• This study compares DNA and culture-based detection of fungi from 15 ericoid mycorrhizal roots of salal (Gaultheria shallon), from Vancouver Island, BC Canada.
• From the 15 roots, we PCR amplified fungal DNAs and analyzed 156 clones that included the internal transcribed spacer two (ITS2). From 150 different subsections of the same roots, we cultured fungi and analyzed their ITS2 DNAs by RFLP patterns or sequencing. We mapped the original position of each root section and recorded fungi detected in each.
• Phylogenetically, most cloned DNAs clustered among Sebacina spp. (Sebacinaceae, Basidiomycota). Capronia sp. and Hymenoscyphus erica (Ascomycota) predominated among the cultured fungi and formed intracellular hyphal coils in resynthesis experiments with salal.
• We illustrate patterns of fungal diversity at the scale of individual roots and compare cloned and cultured fungi from each root. Indicating a systematic culturing detection bias, Sebacina DNAs predominated in 10 of the 15 roots yet Sebacina spp. never grew from cultures from the same roots or from among the > 200 ericoid mycorrhizal fungi previously cultured from different roots from the same site.
Roots of salal (Gaultheria shallon Pursh) a western North American shrub, are usually heavily colonized by mycorrhizal fungi (Xiao & Berch, 1996). In this study, we compare the fungi in an individual mycorrhizal root as detected by direct DNA extraction and analysis, with the fungi as detected by their growth in pure culture. Failure of bacterial cells to grow under standard cultural conditions has been reported repeatedly, and as reviewed by Pace (1997), 99% of bacteria observed microscopically are missed in standard culturing experiments. Similar losses may take place among fungi cultured from natural environments. Instead of using the more traditional isolation in pure culture to detect fungi from the roots of a grass, Arrhenatherum elatius (L.) P. Beauv., Vandenkoornhuyse et al. (2002) extracted total DNA and then used fungal specific PCR primers to amplify and sequence the 18S ribosomal RNA genes of root-associated fungi. In contrast to most culture-based studies that detect primarily fungi in the Ascomycota, this sequence-based approach found that about half of the fungi were in the Basidiomycota or Zygomycota, rather than the Ascomycota and several of the sequences could not be classified into a familiar fungal order (Vandenkoornhuyse et al., 2002).
Like the roots of grasses, ericoid mycorrhizae offer a relatively simple biological environment where fungal DNA sequence diversity can conveniently be assayed. No one had previously studied the diversity of fungal DNAs from these roots. However, the DNAs from the roots could be matched against an extensive database of sequences from fungi cultured from ericoid and epacrid mycorrhizae from around the world (McLean et al., 1999; Bergero et al., 2000; Berch et al., 2002; Cairney & Ashford, 2002).
Our interest in ericoid mycorrhizal fungi has its origin in a biological problem, but is also linked to an economic problem. Conifer seedlings used in reforestation of some sites previously occupied by cedar-hemlock (Thuja plicata Donn./Tsuga heterophylla (Raf.) Sarg.) forests on Vancouver Island, British Columbia, Canada turn yellow and fail to grow while the ericaceous shrub salal thrives. Colonized with fungal mutualists, the roots of the salal may be able to access nutrients, especially nitrogen, that may be otherwise unavailable (Read, 1991; Perotto et al., 2002). To contribute to an understanding of the biology underlying the conifer growth check problem, Xiao & Berch (1996, 1999), Monreal et al. (1999), and Berch et al. (2002) have been investigating the diversity and functioning of some of the mycorrhizal fungi collected from salal roots from the Salal Cedar Hemlock Integrated Research Project (SCHIRP) trial site at Port McNeil, Vancouver Island, British Columbia. Through their series of studies, a succession of different techniques have revealed different mycorrhizal fungi.
Using culturing and morphological identification, Xiao & Berch (1996) recognized four groups of fungi among over 200 mycorrhizal fungal cultures from salal from the SCHIRP site. Two groups could be identified by their sporulation patterns, while two remained unidentified and were nicknamed ‘Unknown 1’ and ‘Unknown 2’. Monreal et al. (1999) determined ITS2 sequences from all fungi that had been confirmed as ericoid mycorrhizal symbionts, including ‘Unknown 1’ and ‘Unknown 2.’ Among additional fungi cultured from the Vancouver Island SCHIRP site, Monreal et al. (1999) found isolates with sequences matching the well-characterized ericoid mycorrhizal fungus Hymenoscyphus ericae. Lynn Sigler (personal communication) later used conidial morphology to confirm Monreal's isolates as H. ericae.
Berch et al. (2002) discussed the phylogenetic relationships of sequences from cultured or cloned fungi from about 800 salal root tips from the Vancouver Island SCHIRP site and noted which of the cultured fungi produced mycorrhizae in resynthesis experiments with salal. Berch et al. (2002) discovered that isolates similar to most of the ericoid mycorrhizal fungi grown in pure culture from salal roots from our Vancouver Island SCHIRP site were also found in Europe or Australia. All the ericoid mycorrhizae from pure cultures from salal were from ascomycetous fungi. However, Berch et al. (2002) also reported that the predominant fungal clones from PCR amplifications of DNA extracts from salal mycorrhizae were from basidiomycetes closest to inconspicuous jelly fungi in the genus Sebacina.
Our objective in this study was to analyze the patterns of fungal diversity, as detected by culturing and by cloning, at the scale of individual roots. We planned to analyze the fungal DNAs that we could detect by direct DNA extraction and PCR amplification from segments of highly colonized, mycorrhizal salal roots. For comparison with these fungal DNAs, we would culture fungi from root segments flanking the segments used for direct DNA extraction, mapping the original position of each segment from each root. From the mapped positions of root segments used for direct DNA extraction and for culturing, we could reconstruct the original position of each fungal colony or DNA in the original root and track which root segments gave no fungal colonies. To help distinguish other saprophytes or parasites from likely mycorrhizal partners, we planned to test the fungi in culture for their ability to form hyphal coils in the epidermal and cortical cells in resynthesis experiments with salal. Discrepancies between the fungi detected as DNAs and the fungi that grew in pure culture would suggest a possible systematic detection bias against some fungal species. Using a combination of detection methods should optimize the chances of identifying the entire suite of mycorrhizal fungi occupying root systems.
Materials and Methods
A total of 15 mycorrhizal hair roots of salal (G. shallon) were selected for detailed analysis, from samples collected from the SCHIRP study site located between Port McNeil and Port Hardy on northern Vancouver Island, British Columbia (50°60′ N, 127°35′ W). The first sample was collected in October 1998 and samples two and three were collected from plants about 30 m apart in January 1999. Each sample, collected with a shovel, consisted of about 500 g of soil, rhizomes and roots. Because salal reproduces with rhizomes, we could not be sure whether samples were from the same or different plants. Samples were kept in plastic bags on ice, transported to the University of British Columbia, and held at 4°C until cleaning. Small roots were removed from the rhizomes, soaked in cool tap water and washed gently to remove soil and plant debris. Using light microscopy, roots were screened for healthy appearance (turgid cells and light color) and for fungal colonization. From these apparently healthy, colonized roots, sections with a total contiguous length of 4 cm were selected for study. Seven roots (#1011–1116) were selected for analysis from sample one from October. From the January collections, from sample two, two roots (#1302–1322) were selected, and from sample three, six roots (#1442–1542) were selected (Fig. 1). Roots 1482 and 1502 were from the same rhizome as were roots 1522 and 1542. Each of the other roots was from a different rhizome. Each root was sterilized in 30% hydrogen peroxide for 1 min and washed three times in sterile distilled water. The 4-cm mycorrhizal root segments were further subdivided into 11 pieces (Fig. 1). The largest piece, a 2-cm section from the middle of the 4 cm root was used for direct DNA extraction and PCR amplification. Ten 2 mm pieces, five from each end of the 2 cm middle section, were set on potato dextrose agar for culturing. One of the pieces included the root tip. Because the hair roots were highly branched, multiple fine lateral root tips were present along the main root axis and some of the segments used for cloning or culturing also included a smaller, lateral root tip.
The original location of each 2 mm piece was recorded to allow reconstruction of the original fungal distribution along each root. Genomic DNA from each fungus that grew in culture was extracted and amplified with the fungal specific primer ITS1-F (Gardes & Bruns, 1993) or universal primers ITS4 or TW13 (http://plantbio.berkeley.edu/bruns/primers.html28s). The PCR product of each isolate was digested with MspI. Some of the products were also digested with RsaI. The fungal isolates were grouped by RFLP pattern.
Fungal isolates were tested for ability to form hyphal coils inside of the root cells of salal in Petri-dish growth chambers. Fungal isolates that did form coils inside cells of living roots were termed ‘mycorrhizal.’ (Studies to confirm mycorrhizal status would have required additional evaluation of whether the fungus benefited the plant.) Seeds of salal were sterilized in 30% hydrogen peroxide for 1 min and placed on water agar for germination. Seedlings were transplanted to a low nutrient medium, modified Melin–Norkrans agar MMN (Xiao & Berch, 1992) in plastic Petri dishes and inoculated with a fungal culture. Roots were extracted from growth chambers and stained with aniline blue. Colonization of the root cortical cells was confirmed with light microscopy. Fungal isolates that formed hyphal coils within root cells were re-tested by culturing the fungus from sterilized roots of colonized plants and inoculating the fungus on new plants.
Direct DNA extraction, PCR amplification, cloning, and sequence analysis
Total genomic DNA was extracted from the middle 2 cm segments of 15, highly colonized, sterilized field collected mycorrhizal root pieces using a DNeasy® Plant Mini Kit (QIAGEN Inc., Mississauga Ontario, Canada). Genomic DNA from sterile leaves from salal was extracted and used as a negative control for each DNA extraction and PCR amplification. Total fungal DNA was amplified with ITS1-F, in combination with either ITS4 or TW13. PCR products from the ITS and large subunit (LSU) ribosomal RNA gene regions were cloned using InvitrogenTM TOPO TA Cloning® kit (Invitrogen Canada Inc. Burlington, ON, Canada) following the manufacturer's instructions. From the first root, we selected approximately 50 colonies with plasmid inserts for further analysis. For subsequent roots, we selected 20 colonies with inserts for further analysis when possible. Some roots gave relatively little fungal PCR product, and as a result, fewer transformants were available for analysis. Plasmid DNAs extracted using a QIAprep® kit (QIAGEN Inc., Mississauga Ontario, Canada) were digested with EcoR I to verify that an insert of the correct size was present, and the cloned insert was amplified by PCR. MspI and in some cases, RsaI digests of the PCR amplified inserts were used to group similar isolates based on their RFLP patterns.
To further characterize cultured fungi or cloned DNAs, we selected fungal DNAs from each RFLP group and used the primer ITS3 to generate single-stranded ITS2 sequences using Applied Biosystems AmpliTaq DyeDeoxyTM terminator kit following the manufacturer's instructions (PE Applied Biosystems, Foster City, CA, USA). The ITS2 sequences were subjected to BLAST searches, and aligned with similar sequences from GenBank, using ClustalW PPC (Higgins et al., 1992) followed by manual adjustment with SeqApp (available through ftp://iubio.bio.indiana.edu/molbio/seqapp/).
Given the extensive database of ericoid and epacrid mycorrhizal fungal ITS sequences, we had expected the ITS2 clone sequences to find matches among GenBank's ITS2 sequences. However, many of the cloned sequences failed to find matches. To provide a more highly conserved sequence region for phylogenetic identification of the clones, we switched from amplifying and sequencing the ITS2 regions alone to amplifying an approximately 1200 bp region that included about 500 bp of the 5′ end of the LSU gene as well as the ITS2 region. We then used sequences including the LSU region for BLAST searches and to cluster clones in a preliminary phylogeny. For phylogenetic analysis, representatives of groups of isolates were selected for more extensive sequencing, using primers ITS3, ITS4, cTB6 and TW13 (http://plantbio.berkeley.edu/bruns/primers.html28s) to obtain double-stranded sequences across the 1200 bp sequence region. The LSU and ITS plus LSU fragments were also aligned with sequences from BLAST searches from GenBank. To help track and sort sequences, each was assigned a number based on its original position in a preliminary alignment and phylogeny. GenBank accession numbers for all isolates in our alignments are listed in Table 1.
Table 1. Fungal isolates or clones in phylogenetic analyses
‘LSU’ indicates that approximately 500 bp of the 5′ end of the large subunit ribosomal RNA gene was available for analysis.
‘Clone’ indicates a sequence from fungal PCR product directly amplified from ericoid mycorrhizal roots of salal. ‘Culture’ indicates a sequence from a fungus that grew in pure culture following isolation from ericoid mycorrhizal roots. Sequences taken from a database are designated ‘GenBank’.
Accession numbers beginning with ‘UAMH’ designate cultures in the University of Alberta Microfungus Collection and Herbarium. All others are GenBank accession numbers.
Some ITS2 sequences were too different to be aligned. Similar subsets of sequences were aligned in blocks, and gaps were inserted to the other, more distant sequences. In this way, for example, ITS sequences of ascomycetes in the Leotiomycetes were aligned with one another, but the basidiomycete sequences were shifted to a different alignment block instead of being aligned with or compared to the Leotiomycetes sequences. Some portions of some of the ITS2 basidiomycete sequences showed no obvious sequence similarity to any of the other sequences and were excluded from all analyses. The complete alignment is available from TreeBASE (accession # S905).
The complete alignment, assembled to help cluster similar sequences, included ITS2 data from most taxa, and about 500 bp of sequence the 5′ end of the LSU rDNA for representative sequences of most phylogenetic groups (Table 1, Fig. 2). The alignment included diverse sequences from fungi in the Ascomycota and Basidiomycota, as well as sets of numerous, almost identical sequences. paup 4.0b10 (PPC) (Swofford, 1999) was used for all analyses. To show similarity among the 191 sequences (ITS2 and/or LSU) in our alignment, we used parsimony bootstrapping without branch swapping and showed the resulting trees as phylograms, so that the similarity of isolates could be inferred from horizontal branch lengths. The complete 191 sequence data set provided little bootstrap support even to clusters of nearly identical sequences. Other kinds of analyses were not feasible for the complete data set because the sequences were numerous and diverse, because LSU data were not obtained for most of the isolates, and because the highly variable ITS2 regions lacked sufficient phylogenetic signal to resolve relationships.
For more critical phylogenetic analysis, sequences with only ITS2 data were excluded from the analysis, and the remaining sequences with LSU or with both LSU and ITS2 data were analyzed. We conducted a series of analyses, variously including only sequences from ascomycetes or only sequences from basidiomycetes, or subsets of the ascomycete or basidiomycete sequences in which numbers of taxa were reduced by excluding all but one to four representative sequences from groups of similar sequences. We used 500 replicated parsimony searches, without branch swapping, for one estimate of bootstrap support. When analyzing smaller numbers of sequences, we used 20 replicated heuristic parsimony searches with random addition of taxa to find the most parsimonious trees, and 500 parsimony heuristic searches, with tree-bisection reconnection branch swapping, for bootstrap analysis. For a neighbor-joining bootstrap analysis, we used distance matrices constructed assuming a general-time reversible model of sequence evolution, with the values for a proportion of invariable sites and for the gamma shape parameter estimated using likelihood and the parsimony bootstrap tree.
The 15 mycorrhizal hair roots of salal chosen for analysis were heavily colonized by fungal mycelia. All the segments chosen for analysis contained typical ericoid mycorrhizal hyphal coils. An average of 3.8 genetically different fungi were detected per root (Fig. 1). Roots from the three different soil samples showed similar patterns of diversity. From the DNA extracts of mycorrhizal salal roots, clones from the genus Sebacina predominated (Figs 1, 2 and 3). With BLAST searches as well as phylogenetic analysis (explained below), we found that none of the over 300 previously determined sequences from ericoid mycorrhizal fungi matched the Sebacina-like clones. The Sebacina-like PCR products were produced from 11 of 15 mycorrhizal roots (Fig. 1). From the PCR product of 10 of the 15 roots, most of the DNAs detected were of the Sebacina type (Fig. 1). The Sebacina-like DNAs prevailed among mycorrhizal roots collected in October and in January. Of the 90 sequenced fungal clones, 46 clustered phylogenetically with the ITS2 and LSU sequences of Sebacina vermifera (Figs 1, 2a and 3). Of the 66 clones characterized by RFLPs, 46 were also of the Sebacina type.
The second most common clones from mycorrhizal root DNA clustered in the Ascomycota genus Capronia (Fig. 2b). Eight of the 15 roots, including roots from both collection dates and from three samples, contained Capronia-like DNA (Fig. 1). Eighteen of the sequenced clones, and nine of the clones characterized by RFLP were of the Capronia type. The third most common clone matched H. ericae, and two roots yielded 14 clones (four sequenced and 10 analyzed by RFLP) of this fungus (Fig. 1). Other fungal clone types were less frequent. One root (1101) gave three clones with sequences similar to those of species of wood decay basidiomycetes in the genus Trechispora (Fig. 2a). A clone sequence (UBCtra1302.14) that appeared to be a chimera of a H. ericae ITS region and a LSU sequence from the Hypocreales was not included in analyses. Clone sequence #9.2 (UBCtra15222A) matched the sequence of known ericoid mycorrhizal fungus Oidiodendron maius (Fig. 2b). Eight clone sequences (not included in the fungal total) were from salal or other plants. The remaining clone sequences were widely scattered among the Ascomycota and Basidiomycota. Of these remaining sequences, clones from the same root usually represented genetically different fungi and clustered in different clades of the phylogram (Figs 1 and 2a,b).
Even though initial light microscopy indicated that mycorrhizal hyphae were abundant in all segments, fungi grew from only 39 of the 150 segments from the 15 roots (Fig. 1). Only 15 of the 39 cultured fungi entered salal cortical cells to form mycorrhizal coils in resynthesis experiments (Fig. 1). The Sebacina-like fungi that predominated among the directly amplified mycorrhizal DNAs were completely absent among the cultured fungi. Among the cultures, the most frequently isolated fungi had Capronia-like DNA sequences that matched the second most common sequence type from the DNAs directly amplified from mycorrhizal roots (Figs 1 and 2b). No sporulation was observed among these Capronia-like isolates. A total of eight of these Capronia-like cultures grew from segments of three different roots (while direct PCR amplification detected Capronia-like sequences in eight roots). Both direct PCR amplification and culturing detected the same Capronia-like fungi from only two roots. In resynthesis experiments, five of the Capronia-like isolates formed hyphal coils inside salal root cells. When the Capronia-like isolates failed to colonize salal root cells, it was usually because the fungus did not grow well enough to contact the plant. Four isolates from three roots were identified as H. ericae (Fig. 1). Both direct PCR amplification and culturing detected H. ericae in one of the roots. All isolates of H. ericae formed hyphal coils in salal root cells in in vitro synthesis experiments. Only two of the 39 cultures from the 15 root segments were from the Basidiomycota and ITS2 sequence of one of these (#173, UBCtra1092C) matched clone sequences from the same root (root 1101) and clustered with Trechispora (Fig. 2a).
Some mycorrhizal fungi that were frequently detected in previous studies from the same Vancouver Island SCHIRP site were absent or infrequent on the 15 roots emphasized in this study. Although we came across the common mycorrhizal species Oidiodendron maius once among the cloned DNAs, we did not find the species among any of the fungal cultures from the 15 roots. We did not detect the sterile and unidentified taxa ‘Unknown 1’ or ‘Unknown 2’ among our 15 roots. However, mycorrhizal isolates close to ‘Unknown 1’ (#2, UBCtra1256C) and ‘Unknown 2’ (#24, UBCtra1128C) were isolated from roots that were not used for cloning, although they were from the same soil samples as the 15 roots used for detailed analysis.
The simple clustering of sequences based on fast parsimony bootstrapping divided 191 sequences in our alignment between the Ascomycota and Basidiomycota (Fig. 2a,b). The alignment included sequences from our cultures and clones of ericoid mycorrhizae, along with other sequences from GenBank. While only two of the sequences from fungi cultured from the 15 ericoid mycorrhizal roots were from the Basidiomycota, most of the clones of PCR amplified DNAs were from the Basidiomycota (Fig. 2a,b). Most of the basidiomycete clone sequences were similar to one another and to sequences from the orchid mycorrhizal fungus Sebacina vermifera (Fig. 2a). The nodes clustering the Sebacina-like fungi in the Basidiomycota received no bootstrap support from the data set of all 191 sequences. To further investigate the relationship of our isolates to Sebacina and to other fungi in the Sebacinaceae, we excluded 20 Sebacina-like sequences for which we had only ITS2 data. We re-analyzed the remaining 26 Sebacina-like sequences along with two ascomycete outgroups, and all other basidiomycete LSU sequences from our alignment. Our salal isolates clustered together (united at Node A, Fig. 3) with 87% fast parsimony bootstrap support. Node B united salal clones with Sebacina vermifera sensu Warcup & Talbot, with 86% bootstrap support, and Node C united all the above isolates with Sebacina vermifera Oberw. (Fig. 3). All sequenced members of the Sebacinaceae are united at Node D, with 59% bootstrap support (Fig. 3). Only nodes C and D appeared in neighbor-joining bootstrap trees, and the reason seemed to be that some of the Sebacina-like sequences were short fragments with data missing from the 5′ end of the LSU gene. These short sequences tended to cluster at the base of Node C, rather than with the other salal Sebacina-like salal sequences. As a further experiment, four of the Sebacina-like sequences having full length ITS2 regions and 550 bp LSU fragments (#124 UBCtra1322.6; #150 UBCtra1011.6; #151 UBCtra1041.2; #153 UBCtra1.04) were included in an analysis along with the other basidiomycetes from the analysis for Fig. 3. The ascomycete outgroups and all other Sebacina-like isolates were excluded. These Sebacina-like sequences that remained in the analysis represented the range of different sequence types within the Sebacina-like cluster (Fig. 3). We analyzed the reduced number of taxa by parsimony heuristic searches with tree-bisection-reconnection (TBR) branch swapping, and 20 replicated searches with random addition of taxa and found two most parsimonious trees with a length of 1611. Nodes A, B, C and D were present in both, most parsimonious trees. We used 500 replicated parsimony bootstrap replicates, with both ITS and LSU data and with TBR branch swapping, and found the following support for nodes: A, 95%; B 93%; C 70%; and D 91% (not illustrated). Excluding the ITS data, support for nodes was: A 91%; B 84%; C 76% and D 51%. With neighbor-joining bootstrap analysis of the ITS and LSU data, assuming a GTR substitution model, an estimate that 20% of sites were invariable, and a gamma shape parameter estimate of .577, support for nodes was: A 86%; B 95%; C 84% and D 74%. Continuing with neighbor joining, excluding the ITS data, 22% of sites were estimated to be invariable, the gamma shape parameter was .504, and support for the four nodes of interest was: A 36%; B 57%; C 63% and D 44% (not illustrated).
Among the ascomycetes, the fast parsimony bootstrap including all 191 sequences revealed a cluster of sequences similar to H. ericae, united by node E (Fig. 2). A second group of sequences appeared to be nested in the genus Capronia (Nodes F, G, and H in Fig. 2). Again, none of these nodes received any bootstrap support when the entire range of sequences was included in the analysis. For a more rigorous analysis, we excluded the basidiomycetes and all sequences lacking LSU data. With the smaller number of sequences and 500 fast parsimony bootstrap replicates, node E, uniting the H. ericae isolates and clones, received 86% bootstrap support (Fig. 4). The monophyly of the Capronia-like samples from salal (node H) received 98% support and the node uniting Capronia villosa with the clones and isolate from salal (Node G) received 93% support (Fig. 4). Neighbor-joining bootstrapping, with a GTR model of substitution, and an estimated 35% of sites invariable with a gamma shape parameter of 0.81 provided 100% support to the H. ericae node E and 96% support to Capronia Node G. When the LSU fragments were analyzed alone (excluding the ITS2 region), bootstrap support for nodes E-H ranged between 53% and 100%, depending on the node and the kind of analysis (not illustrated).
Variation among Sebacina-like and Capronia-like clones
Several of the Sebacina sequences differed from one another (Figs 2a and 3). To minimize sequencing errors, each variable position in the DNA sequences was double-checked by comparing the electropherograms to the alignment. For all of these sequences, the quality of the electropherograms was good and sequencing ambiguities were not the source of the variation. For sequences #124, 150, 151 and 153, most of variable sites were confirmed from sequence in both directions.
If the variation was caused by enzymatic nucleotide misincorporation during the PCR amplification process, the errors should be randomly distributed and not repeated in different PCR reactions. Sequences from different roots originated from different PCR reactions and so PCR error would not be expected to lead to shared variation in different roots. To find out whether substitutions were shared among roots, we began with the 25 Sebacina-like ITS2 sequences. By eliminating identical sequences, we reduced the data set to 15 unique sequences and found all of the 53 possible, equally parsimonious trees, also of length 119, using a branch and bound search. We then used the ‘describe trees’ option in paup to map all nucleotide changes to the branches where they most parsimoniously would have occurred, for one of the 53 trees. Some nucleotide substitutions were shared across roots. For example, fungal sequences #150, 169 and 181, from roots 1502 and 1542, shared 10 nucleotide substitution characters that distinguished them from sequences of the other roots. Sequences from roots 1011, 1116, 1056 and 1041, and sequence #179 from root 1322 shared at least three characters that distinguished them from other sequences, including a second sequence, #124, from root 1322.
Sequences from a root were usually more similar to other sequences from the same root than would be expected if sequence variation had been randomly distributed within and among roots. The most parsimonious trees, constrained so that sequences from each root formed a monophyletic group, required 121 substitutions, only two more than the most parsimonious trees. The most parsimonious, unconstrained trees, and the most parsimonious, constrained trees, were all considerably shorter than the lengths of 100 000 random trees that varied in length from 153 to 226 steps.
Unlike the Sebacina-like sequences, the 18 Capronia-like sequences, including 11 sequences with ITS2 regions, were almost identical to one another (Figs 2b and 4). The most divergent sequence, #79 from root 1522, differed from the others at four positions in the ITS2 region. Explaining all the substitutions among the 18 sequences required 13 steps in parsimony analysis. Unlike the substitutions among the Sebacina-like isolates, substitutions among the Capronia-like sequences did not show any obvious phylogenetic structure. If the substitutions had a phylogenetic structure, then the most parsimonious trees would be much shorter than almost all random trees. Instead, many of the random trees (55 of 1000) were the same length as the most parsimonious, 13-step trees.
Sebacina-like sequences dominated among the directly amplified fungi
Finding that Sebacina-like DNA was present in 11 of the 15 mycorrhizal roots and that 92 out of the 156 cloned DNAs were of the Sebacina type suggests that Sebacina spp. regularly associate with salal roots at our Vancouver Island SCHIRP research site. If Sebacina spp. produced the most abundant fungal DNA in the 15 roots, they were probably also present in at least some of the 150 segments of the same roots that were used for culturing. However, no Sebacina-type fungus has ever been detected in a fungal culture from these or any other ericoid root segments, including the hundreds of other ericoid mycorrhizal fungal cultures from the SCHIRP site (Xiao & Berch, 1996; Monreal et al., 1999; Berch et al., 2002). Most of the ascomycetous mycorrhizal fungi reported from the SCHIRP research site have also been reported from other continents (Berch et al., 2002). Even if the Sebacina-like fungal involvement in ericoid mycorrhizae is similarly geographically widespread, it may be systematically overlooked in surveys of root fungi that detect fungi only when they grow in pure culture under standard conditions.
Other species in the genus Sebacina are probably mycorrhizal and the genus is geographically widespread (Warcup, 1988; Glen et al., 2002; Selosse et al., 2002a; Selosse et al., 2002b; Urban et al., 2003). Lacking a culture of the Sebacina species, we cannot test our fungi for the ability to form mycorrhizae. Most species of Sebacina do not grow readily on artificial media (R. J. Bandoni, pers. comm.). However, our clone sequences cluster phylogenetically with > 87% bootstrap support with Sebacina vermifera, one of the few Sebacina species that has been cultured and shown to be mycorrhizal with orchids and deciduous trees (Warcup, 1988). The LSU and ITS sequences of at least two other Sebacina-like fungi were detected as common ectomycorrhizal associates of eucalyptus (Glen et al., 2002). Selosse et al. (2002a) showed using DNA sequence data and electron microscopic evidence that a Sebacina species appeared to be a mycorrhizal partner of an achlorophyllose orchid, Neottia nidus-avis L. Rich. and of deciduous trees in France. Similarly, Urban et al. used DNA sequence data and electron microscopy to identify Sebacina spp. as ectomycorrhizal partners of deciduous trees from Austria.
As with the Sebacina-like species from mycorrhizae from eucalyptus (Glen et al., 2002) and from achlorphyllous orchids (Selosse et al., 2002b), DNAs of Sebacina-like fungi from salal show patterns of variation consistent with underlying natural variation. The differences in sequences from fungi from a single root could be due to heterogeneity in the sequences of the ribosomal repeats of a single nucleus; to allelic differences in dikaryotic nuclei of a hypha, or to the presence of different genets or different species in different individual mycelia. Enzymatic misincorporations of nucleotides during PCR amplification were not likely to have caused the substitutions shared among independently amplified fungal DNAs from different roots.
As in studies of bacterial cell viability in cultures, lack of fungal growth was the most frequent outcome of attempts to grow cultures from salal mycorrhizae. Of our 150 ericoid mycorrhizal root segments, 74% gave no slow-growing fungal cultures. This is consistent with results from Xiao and Berch (1996), who obtained cultures of ericoid mycorrhizal fungi from 16% of the 1120 salal mycorrhizal root tips they examined, even though 90% of the host roots’ cortical cells were colonized. Possibly because the original population of fungal cells in a root segment was high, or possibly because cultural conditions have already been optimized for recovery of the mycorrhizal fungi, the rates of recovery of fungi in pure culture were still 10 times the 1% recovery rate for bacteria. If the frequency of fungal individuals were proportional to the proportion of clones, then about half of the mycorrhizal root segments may have contained Sebacina species and our inability to culture these Sebacina species could have accounted for a substantial fraction of the missing fungal cultures.
Capronia species as possible mycorrhizal symbionts
In both direct PCR amplification and isolation in culture, Capronia-like fungi were common. Although Capronia-like fungi are not usually considered to be mycorrhizal, Bergero et al. (2000) repeatedly isolated a sterile fungus, ‘Sd2’ in Italy, from Erica arborea L. (Ericaceae) and from Quercus ilex L. (an oak) that falls phylogenetically among the Capronia isolates and Ramichloridium anceps, a related asexual state (Fig. 2b). Isolate ‘Sd2’ formed typical ericoid mycorrhizae in resynthesis experiments. In resynthesis experiments, our Capronia-like isolate clearly colonize the cortical cells of salal roots but whether the species are truly mycorrhizal requires physiological testing to determine whether the salal benefits from the fungal colonization.
Suggesting that the viable part of each fungal individual was small (at least after washing and sterilization), Capronia sp. never grew from adjacent root segments (Fig. 1). Capronia sp. DNA in a root did not reliably predict that a Capronia culture would grow from a neighboring segment of the same root. Instead, Capronia sp. cultures grew from only two of the eight roots that contained Capronia DNAs. Similarly, finding a Capronia sp. in a culture did not necessarily indicate that the genus could be detected in DNA extracts from a mycorrhizal root. Capronia sp. cultures were isolated from both the left and right flanks of root 1056 and yet Capronia sp. was not detected in the DNA extracts from middle segment of the same root (Fig. 1).
Little is known about the natural habitat of the vegetative mycelial stage of Capronia species. Capronia species attract the most attention when causing opportunistic infections of humans, but these infections are uncommon and they are probably an evolutionary dead end for the fungus. Capronia species’ sexual fruiting bodies are usually found on decorticated or well-rotted wood and they may be hyperparasites on other fungi (Untereiner & Malloch, 1999). Long-term accumulation of coarse woody debris and logging on our Vancouver Island research site had left abundant cedar and hemlock wood, and perhaps the Capronia isolates were relatively common because the salal roots were growing in and around a high concentration of well-rotted wood. Untereiner & Malloch (1999) found that Capronia species generally had no ability to use cellulose or starch and suggested that some may be mycoparasites. Our isolates penetrated the roots cells of salal suggesting that they have cellulases.
Evidence for the presence of mixed ascomycetes and basidiomycetes in ericoid mycorrhizae
Most ericoid mycorrhizal root contained mixtures of DNA types. Some of the fungi that contributed to the DNAs, the Trechispora spp. for example, were probably saprobic. Some of the fungi may have been endorhizal, able to penetrate living root cells, without conferring the physiological benefit to the salal that would make the fungi truly mycorrhizal. However, because of the way that ericoid mycorrhizae develop, some roots may have had mixed mycorrhizal fungi. In ericoid mycorrhizae, a hypha usually penetrates an individual cell and then forms a hyphal coil within one cell. The hypha within the cell does not usually grow into adjacent cells (Cairney & Ashford, 2002; Perotto et al., 2002). As a result, unlike most ectomycorrhizae where one fungal species predominates in one root tip, adjacent cells in an ericoid root can be mycorrhizal with different fungal species (Pearson & Read, 1973; Gianinazzi-Pearson et al., 1995; Nafar, 1998; Monreal et al., 1999; Perotto et al., 2002).
Finding mixed ascomycete and basidiomycete DNAs associated with salal roots is consistent with ultrastructural studies of mycorrhizal roots from other plants in the Ericaceae. Ascomycetes can be recognized by their simple septal pores that are often associated with spherical, electron opaque, Woronin bodies. In the Sebacinaceae, as in other ‘jelly fungi’ in the Auriculariales, septal walls have a central pore, surrounded by a torus-like swelling. The pore is bracketed on either side by a ‘parenthesome’ cap of thickened endoplasmic reticulum and the Auriculariales can usually be recognized by the small, irregular perforations in their parenthesomes. Duddridge and Read (1982) found both ascomycete-like and basidiomycete-type septa in mycorrhizal roots of Rhododendron but the ascomycete type septa predominated in the active mycorrhizal cells and basidiomycete septa occurred in dead cortical cells. However Peterson et al. (1980) found basidiomycete type septa in living cells of mycorrhizal Rhododendron. Auriculariales-type septal pores, in addition to ascomycete-like pores (Bonfante-Fasolo & Gianinazzi-Pearson, 1979) were found in apparently living ericoid mycorrhizae of Calluna vulgaris (L.) Hull (Bonfante-Fasolo, 1980) and Pieris (Peterson et al., 1980), and in epacrid mycorrhizae of Dracophyllum secundum R. Br. (Allen et al., 1989).
Finding the Capronia sp. and H. ericae in both direct DNA extracts and among cultures from mycorrhizal roots suggests that both species are consistent symbionts of salal. H. ericae is the best known and most frequently studied of the ericoid mycorrhizal fungi, and finding it frequently hints that it may be important in western Canada, as in Europe. The discrepancy between the high levels of Sebacina spp. DNA detected in roots and absence of Sebacina cultures from any of the hundreds of fungi from salal roots indicates a systematic detection bias. We plan to explore alternative culturing techniques to recover Sebacina sp. like isolates in culture and to subject them to mycorrhizal testing.
We thank Dr Keith Egger for assistance with phylogenetic analysis, Dr Sara Landvik for critical advice on manuscript and Dr Lee Taylor for access to unpublished sequences of Sebacina vermifera. Sea Ra Lim contributed to the sequencing. Lynn Sigler (University of Alberta Microfungus Collection and Herbarium) verified the identity of H. ericae cultures deposited in their collection. Funding for this research was provided by Forest Renewal B.C and Forestry Innovation Investment BC grants for the Salal Cedar Hemlock Integrated Research Program and by a grant to M. Berbee from the Canadian National Science and Engineering Research Council.