Host effects on ectomycorrhizal fungal communities: insight from eight host species in mixed conifer–broadleaf forests

Authors


Author for correspondence: Takahide A. Ishida Tel: +81 42 465-5601 Fax: +81 42 465-5616 Email: ishida@anesc.u-tokyo.ac.jp

Summary

  • • To advance our understanding of host effects on the community structure of ectomycorrhizal fungi (EMF), EMF communities were compared among different host species, genera and families in two mixed conifer–broadleaf forests in Japan.
  • • Using molecular identification methods we examined EMF root tips of eight coexisting species belonging to six genera (three families): Abies and Tsuga (Pinaceae), Betula and Carpinus (Betulaceae) and Fagus and Quercus (Fagaceae).
  • • In total, 205 EMF species were detected, and the total richness was estimated to exceed 300 species using major estimators. Of the 55 EMF species occurring three or more times, eight showed significantly biased host preference. A Mantel test showed a significantly negative correlation between EMF community similarity and host taxonomic distance. Detrended correspondence analysis separated EMF communities mainly by host taxonomic and successional status.
  • • Thus, EMF communities are similar on hosts with similar taxonomic and successional status. A significant proportion of EMF exhibited host specificity, which may contribute to the extremely diverse EMF community in conifer–broadleaf forests.

Introduction

Ectomycorrhizal fungi (EMF) improve host performance by enhancing nutrient and water uptake from the soil and protecting host roots from pathogens and toxic compounds (Smith & Read, 1997). In a developed forest ecosystem, several tens or more of ectomycorrhizal fungal species coexist and comprise diverse EMF communities, even in association with a single host species. Because EMF communities affect host productivity and seedling establishment (Baxter & Dighton, 2001; Jonsson et al., 2001; Dickie et al., 2002), it is important to understand the determinants of EMF community structure.

The EMF community structure is affected by a number of biotic and abiotic factors such as EMF succession (Nara et al., 2003), interspecific interaction of EMF (Koide et al., 2005), litter quality (Conn & Dighton, 2000), climate (O'Dell et al., 1999), soil type (Moser et al., 2005; Walker et al., 2005) and soil nutrient status (Nantel & Neumann, 1992). Host species may also be important in structuring EMF communities, although this has not been examined explicitly.

Comparisons of EMF among host species within the same environmental conditions would help elucidate the effects of host species on EMF communities. Below-ground EMF communities have been compared between two host species within the genus Quercus (Walker et al., 2005), two species from different genera in Pinaceae (Horton & Bruns, 1998; Cullings et al., 2000), two species from different families (Kennedy et al., 2003; Richard et al., 2005) and among three host species of different families (Nara, 2006) using molecular techniques. These studies demonstrate that even though a substantial number of EMF species are shared among coexisting host species, some EMF are not shared, especially among host families. Such EMF specificities to a certain host taxon may affect EMF community structure.

The extent of this host effect remains largely unknown, because the frequency of occurrence of most EMF is too rare to be evaluated statistically. Moreover, no studies have compared EMF communities among hosts of different species, genera, and families concurrently. Thus, little is known about whether host species belonging to the same family share more EMF than do host species of different families, and if so, to what degree. To advance our understanding of the effect of host taxonomic distance on EMF communities, we should examine EMF communities of host species, genera, and families occurring within the same site conditions.

Trees of the families Betulaceae, Fagaceae and Pinaceae are among the major elements of temperate to cool forests in the Northern Hemisphere. These trees are obligatorily ectomycorrhizal in nature, although previous studies of EMF communities have concentrated more on species of Pinaceae than on broadleaf species. In addition, most EMF community studies have been conducted in relatively simple forests composed of one or a few dominant ectomycorrhizal host species. In Japan, complex forests in which a variety of tree species codominate are widespread. This allowed us to study the host effect on EMF communities within a site. We compared EMF communities associated with eight codominant tree species: Abies homolepis and Tsuga sieboldii (Pinaceae), Betula maximowicziana, B. grossa and Carpinus japonica (Betulaceae), and Fagus crenata, F. japonica and Quercus crispula (Fagaceae).

Materials and Methods

Study sites

This study was conducted in two conifer–broadleaf forests, i.e. a secondary forest in the Iriyama district and a primary forest in the Irikawa district, in the Chichibu Experimental Forest of the University of Tokyo in central Japan (35°56′−35°57′ N, 138°48′−138°49′ E, 1350–1500 m above sea level). The canopy of the secondary forest site, which was clear-cut in 1927, was composed mainly of Q. crispula, B. maximowicziana, F. crenata, Castanea crenata, and Acer spp. Abies homolepis was the only coniferous species at this site, and most individuals were still small in diameter at breast height (d.b.h.; Table 1). The primary forest site was located 2 km from the secondary forest and supported an old growth forest. The canopy species were mainly dominated by ectomycorrhizal hosts (listed in Table 1) and were intermixed with other tree species (i.e. Carpinus cordata, Carpinus laxiflora, Acer spp., Styrax obassia, Ilex macropoda, Abies firma, and Fraxinus lanuginosa). Detailed descriptions of the tree species in and around the primary forest were given by Sawada et al. (2006). The soils were sandy clays with thin organic layers (Sawada et al., 2006). Both sites were located on convex south-facing slopes in the cool temperate zone. The annual mean temperature near the sites was 10.9°C (2001–04) and the mean annual precipitation was 1596 mm (2003–04; available at http://www.uf.a.u-tokyo.ac.jp). The sampling area was approx. 5.8 ha and 4.0 ha in the secondary and primary forests, respectively.

Table 1.  Species, diameter at breast height (d.b.h.) and number of trees examined, and number of observed root tips
SiteHost speciesAbbreviationnrange of d.b.h.Number of root tips
Secondary forestAbies homolepisAh10 8.3–13.7 644
Betula maximowiczianaBm1036.7–62.11110
Carpinus japonicaCj10 9.0–18.21163
Fagus crenataFc10 6.7–24.91553
Quercus crispulaQc1018.8–29.71164
Primary forestAbies homolepisAh1018.8–89.2 755
Tsuga sieboldiiTs1012.8–58.0 594
Betula maximowiczianaBm1030.0–38.91100
Betula grossaBg1019.5–33.8 942
Fagus crenataFc10 5.5–57.31603
Fagus japonicaFj1023.3–54.22031
Quercus crispulaQc1024.2–33.51193

Sampling

From late July to early August 2004, five trees of each species were randomly chosen at each site (Table 1). From each selected tree, we sampled two separate root systems, including a main root (c. 20–30 cm in length and < 5 mm in diameter) and many attached fine root tips. To avoid contamination from other tree roots, all root systems were traced to the trunk of the selected trees. The root systems sampled were placed separately in plastic bags and kept at 4°C for further analyses. In 2004, EMF sporocarps were sampled biweekly in and around the research sites and used for the identification of below-ground EMF.

Ectomycorrhizal tips for morphotyping and molecular analysis

The root systems were washed carefully with tap water. Each EMF root tip was assigned to a morphotype based on its surface color, texture, and emanating hyphae by examination under a dissecting microscope. If a root system had an extensive number of root tips, we randomly subsampled about half of the root tips. For the molecular identification of EMF, we used 1–13 replicate EMF root tips from each morphotype from each root system. Root tips were placed individually in 2.0-ml tubes and dried for DNA extraction. In total, 13 852 root tips from 120 root systems were examined under a dissecting microscope, and 1550 EMF root tips were used for molecular identification.

DNA preparation and internal transcribed spacer fragment analysis

Extraction of DNA was performed using the method described by Nara et al. (2003). Briefly, each sample was pulverized in a 2.0-ml tube containing a zirconia ball using a homogenizer and subjected to the modified cetyltrimethylammonium bromide (CTAB) method.

The rDNA internal transcribed spacer (ITS) regions were amplified using Amplitaq Gold (Applied Biosystems, Foster City, CA, USA); KOD Plus DNA polymerase (Toyobo, Osaka, Japan) was used when a polymerase chain reaction (PCR) product was faint or absent using Amplitaq Gold. We followed the manufacturer's instructions and the methods described by Nara et al. (2003) and Nara (2006). The ITS1F primer (Gardes & Bruns, 1993) with a 5′ D2 fluorescent label (Proligo, Kyoto, Japan) and the ITS4 primer (White et al., 1990) with a 5′ D3 fluorescent label (Proligo) were used.

We incubated 2 µl of each PCR product with 1.5 U of either HinfI or AluI (Takara Shuzo, Shiga, Japan, or Toyobo) in the reaction mixture. The digested products were purified by ethanol precipitation with sodium acetate and diluted with 25 µl of HiDi formamide (Applied Biosystems) containing 0.06 µl of CEQ 600 size standard (Beckman Coulter, Fullerton, CA, USA).

Capillary gel electrophoresis was performed using a CEQ 8800 (Beckman Coulter). The diluted DNA solutions were denatured for 2 min at 95°C, injected for 50 s at 2.0 kV, and separated for 65 min at 35°C for HinfI-digested samples and 50°C for AluI-digested samples. The terminal restriction fragment lengths were determined using FRAGMENTS implemented in a CEQ 8800 genetic analysis system (Beckman Coulter). Two DNA polymerases (Amplitaq Gold and KOD Plus) showed no difference in results when the same genomic DNA was used.

Four fragments were obtained for each sample. When samples had fragments within ±2 bp in all four fragments, they were considered to be the same terminal restriction fragment length polymorphism (T-RFLP) type. When more than two fragments were observed for a fluorescent dye in one sample, the highest peak was used.

Identification of fungi

The T-RFLP types of EMF root tips were compared with sporocarp T-RFLP types. When the T-RFLP type of an EMF root tip matched that of a sporocarp within ±2 bp, the EMF root tip was assigned to the sporocarp species. If no matching sporocarp was found, the T-RFLP types were sequenced. A PCR product amplified using ITS1F and ITS4 was purified using a PCR product presequencing kit (USB Co., Cleveland, OH, USA). If we could not obtain a clear sequence from a sample, we tried another sample belonging to the same T-RFLP type. When the ITS sequence of a major T-RFLP type could not be determined by direct sequencing, the PCR product of ITS1F and ITS4 was subcloned with a pT7Blue Perfectly Blunt cloning kit (Novagen Inc., Madison, WI, USA) following the manufacturer's instructions.

The obtained sequences were compared with the sequences of known species in the UNITE database (Kõljalg et al., 2005) using Galaxy blast. When close matches were unavailable, a BLAST search was conducted in GenBank. When different T-RFLP types resulted in the same closest match, their sequence homology was analysed using paup 4.0b2. When the sequence homology was > 99%, the T-RFLP types were regarded as the same taxon. All sequences were deposited in the DNA Data Bank of Japan (DDBJ; Table S1).

Analyses

The numbers of root tips colonized by individual EMF species were determined for each root system using the molecular identification results. The relative abundance of an EMF species in a root system was calculated as the percentage of EMF root tips in the root system found to have that EMF. The relative abundance in two replicates from each host was pooled to calculate the relative abundance in this individual host. The relative abundance in five trees from a host species/site was pooled to calculate the relative abundance in a host species/site. Similarly, the relative abundance in all hosts within each site was averaged and used as the relative site abundance. The frequency of each EMF species was the total number of trees in which a given EMF was found.

The diversity of EMF communities was analyzed using richness, Simpson's diversity (1/D), and Shannon–Wiener information (H′) indices. For the estimation of richness, first- and second-order jackknife and Chao2 richness estimators were calculated using estimates version 7.5 (Colwell, 2005), based on the frequency of EMF taxa. To compare richness estimators between the sites, we used the rarefaction method without replacement adjusted to the lower number of samples (25 trees in the secondary forest).

To compare EMF communities between hosts, Morisita–Horn similarity indices were calculated for each pair of host species using estimates version 7.5 (Colwell, 2005) and the total number of host species root tips with a particular species of EMF. The relationship between EMF community similarity and the taxonomic distance separating host species was analysed using a Mantel test. Taxonomic distance classes were assigned to each host pair, where distance classes were: 1, the same species from each site; 2, two different host species within the same genus; 3, a pair of host species belonging to different genera within the same family; 4, a host pair made up of a Fagaceae and a Betulaceae species; and 5, a host pair consisting of a conifer and a broadleaf species. The autocorrelation analysis was conducted on the distance matrix filled with integers from one to five. For each of the five distance classes, a binary taxonomic distance matrix filled with only 0s and 1s was constructed. For example, the binary matrix for distance class 2 had a 0 for every object pair (distance class value) except those that were originally assigned a value of 2, and the 2′s were replaced with 1s. The Mantel test was then conducted using the EMF similarity matrix and the binary taxonomic matrix for each distance class (Oden & Sokal, 1986; Ishida & Kimura, 2003). In addition, a simple Mantel test (Mantel, 1967) was also applied. These Mantel tests were performed using r 2.4 software (R Development Core Team, 2006) with the packages ‘vegan’ version 1.6 and ‘ecodist’ version 1.01 (available at http://cran.za.r-project.org/src/contrib/Descriptions), with 1000 permutations on the distance matrix elements.

Detrended correspondence analysis (DCA) was also used to visualize the similarity of host species EMF communities. The analysis was performed using r 2.4 software (R Development Core Team, 2006) with the package ‘vegan’, using the standardized relative abundance data after arcsine square-root transformation. Taxa that appeared in a single sample were excluded from the analyses.

To assess the effects of site on EMF communities, four commonly studied host species were compared between the sites. To assess the specificity of each EMF, Fisher's exact test was used after combining EMF frequency data for the same host species in both sites.

To test the independence of samples collected from the same trees, we compared Morisita–Horn similarity indices for sample pairs from the same trees to those for sample pairs from different trees within the same species/sites by Monte Carlo analysis using 1000 randomized permutations.

Results

General descriptions

The T-RFLP patterns were obtained from 1396 root tips and were assigned to 209 T-RFLP types. Polymerase chain reaction (PCR) amplification was unsuccessful for 154 samples, which were removed from further analyses. Of 145 EMF sporocarps, the ITS regions of 137 were successfully amplified and classified into 96 T-RFLP types. Among these, only 28 T-RFLP types matched those of ectomycorrhizas. Of the 150 T-RFLP types for which ITS sequences were successfully determined, four T-RFLP types showing high sequence homology to four other T-RFLP types (99%) were not regarded as unique fungal species. Many of these sequenced samples showed high similarities to European species, but most of them were not similar enough to be considered the same species (see the Supplementary Material, Table S1). A total of 205 T-RFLP types were regarded as putatively different EMF species (Table S1). Of the 205 EMF species, 105 species were found in single root systems and another seven species were found in single trees. When all 60 individuals were included, total EMF richness was estimated to be 362, 315, and 387 using Chao2 and Jackknife1 and 2, respectively.

The average Morisita–Horn similarity index for the pairs of samples from the same trees was 0.210, which was significantly higher than the value for the sample pairs from different trees within the same species/site (0.044; P < 0.01). Thus, we did not treat each root sample as an independent sample. Instead, data of two samples from each tree were combined and used as an independent replicate for further analyses.

The EMF richness was not significantly different among host trees (nested anova, P = 0.383). The result was the same when each host species/site was separately subjected to one-way anova (P > 0.05, after sequential Bonferroni correction). The most abundant and most frequent EMF was Cenococcum geophilum, which colonized 7.4% of root tips and was found in 40 trees. This species was the only EMF found on every host species in both sites. Russulaceae was the richest EMF family, followed by Cortinariaceae, Thelephoraceae, and Exidiaceae (Sebacinaceae; Table 2). Together, these families occupied more than 50% of the root tips.

Table 2.  Ectomycorrhizal fungal (EMF) communities on eight host species in two conifer–broadleaf mixed forests, Japan
Host species/siteEMF richness per host treea,bObserved EMF richnessShannon'sH′Simpson's 1/DBasidiomycotaAscomycotaUnidentified
BoletaceaeCortinariaceaeExidiaceaeRussulaceaeaThelephoraceaeaAmanitaLaccariaOtherscPezizalesC. g.dOtherse
  • EMF species were grouped according to their genera or families. Number of EMF species within the category and relative abundance in the parenthesis.

  • a

    Significant difference (P < 0.05) among host species/site was indicated by different letters.

  • b

    Mean of five trees without correction for the number of tips sampled.

  • c

    Naucoria, Atheliaceae, Clavulinaceae, Cantharellales and unidentified Agaricales.

  • dCenococcum geophilum.

  • e

    Phialophora, Tuber, Elaphomyces and Hymenoscyphus.

Secondary forest
Abies homolepis 6.4 ± 1.9ab 272.711.11 (10.8) 5 (26.4) 3 (11.7) 2 (6.7b) 6 (22.6ab)00 02 (1.1)(11.3) 2 (2.9) 5 (6.5)
Betula maximowicziana 6.4 ± 3.1ab 272.913.21 (10.0) 3 (18.8) 1 (8.7) 4 (14.7ab) 1 (4.4b)00 3 (6.5)2 (3.4)(3.5) 3 (7.9) 8 (22.1)
Carpinus japonica10.2 ± 2.9ab 423.422.61 (4.5) 4 (9.1) 6 (13.3) 9 (26.7ab) 2 (2.6b)02 (3.0) 4 (14.1)3 (2.7)(9.2) 5 (9.9) 5 (4.9)
Fagus crenata12.0 ± 3.2a 473.526.12 (3.6) 4 (2.9) 5 (15.6) 5 (15.0ab) 4 (6.9b)1 (1.3)2 (1.8) 5 (13.2)2 (1.5)(3.6) 6 (9.9)10 (24.7)
Quercus crispula 9.6 ± 2.5ab 393.220.03 (10.8) 5 (6.8) 4 (13) 5 (18.5ab) 4 (11.5b)02 (1.7) 2 (9.6)2 (5.5)(7.9) 3 (5.6) 8 (9.0)
Primary forest
Abies homolepis 8.8 ± 3.3ab 363.115.83 (7.3) 3 (2.7) 6 (10.5) 5 (4.9b) 7 (48.8a)1 (2.3)2 (5.2) 2 (5.1)0(3.1) 3 (5.1) 3 (5.0)
Tsuga sieboldii 7.4 ± 2.9ab 323.015.61 (2.6) 3 (10.2) 2 (11.1) 5 (8.1b) 6 (17.0ab)2 (1.7)0 1 (4.9)0(4.0) 4 (10.5) 7 (29.9)
Betula maximowicziana 5.8 ± 1.5b 232.711.30 6 (14.2) 1 (0.1) 5 (58.4a) 1 (4.5b)00 01 (5.6)(2.2) 2 (4.3) 6 (10.8)
Betula grossa 5.4 ± 3.4b 232.813.20 4 (9.8) 2 (10.1) 6 (34.1ab) 1 (0.5b)1 (1.3)1 (3.4) 02 (1.0)(9.2) 2 (21.4) 3 (9.1)
Fagus crenata 8.8 ± 1.3ab 332.911.90 3 (5.1) 3 (12.4) 9 (25.0ab) 3 (10.9b)1 (2.5)0 2 (2.1)1 (0.5)(18.9) 4 (8.5) 6 (14.1)
Fagus japonica 9.6 ± 1.5ab 363.218.02 (15.4) 4 (2.9) 4 (10.7) 5 (18.5ab) 3 (10.9b)01 (1.4) 2 (7.1)3 (4.2)(5.3) 3 (10.6) 8 (12.9)
Quercus crispula 8.2 ± 2.8ab 333.117.62 (9.8) 2 (5.1) 6 (10.2) 5 (23.8ab) 1 (2.1b)01 (1.5) 5 (8.8)1 (1.6)(11.5) 1 (1.4) 8 (24.1)
Totals in secondary forest 1214.460.15 (7.9)15 (12.8)10 (12.5)20 (16.3)11 (9.6)1 (0.3)4 (1.3)10 (2.9)6 (2.9)(7.1)10 (10.0)28 (13.4)
Totals in primary forest 1374.352.25 (5.0)17 (7.1)14 (9.3)24 (24.7)14 (13.5)5 (1.1)2 (1.6) 9 (8.2)5 (1.9)(7.7)12 (6.9)29 (15.1)
Totals in both sites 205  7251735195517811749

Site effects on EMF communities

The observed EMF richness was 121 species in the secondary forest and 137 species in the primary forest. Because the species-area curves had not begun to level off (Fig. 1), these richness values were far lower than the potential EMF richness. Species richness was estimated to be 214, 191 and 235 in the secondary forest and 243, 215 and 265 in the primary forest, using Chao2 and Jackknife1 and 2, respectively. When a rarefaction analysis was applied, the observed richness in the primary forest was 112, and that estimated using Chao2 and Jackknife1 and 2 was 219, 180 and 225, respectively. These values were similar to those in the secondary forest. EMF richness per host tree in the four host species studied at both sites (A. homolepis, B. maximowicziana, F. crenata and Q. crispula) was 8.6 ± 3.5 and 7.9 ± 2.5 species in the secondary and primary forest, respectively, and this difference was not significant (anova, P = 0.472). Shannon's H′ and Simpson's 1/D were 4.52 and 83.6 in the secondary forest, and 4.61 and 78.8 in the primary forest, respectively.

Figure 1.

Species–area curves for ectomycorrhizal fungal (EMF) species in the secondary forest (open circles) and the primary forest (closed circles) in central Japan. The mean cumulative number of EMF species in each site is plotted after 50 randomizations using estimates version 7.5.

To assess the site specificity of each EMF species, data for the four commonly studied species were pooled within each site and subjected to Fisher's exact tests. No EMF species showed significantly biased occurrence at either site (P > 0.05). The mean EMF community similarity between host species at different sites was 0.122 (pairs of the same species were excluded), which was similar to that within the sites (0.110).

Comparisons of EMF communities among host species

At both sites, the observed EMF richness on Betula species was lowest, whereas that on Fagaceae species was relatively high (Fig. 2). The estimated richness showed a similar tendency. The species richness on A. homolepis was high in the primary forest, but as low as that on B. maximowicziana in the secondary forest, where A. homolepis trees were small and infrequent (Table 1). Shannon's and Simpson's indices of EMF diversity in the secondary forest were lowest for A. homolepis (Table 2).

Figure 2.

Ectomycorrhizal fungal (EMF) richness on eight host species in the secondary and the primary mixed conifer–broadleaf forest. Observed EMF richness is indicated by open circles. Estimated EMF richness using Jackknife1 (bold bars) with 99% confidence intervals (Heltshe & Forrester, 1983) and Jackknife2 (squares) was calculated for each host species after 50 permutations without replacement using estimates version 7.5. Ah, Abies homolepis; Ts, Tsuga sieboldii; Bm, Betula maximowicziana; Bg, B. grossa; Cj, Carpinus japonica; Fc, Fagus crenata; Fj, Fagus japonica; Qc, Quercus crispula.

In general, Morisita–Horn similarity indices (Table 3) were high for host pairs from the same genus or family, especially for Betula or within Fagaceae. The simple Mantel test revealed a significant negative correlation between EMF similarity and taxonomic distance (Kendall's τb = −0.338, P < 0.001; Fig. 3). Mantel's r generally decreased from positive to negative with increases in taxonomic distance and deviated significantly from zero in congeneric pairs (positive) and conifer–broadleaf species pairs (negative; P < 0.05 after sequential Bonferroni correction; Fig. 3).

Table 3.  The number of shared ectomycorrhizal fungi (upper triangle matrix) and Morisita-Horn simirality index (lower triangle matrix) between host species
SitesHost speciesSecondary forestPrimary forest
AhBmCjFcQcAhTsBmBgFcFjQc
  1. Ah, Abies homolepis; Ts, Tsuga sieboldii; Bm, Betula maximowicziana; Bg, B. grossa; Cj, Carpinus japonica; Fc, Fagus crenata; Fj, Fagus japonica; Qc, Quercus crispula.

Secondary forestAh 39101010745 9 7 7
Bm0.13 7 7 6 3354 2 6 3
Cj0.150.10 15 9 5672 7 8 6
Fc0.110.070.16 14 7885121414
Qc0.130.080.13 0.25  7565 610 6
Primary forestAh0.160.030.05 0.05 0.09 635 6 7 7
Ts0.090.030.05 0.08 0.07 0.04 41 9 6 3
Bm0.030.090.09 0.12 0.09 0.010.08 4 6 7 5
Bg0.110.170.09 0.10 0.15 0.060.030.30  6 4 7
Fc0.220.040.17 0.29 0.22 0.130.100.070.20 1010
Fj0.090.170.20 0.29 0.19 0.090.100.100.16 0.21 11
Qc0.150.050.22 0.23 0.15 0.100.070.060.19 0.32 0.30 
Figure 3.

Mantel autocorelogram of ectomycorrhizal fungal similarity based on the Morisita–Horn similarity index (Table 3) against taxonomic distance (see the Materials and Methods). Congeneric and confamilial pairs indicate comparisons between hosts in the same genus and family, respectively. The overall correlation was significantly negative (Kendall's τb = −0.338, P < 0.001). Asterisks indicate significant deviations from zero (P < 0.05, after sequential Bonferroni correction).

Detrended correspondence analysis showed that EMF communities within the same host family were located in clearly distinguishable clusters along the primary axis (Fig. 4), which had an eigenvalue of 0.501 and explained 13.2% of the variance in the data. By contrast, hosts from different sites were somewhat clustered along the secondary axis, which had an eigenvalue of 0.375 and explained 9.9% of the variance. These two axes together explained 23.1% of the variance in the data.

Figure 4.

Detrended correspondence analysis of ectomycorrhizal fungal communities on eight host species in the secondary (closed circles) and the primary (open circles) forest. Ah, Abies homolepis; Ts, Tsuga sieboldii; Bm, Betula maximowicziana; Bg, B. grossa; Cj, Carpinus japonica; Fc, Fagus crenata; Fj, Fagus japonica; Qc, Quercus crispula.

Host specificity

Because no EMF species showed significantly biased occurrence toward the secondary or primary forest, the frequencies of each EMF species in both sites were combined to assess EMF host specificity. Of the 37 EMF species that appeared on both conifer and broadleaf trees, 19 species (24.2% in mycorrhizal abundance) were found on three host families, six species (3.7%) were common on Pinaceae and Betulaceae, and 12 species (8.7%) were shared on Pinaceae and Fagaceae. Twenty-four EMF species (16.5% of mycorrhizas) were found on both Betulaceae and Fagaceae, but not on Pinaceae (Fig. 5). In total, generalists that were associated with two or more host families accounted for 29.8% of the EMF species, but 53.0% of the total EMF abundance (Fig. 5).

Figure 5.

Percentage occurrence of ectomycorrhizal fungal (EMF) species with reference to their host ranges. The relative abundance of mycorrhizas in each host range group was calculated after pooling the relative abundance data for the same host species in both sites (i.e. secondary and primary forest). To test the bias of the observed proportion of each host range class, observed values were compared with values obtained after 1000 randomizations. In a randomization, each EMF species was randomly assigned to the number of host trees equal to its observed frequency. Asterisks indicate significantly biased values (P < 0.05). Mean numbers of EMF species obtained from randomizations were shown in the parentheses.

Of the 14 EMF species observed on only one family, two, four and eight were found on Pinaceae, Betulaceae, and Fagaceae, respectively. In total, 130 EMF species were found on a single genus, and the number of EMF species detected on single host families represented 47.0% of mycorrhizas, although 112 EMF species occurred on only single trees.

To demonstrate EMF host specificity with statistical evidence, we used Fisher's exact test. First, the frequency of each EMF species was compared between host species within the same genus. Only one EMF, Sebacina sp. 1, was significantly more frequent on F. crenata than on F. japonica (P = 0.044; Table 4). However, this species was found on other host species (four, one and two trees of Quercus, Betulaceae and Pinaceae, respectively).

Table 4.  Ectomycorrhizal fungal (EMF) species showing biased occurrence between host species, genera or among families
EMF taxon comparedFisher's PHost taxa (frequency)
  1. Total trees studied: n = 5 for F. japonica, and Carpinus; n = 10 for F. crenata; n = 15 for Betula and Pinaceae; n = 20 for Betulaceae; n = 25 for Fagaceae. Frequency is the number of trees in which a given EMF species was found.

  2. When significant difference was observed among host families, Fisher's exact tests were applied for all pairs of the families. Different letters indicate the significant difference between host taxa. ns, pair-wise Fisher's exact test was not significant (P > 0.05) in any pairs of the families.

Between species within the same genus
 Sebacina sp. 10.044Fagus crenata (6)a, Fagus japonica (0)b
Between genera within the same family
 Agaricales sp. 40.009Betula (0)a, Carpinus (3)b
 Piloderma sp. 30.032Betula (1)a, Carpinus (3)b
Among families
 Boletus sp. 50.013Pinaceae (3)a, Betulaceae (0)ab, Fagaceae (0)b
 Cortinarius sp. 90.047Pinaceae (0)ns, Betulaceae (3)ns, Fagaceae (0)ns
 Lactarius sp. 10.047Pinaceae (0)ns, Betulaceae (3)ns, Fagaceae (0)ns
 Sebacina sp. 10.016Pinaceae (2)ab, Betulaceae (1)a, Fagaceae (10)b
 Thelephoraceae sp. 30.031Pinaceae (3)a, Betulaceae (1)ab, Fagaceae (0)b
 Tomentella sp. 100.035Pinaceae (1)ab, Betulaceae (0)a, Fagaceae (6)b

Next, EMF frequencies were compared between host genera within the same family. Two EMF species (Agaricales sp. 4 and Piloderma sp. 3) were more frequent on Carpinus than on Betula (P = 0.009 and P = 0.032, respectively). Because Agaricales sp. 4 was not confirmed in other host families, the species is considered to be specific to Carpinus. By contrast, Piloderma sp. 3 was found in two trees of Fagaceae, indicating weak host aversion to Betula.

Third, EMF frequencies were compared among the three host families. We excluded Agaricales sp. 4 from the analysis. Significantly biased occurrence was found (P < 0.05) on Pinaceae for two EMF species (Boletus sp. 5 and Thelephoraceae sp. 3), on Betulaceae for two EMF species (Cortinarius sp. 9 and Lactarius sp. 1), and on Fagaceae for two EMF species (Tomentella sp. 10 and Sebacina sp. 1; Table 4). Of these, one species (Boletus sp. 5) and two species (Cortinarius sp. 9 and Lactarius sp. 1) are likely specific to Pinaceae and Betulaceae, respectively, because they were not confirmed on other host families. The remaining three species showed significant host preference, but were not specific to a family, showing rare occurrence in other host families. Among them, Sebacina sp. 1 may exhibit preference for a particular host species in Fagaceae, because it showed host preference to Fagaceae but aversion to F. japonica.

Of 55 EMF species that appeared in more than two host trees, eight showed significantly biased occurrence. Note that an EMF species had to be detected in at least three trees to demonstrate significant host specificity. Because 150 EMF occurred in less than three hosts, the host preferences of these EMF remain unknown.

Discussion

Host specificity

Previous studies based on sporocarp surveys (Lee & Kim, 1987; Molina et al., 1992; Newton & Haigh, 1998) have revealed the ecological specificity and host ranges of a variety of EMF. However, the absence of sporocarps does not necessarily indicate a lack of colonization. Moreover, sporocarp approaches are problematic when different host species are in close vicinity or occur in different field conditions. Contrary to sporocarp approaches, molecular approaches can be applied to EMF on individual host species in close vicinity within the same site (Gardes & Bruns, 1993; Horton & Bruns, 1998).

A number of studies have examined EMF host specificity at various host-taxon levels using molecular approaches (species, Walker et al., 2005; genus, Horton & Bruns, 1998; Cullings et al., 2000; family, Kennedy et al., 2003; Richard et al., 2005; Nara, 2006). Each of these studies examined only one level of host specificity (species, genus or family) using two or three host species. EMF species exhibit biased occurrence between host species belonging to different genera within the Pinaceae (Horton & Bruns, 1998) and among host species from different families (Nara, 2006). However, these studies cannot be used to determine which host levels (i.e. species, genus or family accounted for these specific occurrences). To better understand EMF host specificity, we should examine EMF occurrence at multiple host-pair taxonomic levels. To our knowledge, ours is the first study to examine EMF host specificity at various host taxonomic levels within a site.

Some sporocarp studies have suggested that host specificity at higher levels of the host taxon (i.e. genus or family) is relatively common (Molina et al., 1992; Newton & Haigh, 1998; Massicotte et al., 1999). A similar tendency was also observed in our study (Table 4). This may indicate that biased occurrence of EMF is more common at the host family level than at the host species level. However, this pattern may also be confounded by differences in statistical power among the host taxa compared. Within the genus level, comparisons were made on as few as 15 observations, requiring a frequency of at least 60% of samples within a species to achieve significance, while among-family comparisons were made on a total of 60 observations, requiring a frequency of 20% of samples within a family to achieve statistical significance.

Significantly biased EMF occurrence did not always indicate specificity to a host taxon, because a low level of colonization was sometimes observed on other host taxa. However, a preference for a certain host taxon over other taxa would still affect EMF communities in forest ecosystems.

Host effects on EMF communities

Newton & Haigh (1998) observed similar EMF sporocarp communities among genera within the Fagaceae and Pinaceae (except Abies), although host taxonomic closeness did not always lead to EMF sporocarp community similarity. Because sporocarp communities are poor representatives of below-ground EMF communities (Richard et al., 2005), sporocarp studies are insufficient to reveal the relationship between host taxonomic status and below-ground EMF communities. In the present study, we clearly showed that taxonomically close host species harbor similar EMF communities (Table 3, Figs 3 and 4). What could cause the positive correlation between host taxonomic closeness and EMF similarity? One factor may be the evolutionary history of EMF–host associations. Molecular phylogenetic studies of some fungal genera have demonstrated that closely related EMF species that evolved from a common ancestor show specificity to taxonomically related hosts (Kretzer et al., 1996; Hibbet et al., 2000; Grubisha et al., 2002; den Bakker et al., 2004). If such a fungus–host relationship is common in EMF, similar EMF communities will be found in closely related host species.

In the DCA (Fig. 4), B. maximowicziana, an early successional species, was located near the left end of the primary axis, followed by B. grossa and C. japonica, two mid-successional species (Koike, 1988; Seiwa et al., 2006). Quercus crispula, a mid- to late-successional species, appeared next to C. japonica, followed by Fagus and Pinaceae species, all of which are shade-tolerant late-successional species (Hashizume et al., 1993). Thus, the primary DCA axis corresponds well with the successional status of tree species, indicating that host successional status may partly determine the EMF community.

The successional status of host species may also affect the number of EMF species associated with an individual host species. The observed EMF richness was lowest on early successional B. maximowicziana at both sites. By contrast, most of the late-successional species (Fagaceae and Pinaceae) were associated with higher numbers of EMF species. Because both plant productivity and seedling establishment may be greater with more diverse EMF communities (Baxter & Dighton, 2001; Jonsson et al., 2001; Dickie et al., 2002), diverse EMF communities are advantageous to late-successional species, which dominate in the competitive conditions of mature forests.

Another host effect on the EMF community is the time elapsed after the establishment of the host species. The EMF richness associated with A. homolepis was lower in the secondary forest, where A. homolepis individuals were small and infrequent, than in the primary forest (Fig. 2). Given that A. homolepis is usually associated with a variety of EMF species (Lee & Kim, 1987; Kranabetter et al., 1999), as in the primary forest studied here, the EMF richness on this host in the secondary forest appears to be lower than its potential. Moreover, the EMF community of A. homolepis in the secondary forest was placed closer to EMF communities of broadleaf trees than that of A. homolepis in the primary forest in the DCA plot (Fig. 4). These findings may indicate that A. homolepis initially shares EMF with broadleaf trees and becomes colonized by specific EMF many years after its establishment. This view is supported by data that indicate the proportions of EMF species with a broad host range (found in both angiosperms and gymnosperms) and a narrow host range (found only in Pinaceae). We observed 21 broad-host-range and six narrow-host-range fungi in the secondary forest, whereas 15 broad-host-range and 21 narrow-host-range fungi were found in the primary forest.

EMF diversity in mixed conifer–broadleaf forests

The total observed species richness was high at both sites, with 121 and 137 EMF species in the secondary and primary forest, respectively. Although the comparison of EMF richness among studies is difficult because of differences in sampling strategy, the EMF community described here, containing 205 observed species (overall total), is the richest among studies on below-ground EMF (Luoma et al., 1997; Gehring et al., 1998; Jonsson et al., 1999; Cullings et al., 2000; Grogan et al., 2000; Peter et al., 2001; Jumpponen, 2003; Smith et al., 2004; Horton et al., 2005; Moser et al., 2005; Richard et al., 2005; Walker et al., 2005; Korkama et al., 2006; Nara, 2006; Tedersoo et al., 2006; Toljander et al., 2006). Moreover, the species–area curves did not reach an asymptote at the maximum sample size; total diversity estimates ranged from 315 to 387 species, implying that nearly 50% of species may have remained undetected (Fig. 1).

The EMF host preference and specificity were found for a considerable portion of the EMF species, suggesting that the presence of a variety of host taxa contributes to such EMF hyperdiversity in mixed conifer–broadleaf forests. This finding supports the hypothesis that host diversity contributes to EMF diversity (Nantel & Neumann, 1992; Kernaghan et al., 2003). In addition, EMF communities differed significantly among codominant tree species, indicating EMF spatial heterogeneity. The EMF colonization on seedlings is related to the surrounding EMF communities (Cline et al., 2005); dissimilar EMF communities can have different effects on different host species (Jonsson et al., 2001). Thus, the heterogeneity of the EMF community may contribute to the establishment of various host species. Ectomycorrhizal fungal hyperdiversity in mixed conifer–broadleaf forests may be maintained by host diversity, and the coexistence of various host species may in turn be supported by diverse and spatially heterogeneous EMF communities.

Acknowledgements

This research was supported by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists (No. 0410304) and Grants-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Ancillary