analyses were performed with garli (Zwickl, 2006). mr modeltest was used to determine that the GTR + I + G model was the best model for all analyses (Nylander, 2004). Bootstrap support was assessed with 1000 MP boostrap replicates using the ‘fast stepwise addition’ option in paup* whereas ML bootstrap analysis was performed with 500 bootstrap replicates using default settings in garli. Bayesian analyses were conducted with mr bayes v. 3.12 (Huelsenbeck & Ronquist, 2001). Posterior probabilities were obtained by running four Markov–Monte Carlo (MCMC) chains for 20 000 000 generations, sampling every 1000 generations and discarding the initial 15% of trees as the burn in.
Analysis of ectomycorrhizal diversity, host preferences and spatial patterns
The Shannon–Wiener diversity index (H′) and the Simpson’s diversity index (1-D) were calculated for ECM fungi using pc-ord 6 (McCune & Mefford, 2011). To compare the effect of site vs host tree species on the ECM fungal community structure, we calculated MRPP (Multiple Response Permutation Procedure) and perMANOVA (permutational multivariate analysis of variance) using pc-ord 6 (McCune & Mefford, 2011). Both MRPP and perMANOVA were performed for each grouping variable comparison using 999 randomized runs and four different distance measures (Euclidean, relative Euclidean, Jaccard and Sørensen) to ensure that results were consistent across distance measures. The MRPP is a nonparametric, multivariate procedure that tests the null hypothesis of no difference between groups (site and host species). The test statistic (T ) explains the separation between groups in multidimensional ‘species’ space; the more negative the T-value, the greater the separation between groups. The P-value quantifies separation between groups when compared with what is expected by chance and A represents the chance-corrected within group agreement and is a measure of effect size (McCune & Grace, 2002). Similar to conventional ANOVA, perMANOVA calculates an F-value statistic by comparing the ratio of among-group sums of squares with within-group sums of squares and does not assume multivariate normality. In perMANOVA, probabilities were calculated with Monte Carlo randomization procedures (999 runs), where P-values are the probability of obtaining an F-value larger than expected based on random permutations of the original data (McCune & Grace, 2002). To test for statistically significant differences in frequency of individual ECM fungal species, we used the software package r (R Core Development Team, 2007) to perform a Fischer’s exact test for each fungal species across each of the three pairwise tree comparisons. The Fisher exact test is a significance test used in the analysis of contingency tables with low sample sizes. We applied a Bonferroni correction to determine significance at P = 0.01667 (P = 0.05/3).
Biodiversity and biogeography of the /clavulina lineage
The monophyletic /clavulina lineage is essentially equivalent to the coral mushroom genus Clavulina (Cantharellales) (Tedersoo et al., 2010a; J. K. Uehling et al., unpublished). The high species diversity of this lineage in tropical South America based on sporocarps (Henkel et al., 2011) and ECM roots in this study prompted us to examine global /clavulina diversity and biogeographical patterns. All /clavulina ITS sequences in GenBank were identified and downloaded using the Emerencia website (Ryberg et al., 2009). This search produced 535 sequences supplemented by 53 unpublished sequences obtained from L. Tedersoo (Tartu University, Estonia). The ITS barcode was selected because it was the only DNA region with global representation of /clavulina sequences. All sequences were examined in a preliminary alignment and with blast searches. Sequences that were putatively chimeric, were < 300 bp long, had many ambiguities or aberrant 5.8 s sequences, or that had blast matches to genera other than Clavulina were discarded. The 588 ITS sequences were combined with 44 unique /clavulina sequences from Guyana (23 from sporocarps and 21 from roots). The resulting 632 sequences were partitioned into operational taxonomic units (OTUs) meant to approximate fungal species by grouping sequences at the ≥ 97% similarity level using the ‘clean data’ algorithm in sequencher (Smith et al., 2007). For each OTU, one high-quality sequence was included in the final analysis. Frequent indels in /clavulina ITS sequences prevented reliable phylogenetic analysis so two alternative methods were used to analyse diversity: comparison of ITS length variation in paup*; and calculation of uncorrected pairwise genetic distances between all ITS sequences using mothur (Schloss et al., 2009). In mothur, each of the 143 global OTUs was compared with every other OTU using a ‘local alignment’. As each pair of sequences was independently aligned, this provided more accurate comparisons than those obtained with a global alignment. The ‘pairwise.seqs’ command was used with the blast algorithm and the following settings: +1 reward for matching bases, −2 penalty for mismatching bases, terminal gaps not penalized, and gap costs = linear. To assess the diversity of /clavulina in different biogeographical regions, /clavulina OTUs were sorted into four areas of origin: temperate zones (including Northern and Southern Hemispheres); tropical zones; tropical South America and global. For each region, mean genetic distances and standard deviations for the /clavulina sequences were calculated and compared. We also examined how two important morphological features, sporocarp morphology and sterigmata (slender projections where the basidiospores are produced) per basidium, were distributed in the different world regions.
Diversity and structure of the ectomycorrhizal fungal community
Internal transcribed spacer sequences of ECM fungi were successfully generated from c. 90% (1020/1140) of the roots sampled from the 57 host trees. We detected 118 distinct OTUs that corresponded with ECM fungal species. These represent 17 different phylogenetic lineages (Table S1). Approx. 41% of the fungal OTUs from roots were matched to sporocarps (49/118). Sixty-eight fungal OTUs were detected three times or more, along with 20 doubleton and 32 singleton OTUs. We detected 9–18 fungal OTUs per individual tree (mean = 13.8, SD = 2.3). The high species richness, high Shannon–Wiener Index value of H′ = 2.648, and the high Simpson’s Diversity Index value of 1-D = 0.925 indicate that ECM fungal diversity was comparable to that found in a number of species-rich temperate and tropical sites (Peay et al., 2010b; L. Tedersoo et al., unpublished data). The most OTU-rich lineages were /clavulina (21), /russula-lactarius (21), /boletus (19) and /tomentella-thelephora (15). The fungal community was dominated by Basidiomycota with only three Ascomycota OTUs from the /elaphomyces and /sordariales lineages.
We detected five fungal OTUs that could not be assigned to any of the ECM lineages defined by Tedersoo et al. (2010a). The closest sequence matches for four of these OTUs were ECM fungi from Bornean dipterocarp roots (Table 1) (Peay et al., 2010b). These include one OTU in Atheliales, two OTUs within a lineage of Polyporales and two OTUs in separate lineages of Agaricales (Tables 1, S1). Although ITS sequences resolve these ECM fungi at the ordinal level, their closest relatives are not completely clear. ITS sequences of the two Polyporales OTUs are c. 90% similar to the ITS sequence from vouchered ECM roots of D. corymbosa (TH70815). Microscopic examination of ECM collection TH70815 revealed normal ECM formation, including a fungal mantle, Hartig net and otherwise healthy root tissue (Fig. S1).
Table 1. Closest blast matches and frequency of occurrence of five root-inhabiting ectomycorrhizal (ECM) fungal operational taxonomic units (OTUs) that could not be assigned to any previously known ectomycorrhizal fungal lineages
ECM taxon name
Closest blast match
Frequency on roots
1Sequences of 28S rDNA from ECM259 matched closest with Phlebiella spp. and Polyporales ECM LH80 (Malaysia) GQ268627.
Most ECM fungal OTUs, including all of the 28 most frequently detected fungi, were generalists associated with all three of the host plant species (Fig. 2). Similar richness was detected on the three hosts (D. corymbosa = 87 OTUs, D. altsonii = 89 OTUs and A. insignis = 82 OTUs). Sampling effort curves indicate that the community was not exhaustively sampled, but the three hosts were equitably sampled (Fig. 3). We directly sequenced trnL from 33 randomly selected ECM roots to corroborate host plant identifications. Sequences from each of these matched that obtained from leaves of the expected host species (data not shown).
There was no evidence of host specialization among ECM fungi (Fig. 2). Among the 88 nonsingleton OTUs, 73 (83%) were detected on all three of the host plant species and 81 (92%) were found on at least two of the host species. Thus, only seven nonsingleton OTUs were restricted to a single host species and all of these fungi were rare (i.e. found on only two trees). The Fisher’s exact tests showed that 117 out of 118 ECM species were not statistically different in their associations with the three host plant species. Only one ECM fungus, Tomentella MES348, was statistically more frequent on D. corymbosa than on D. altsonii.
Both MRPP and perMANOVA tests indicated that variance in the fungal community was not significantly partitioned by host species regardless of the distance measure used. The sampling site of a trio, however, was a significant explanatory variable (Table 2). Although some of the host trios had ECM communities that were statistically different from other trios, the variance explained by site was, nonetheless, low based on MRPP (Euclidean T = −2.45, Sørensen T = −2.41) and perMANOVA (variability explained, Euclidean = 3.7%, Sørensen = 7.4%). These measures imply that a modest degree of geographical structure is present in the ECM fungal community at sampling distances of < 1 km. The local spatial patterning in the ECM fungal community was probably influenced by several cases where uncommon fungal OTUs were present on two or three trees in a given trio, but were rare or absent from other nearby trios.
Table 2. Results of two statistical tests, MRPP (Multiple Response Permutation Procedure) and perMANOVA (permutational MANOVA), used to examine potential impact of host tree species and sampling site on the structure of the ectomycorrhizal fungal community on three leguminous tree species (Dicymbe corymbosa, Dicymbe altsonii and Aldina insignis) in a tropical rainforest in Guyana. The results of each test are shown using both Euclidean and Sørensen distance measures
T, the separation between groups in multidimensional ‘species’ space; the more negative the T-value, the greater the separation between groups.
A, the chance-corrected within group agreement and is a measure of effect size (McCune & Grace, 2002).
Host tree species
Variability explained (%)
Host tree species
Biodiversity and biogeography of the /clavulina lineage
The pruned dataset of 143 global /clavulina OTUs consisted of 58 OTUs from temperate zones and 85 OTUs from the tropics. The tropical South American /clavulina assemblage accounted for 48 of 143 OTUs, c. 34% of the global total (Table 3). The average genetic distance for tropical South American /clavulina was 0.51, only slightly less than the global average genetic distance of 0.52. Tropical South American /clavulina were also more variable in the full ITS length than those from the temperate regions and were almost as variable in length as the /clavulina sequences from the entire world. There was a similar pattern when ITS1 and ITS2 were analysed separately. Furthermore, there is greater morphological variation in Clavulina species from tropical South America than in those from other geographical regions (Table 3).
Table 3. Global diversity patterns in the /clavulina lineage based on morphology and on internal transcribed spacer (ITS) sequence data from GenBank and this study
Operational taxonomic units (OTUs, % of total)
Mean genetic distance1 (SD)
Sterigmata per basidium3
Mean full ITS length4 (range)
Mean ITS1 length4 (range)
Mean ITS2 length4 (range)
1Genetic distance based on pairwise ITS comparisons using length-adjusted blast scores (see the Materials and Methods section for details).
2Morphological diversity includes four main sporocarp types: clavarioid (C), resupinate (R), infundibuliform (I) and monopodial (M).
4ITS length calculations based on taxa with complete sequences for each comparison; full ITS = 124 spp., ITS1 only = 134 spp., ITS2 only = 127 spp.
C, R, I, M
Tropical South America
C, R, I, M
C, R, I, M
Phylogenetic analysis of ectomycorrhizal Fabaceae
For the two-gene plant dataset (MatK and trnL), the final alignment of 2203 bp yielded 604 parsimony-informative characters. Parsimony analysis resulted in 19 368 equally parsimonious trees of 3163 steps (CI = 0.592, RI = 0.821). Both MP and ML bootstrap analyses and Bayesian posterior probabilities supported relationships shown by published Fabaceae phylogenies. These relationships include several monophyletic clades within Caesalpinioideae as well as monophyly of Papilionoideae and Mimosoideae (Wojciechowski et al., 2004; Bruneau et al., 2008). The overall topology of the consensus tree (Fig. 4) was similar to the best ML tree (−loge 10 124.98, phylogram not shown). The phylogeny identified four distinct lineages of ECM-forming Fabaceae: the Berlinia + Intsia clade (Caesalpinioideae); the Australian Acacia clade (Mimosoideae); the Australian Mirbelieae clade (Papilionoideae); and the genus Aldina (Papilionoideae). All analyses showed that each of these ECM lineages was nested within a larger nonECM clade, suggesting that the ECM habit has evolved at least four times. Dicymbe spp. were nested within the Berlinia + Intsia clade whereas A. insignis was not closely related to any other ECM taxa.
The three-gene dataset included fewer taxa but used an additional locus (ITS rDNA). Analysis of this dataset generated 108 equally parsimonious trees (2312 steps, CI = 0.668, RI = 0.802) and an optimal ML tree with likelihood of loge−16 743.28171 (Fig. S2). The ML tree and the strict MP consensus were similar overall and showed the same basic relationships with only minor arrangements in the Berlinia clade. The Intsia + Afzelia clade was resolved as phylogenetically distinct but sister to the Berlinia clade. Also included within this crown Berlinia + Intsia group of mostly ECM-forming Caesalpinioideae are putatively nonECM taxa Brownea coccinea, Hymenostegia afzelii and Tamarindus indica. In the Berlinia group, all genera and species are known to form ECM except for Humboldtia vahliana (Brundrett, 2009).
Diversity and structure of the ectomycorrhizal fungal community
Trees in the genera Dicymbe and Aldina are known to form ECM symbioses (Singer & Araujo, 1979; Henkel et al., 2002) but this is the first in-depth molecular study of fungal ECM symbionts of leguminous trees in the Neotropics. At a single site we detected 118 ECM fungal OTUs on the roots of D. corymbosa, D. altsonii, and A. insignis, each of which harbored 82–89 OTUs. The diverse ECM assemblage included members of 17 fungal lineages and, similar to other tropical studies, we found a high species richness in the /russula-lactarius, /boletus, and /tomentella-thelephora lineages. Unique to this study was the exceptional diversity in the /clavulina lineage. The high overall fungal species richness and the concurrently high Shannon–Wiener and Simpson’s diversity estimates indicate that the ECM fungal diversity of this Guiana Shield site is on a par with dipterocarp forests of Borneo and many ECM-rich forests from temperate and boreal zones (Peay et al., 2010b; Tedersoo & Nara, 2010; L. Tedersoo et al., unpublished data).
Despite the phylogenetic differences and independent evolution of the ECM symbiosis in Dicymbe and Aldina, we found almost no evidence of ECM fungal host preferences at any taxonomic level (Table 2). Nearly 92% of fungal OTUs were multihost generalists detected on the roots of two or three of the tree species (Fig. 2). Furthermore, Fisher’s exact tests illustrated that, at the species level, 117 out of 118 ECM species were not statistically associated with one host tree over another. This situation is similar to that in West African forests of Fabaceae and Phyllanthaceae where ECM fungal symbionts exhibited almost no preferences among five sympatric host tree species (Diédhiou et al., 2010). By contrast, other studies have demonstrated fungal host preference in communities with multiple sympatric ECM hosts in both the temperate zones (Ishida et al., 2007; Tedersoo et al., 2008; Smith et al., 2009) and the lowland Neotropics (Tedersoo et al., 2010c). In an Ecuadorian rainforest Tedersoo et al. (2010c) detected only 38 ECM fungal species from seven lineages, but the fungi exhibited strong preferences among the 26 sympatric host species of Coccoloba (Polygonaceae), Neea and Guapira (Nyctaginaceae). Although the co-occurring host plant species were diverse at the Ecuadorian site, the plant populations consisted of widely dispersed understory plants scattered throughout a forest overwhelmingly dominated by AM canopy trees. Tedersoo et al. (2010c) hypothesized that the low ECM fungal diversity and strong host preferences were partly a result of resource and dispersal limitations for the mycobionts. They suggested that these isolated host plants functioned as discrete habitat islands with limited photosynthetic resources and inter-connectivity.
The situation in many Guiana Shield rainforests is quite different. The high ECM fungal diversity on only three sympatric host plant species reported here may be attributable to the large resource pool locally available to mycobionts. The study site is nested within a part of Guyana that is codominated by ECM Dicymbe species, along with large, scattered individuals of A. insignis, which together can account for 65–85% of the canopy trees (Degagne et al., 2008). Dicymbe species dominate these central Guiana Shield forests via a combination of life-history traits including the ECM symbiosis, mast fruiting and resulting high recruitment, large and persistent root biomass, and reiterative growth that yields multistemmed individuals with indeterminate life-spans (Zagt, 1997a,b; Henkel et al., 2005a; Mayor & Henkel, 2005; Woolley et al., 2008). Dicymbe corymbosa may also restrict recruitment of other trees by creating highly shaded understory conditions in conjunction with large litter accumulations, rendering it difficult for AM trees to establish (McGuire, 2007). Dominance of closed-canopy forests by Dicymbe species is currently documented only for western Guyana and at some locations Dicymbe species codominate stands with other canopy host genera such as Aldina and Pakaraimea (Dipterocarpaceae) (Henkel et al., 2002; T. Henkel et al. unpublished data; this study), ECM genera that also occur more widely in the Guiana Shield region (Moyersoen, 2006; Mardegan et al., 2009). In this ecological setting, barriers to ECM fungal success and diversification owing to resource and dispersal limitations should be minimal. We hypothesize that in the central Guiana Shield, local forest dominance of ECM plants along with regional prevalence of several ECM canopy tree genera are major drivers of ECM fungal diversity.
Phylogenetic relationships of ectomycorrhizal Fabaceae
Trees in the Fabaceae are dominant components of the woody flora in much of the Neotropics, West Africa and Australia, and to a lesser extent in Southeast Asia (Newbery et al., 1997, 2002; Torti et al., 2001; Crisp et al., 2004; ter Steege et al., 2006). Although phylogenetic studies have addressed morphology, taxonomy and root nodulation in this ecologically important plant family, the ECM symbiosis is rarely mentioned in the Fabaceae literature and this trait has not been studied in a phylogenetic context (Sprent & James, 2007). This aspect of Fabaceae biology was likely neglected because of conflicting reports on ECM formation in some species and because ECM-forming taxa are primarily tropical or Austral and infrequently collected.
Our phylogenetic analysis indicates that the ECM symbiosis has evolved in at least four distinct lineages of Fabaceae (Fig. 4). The phylogeny depicts A. insignis among nonECM Papilionoideae and confirms that this species is phylogenetically distant from the caesalpinioid Dicymbe species. As shown by Bruneau et al. (2008), Dicymbe is related to a diverse clade of African ECM Caesalpinioideae (Figs 4, S2). The separation of different ECM lineages (e.g. Dicymbe, Intsia + Afzelia and Berlinia clades) in various phylogenetic analyses that have used different data and methods (Bruneau & Forest, 2001; Bruneau et al., 2008; this study) indicates the possibility of multiple, independent ECM origins within the Caesalpinioideae. An alternate hypothesis is that the ECM symbiosis may have been lost in some caesalpinioid taxa (e.g. B. coccinea and T. indica; Fig. 4). More phylogenetic data and ECM root sampling of Caesalpinioideae species will be necessary to resolve the evolutionary history of the ECM symbiosis in this group.
Origins and diversification of tropical ECM fungi
Tedersoo & Nara (2010) suggested that ECM fungi reach peak diversity at higher latitudes and that most ECM lineages evolved in the temperate zones. In their analysis, only the /russula-lactarius lineage was more diverse in tropical than in temperate habitats. By contrast, the /inocybe lineage was more diverse in the temperate zones, despite the fact that this group apparently arose in the paleotropics (Matheny et al., 2009; Tedersoo & Nara, 2010). Tedersoo et al. (2010a) could find no evidence of endemic tropical ECM fungal lineages and noted that several temperate ECM lineages are apparently absent or species-poor in the tropics (e.g. /albatrellus, /cenococcum, / endogone and /hygrophorus). However, these assessments were based on a limited amount of data from the tropics and additional studies from tropical ecosystems may alter this view of ECM fungal biogeography and biodiversity.
We detected several fungal OTUs in the Agaricales, Polyporales and Atheliales that could not be assigned to any of the monophyletic ECM lineages defined by Tedersoo et al. (2010a). The closest blast matches for these enigmatic OTUs were mostly sequences from ECM roots of Bornean dipterocarps (Table 1; Peay et al., 2010b). The occurrence of these fungi on healthy ECM roots in two widely disjunct tropical sites suggests that they may represent previously undocumented tropical ECM fungal lineages. Some morphological data also support their ECM status; ITS sequences of the Polyporales OTUs were similar to those of vouchered ECM specimens that have a typical mantle and Hartig net (Fig. S1). Future morphological studies of ECM roots in the Guiana Shield and other tropical sites will help clarify the identity and symbiotic functioning of these putatively ECM fungi.
More important was our finding that the /clavulina lineage was highly diverse and abundant at the Guyana site. Several lines of evidence suggest that the /clavulina lineage, which is more-or-less equivalent to the coral mushroom genus Clavulina (Cantharellales) (Tedersoo et al., 2010a; J. K. Uehling et al., unpublished), has a previously unrecognized center of diversity in tropical South America. We found 21 distinct /clavulina OTUs on roots of the three host plant species. All but one of the 57 trees hosted at least one /clavulina OTU and each individual tree hosted an average of 3.5 /clavulina OTUs. One individual D. altsonii tree hosted seven /clavulina OTUs, compared with the six phylogenetically delimited species documented for the continent of Europe (Olariaga et al., 2009).
Our analysis of global /clavulina ITS sequences also suggests that this lineage is most diverse in tropical habitats. Out of 143 global /clavulina OTUs, 59% were from tropical regions, despite much lower sampling effort in the tropics. Tropical South American sequences accounted for one-third of the world’s /clavulina OTUs and the mean genetic distance for tropical South American OTUs was almost as high as that for the global /clavulina dataset (0.51 vs 0.52, Table 3). The wide variability in ITS length among the tropical South American OTUs is an additional indicator of high diversity (Table 3). Furthermore, ECM fungal sporocarp surveys in Dicymbe forests have uncovered many new and morphologically divergent species in the /clavulina lineage and shown that /clavulina sporocarp diversity is higher at the Guyana sites than anywhere else in the world (Thacker & Henkel, 2004; J. K. Uehling et al., unpublished; T. Henkel et al., unpublished data). Molecular analyses based on rpb2 and 28S rDNA also suggest that the /clavulina lineage is phylogenetically diverse in tropical South America (Thacker & Henkel, 2004; J. K. Uehling et al., unpublished). While the /clavulina lineage is globally distributed (Tedersoo et al., 2010a), the molecular and morphological data presented here suggest that this lineage likely evolved or underwent diversification in tropical South America. As far as we know, this is the most convincing case thus far for an evolutionary radiation among ECM fungi in the tropics.
The high diversity and absence of host preferences for ECM fungi in this Guiana Shield rainforest alter the current view of ECM communities in lowland Neotropical forests. Although host identity and dispersion strongly influences the ECM fungal community in tropical forests with scattered understory hosts that are phylogenetically distant (Tedersoo et al., 2010c), this is not the case in Fabaceae-dominated tropical forests where the hosts are large canopy trees growing in dense stands. A high degree of host sharing by ECM fungi was prevalent whether hosts belonged to the same genus (i.e. Dicymbe species) or represented independent evolutionary acquisitions of the ECM symbiosis (i.e. Dicymbe vs Aldina). The robust diversity of ECM fungi, the discovery of a center of diversity for the /clavulina lineage and presence of unique, putatively tropical-endemic ECM fungal lineages in the Guiana Shield emphasize the need for expanded ECM studies in the tropics.
Funding was provided by the National Science Foundation grants DEB-0918591 to T.W.H. and DEB-3331108 to R.V. Permits were granted by the Guyana Environmental Protection Agency. Field assistance was provided by D. Husbands, P. Joseph, C. Andrew, F. Edmond, L. Edmond, R. Edmond and P. Joseph. Laboratory support was provided by G. Williams and E. C. Smith provided invaluable logistical support during fieldwork. Leho Tedersoo kindly provided unpublished sequence data for Clavulina species. Anonymous reviewers improved this work with constructive editorial critiques.