Author for correspondence: Brian J. Pickles Tel: +1 250 960 5308 Email: firstname.lastname@example.org
•Spatial analysis was used to explore the distribution of individual species in an ectomycorrhizal (ECM) fungal community to address: whether mycorrhizas of individual ECM fungal species were patchily distributed, and at what scale; and what the causes of this patchiness might be.
•Ectomycorrhizas were extracted from spatially explicit samples of the surface organic horizons of a pine plantation. The number of mycorrhizas of each ECM fungal species was recorded using morphotyping combined with internal transcribed spacer (ITS) sequencing. Semivariograms, kriging and cluster analyses were used to determine both the extent and scale of spatial autocorrelation in species abundances, potential interactions between species, and change over time.
•The mycorrhizas of some, but not all, ECM fungal species were patchily distributed and the size of patches differed between species. The relative abundance of individual ECM fungal species and the position of patches of ectomycorrhizas changed between years.
•Spatial and temporal analysis revealed a dynamic ECM fungal community with many interspecific interactions taking place, despite the homogeneity of the host community. The spatial pattern of mycorrhizas was influenced by the underlying distribution of fine roots, but local root density was in turn influenced by the presence of specific fungal species.
Ectomycorrhizal (ECM) fungi are a vital functional component of many of the world’s forest ecosystems (Smith & Read, 2008). The mutualistic association between ECM fungi and their hosts plays an important role in both carbon flow (Högberg et al., 2001) and nutrient dynamics (Read & Perez-Moreno, 2003). However, a great deal of basic information about the ecology of ECM fungi is still lacking, largely because of the inherent difficulties in studying the belowground spatial distribution and abundance of these cryptic soil organisms (Horton & Bruns, 2001; Taylor, 2002). Spatial ecology offers the potential to understand the complex interactions that can influence an individual species within a community (Tilman & Kareiva, 1997). Indeed, Lilleskov et al. (2004) have highlighted the importance of determining whether individual ECM fungal species exhibit distinctive spatial distribution patterns as a basis for investigating the ecological processes generating such patterns.
The analysis of species’ spatial distribution patterns is one of the most challenging aspects of studying below-ground organisms (Ettema & Wardle, 2002). As with any species, it involves quantifying the scale or scales (if any) at which spatial patterning exists and then, where possible, relating that pattern to the life history and growth form of the species, its autecology, and its possible interspecific interactions (facilitation, interference, resource competition; Levin, 1992; Tilman & Kareiva, 1997). Robust modern spatial statistical approaches (Legendre et al., 2002; Pickles et al., 2009) have the potential to reveal significant biologically relevant information about the spatial and temporal ecology of ECM fungi. Recently, for example, geostatistics have been used to examine the spatial distribution of ECM infection (Scattolin et al., 2008) and inoculum (Thiet & Boerner, 2007).
By providing a relatively fast-throughput and reliable means of confirming species identification, recent advances in molecular methods (Horton & Bruns, 2001) have allowed ECM research to open up new lines of inquiry into belowground spatial and temporal dynamics. Differences in the spatial distribution of sporocarps vs root tips (Kikuchi & Futai, 2003) and root tips vs mycelium (Genney et al., 2006) have shown that a range of spatial scales are relevant to studying ECM fungi. There have been several detailed belowground analyses of ECM fungal community composition at coarse spatial scales (50–100 m) in response to, for example, soil chemistry (Toljander et al., 2006) or elevated CO2 (Parrent et al., 2006). Vertical changes at a fine scale have been observed between soil horizons (Dickie et al., 2002; Rosling et al., 2003). However, horizontal variation in ECM fungal species distributions at fine scales (< 20 m) has been less well studied, particularly within homogeneous forest stands, although changes in species diversity and abundance have been observed across a forest edge (Dickie & Reich, 2005). Substrate preferences (Tedersoo et al., 2003) and positive or negative species’ interactions (Agerer et al., 2002; Koide et al., 2005a) have been identified. It is also well established that, at a range of scales, the spatial distribution of the hyphae of ECM fungi does not follow that of their mycorrhizas, and that the relative abundance of mycorrhizas and mycelium differs between fungal species (Koide et al., 2005b; Genney et al., 2006; Kjøller, 2006; Anderson & Cairney, 2007).
There is evidence that ECM fungi display temporal partitioning on root systems (Koide et al., 2007; Courty et al., 2008) and that a single ECM fungal species may completely disappear from an area in the space of a single year (Guidot et al., 2003). Over large spatial extents (25–200 m), ECM fungal communities appear to be relatively stable over time but at fine scales (5 cm) similarity between years is significantly reduced (Izzo et al., 2005), indicating turnover in species populations. At these finer scales the spatial distribution and temporal turnover of the tree fine roots themselves, which in boreal forests can be in the order of months (Högberg et al., 2002; Guo et al., 2008), can also be expected to influence spatial and temporal patterns of ECM fungal distributions.
In the study reported here we applied spatially explicit sampling and geostatistics to analyse and describe spatial and temporal patterns of abundance of ectomycorrhizas formed by different ECM fungal species in a Scots pine plantation. By selecting a uniform, even-aged, monospecific stand with no understory the emphasis in this study is on the emergence of structure in the ECM fungal community as a result of stochastic processes, species interactions, or fine-scale niche partitioning, rather than host changes or coarser scale environmental gradients.
Specifically, we addressed the following questions. How is the below-ground ECM fungal community structured in a homogeneous forest stand selected to minimize environmental variables? Are the mycorrhizas of individual ECM fungal species within a homogeneous host community patchily distributed and, if so, are there differences in the scale of patchiness between ECM fungal species? To what extent is patchiness (or the lack thereof) attributable to underlying patterns of root distribution, species interactions or species’ effects on root density? Does community structure, root distribution, location of patches and/or interactions between species change over time?
Materials and Methods
The study site was a 120-yr-old Scots pine (Pinus sylvestris L.) plantation in Culbin Forest, Morayshire, Scotland (57°38′08″N, 03°42′07″W). The site was flat and there were no obvious environmental gradients. The soil profile consisted of an organic horizon (bryophyte/litter (L, c. 0–2 cm); fermentation (F, c. 2–4 cm) and humic (H, c. 4–12 cm)) above deep Aeolian sand deposits (Gauld, 1981). The dominant bryophytes were Rhytidiadelphus triquetrus (Hedw.) Warnst. and Hylocomium splendens (Hedw.) B.S.G. Vascular plants were absent.
Abundance of ECM species
The abundance of ectomycorrhizas formed by each ECM fungal species was determined by counting live mycorrhizas in soil cores and allocating them to species. Mycorrhizas were extracted from the substrate by hand under a ×40 stereomicroscope and classified into different, distinguishable groups using morphotyping (Agerer, 1987–1998). Five to ten reference samples of each provisional morphotype were stored in 200 μl of CTAB buffer (2% hexadecetyltrimethylammonium bromide (w : v), 100 mM Tris-HCl (pH 8), 1.4 M NaCl, 20 mM EDTA and 1% polyvinyl pyrrolidone) at −20°C. DNA extracted from these samples was used to confirm the final morphotype categories and to assign the morphotypes to fungal taxa. Overall, 30–60 tips were analysed for each final morphotype.
DNA extraction, PCR and sequencing
Mycorrhizal root tips were homogenized using a micropestle and DNA was extracted using the DNeasy Plant Mini Kit according to the manufacturer’s instructions (Qiagen). The internal transcribed spacer (ITS) region was amplified for each DNA sample using the primers ITS1F and ITS4 or, where this resulted in a product with double bands, ITS4B (White et al., 1990; Gardes & Bruns, 1993). Amplifications were performed on a PTC-220 DYAD Thermal Cycler (MJ Research Inc., Waltham, MA, USA) in 50 μl volumes containing: 2.0 mM MgCl2, 5.0 μl buffer (16 mM (NH4)2SO4, 67 mM Tris-HCl (pH 8.8 at 25°C), 0.01% Tween-20) 2.0 mM dNTP, 20 pmol each primer, 2.5 units BioTaq DNA polymerase (Bioline, London, UK), and 100 ng DNA. The PCR cycling conditions consisted of 95°C for 5 min followed by 29 cycles of 30 s at 95°C, 30 s at 55°C and 30 s at 72°C with a final extension step at 72°C for 5 min. The PCR products were purified using the QIAquick PCR Purification Kit (Qiagen) and sequenced using BigDye Terminator v3.1 chemistry and an ABI PRISM 3130xl genetic analyzer (Applied Biosystems, Warrington, UK). Raw sequence data were analysed using the sequencher software package Version 3.0 (Gene Codes Corp., Ann Arbor, MI, USA) before being compared against the UNITE database (Kõljalg et al., 2005) using blast to identify the ectomycorrhizal fungus. For those ectomycorrhizas that produced multiple PCR bands on agarose gels, products were cloned with the pGEM-T Easy vector system (Promega). Colony PCR was performed using the M13 forward and reverse primers (Promega) with the cycling conditions described earlier to screen white colonies for the presence of an insert. Clones containing an insert were sequenced using the vector primers T7 and SP6 (Promega) as already described.
Spatially explicit sampling
A 20 m × 20 m plot was marked out on the most homogeneous part of the study site. Samples were taken with a 5 cm × 5 cm corer. This core size was the optimum trade-off between the number of morphotypes captured per core and the processing time per core (Pickles et al., 2009). Cores were taken to a depth of 20 cm on a grid at 2-m intervals. This primary separation distance was chosen because preliminary studies showed autocorrelation in the abundance of some common ectomycorrhizal morphotypes at distances between 1.4 m and 4.0 m (Pickles, 2007). To gather information on finer-scale distribution patterns, sets of four additional cores were taken 5 cm and 10 cm away from 24 randomly chosen grid points on an E–W or N–S axis. In total, there were 217 spatially explicit samples. Cores were collected on 9 June 2004, wrapped in polythene and stored at 3°C. Because the abundance of mycorrhizas declined sharply below the organic soil layers (Genney et al., 2006), the sandy mineral layer was discarded. The depth of the organic horizons was recorded and all cores were processed within 10 wk.
One year later (14 June 2005) a second set of cores was taken from the same plot. A 10 m × 10 m grid was overlaid on part of the original 20 m × 20 m grid, offset by 10 cm from the previous year’s primary grid points (Fig. 1). On this occasion, cores were taken on a 1-m grid, as the results in 2004 indicated that this would improve the resolution of the semivariograms and cluster analyses. In addition, for improved fine-scale resolution, 50-cm long transects made up of 10 adjacent cores were taken in a randomly assigned N–S or E–W orientation from 15 randomly chosen points on the main grid, giving 271 spatially explicit samples in all. Cores were processed as already described.
The absolute values of root tip abundance were used for data analysis (from 217 cores in 2004 and 271 cores in 2005). Semivariograms were used to measure how organic horizon depth, abundance of all root tips and abundance of ectomycorrhizas of individual ECM species were related to distance between samples. In all cases the data were log-transformed to approximate normality. Each variogram provides information on the variance caused by spatial structure in the data (C1) and the variance resulting from lack of autocorrelation, autocorrelation at finer scales than were measured, or measurement or sampling error (the ‘nugget effect’, C0). The proportion of variance resulting from spatial structure (C1/(C1 + C0)) varies from 0 to 1, with 0 indicating no measurable spatial structure and 1 indicating that all variance is caused by spatial structure. The ‘range’ indicates the distance at which data is no longer spatially autocorrelated (i.e. estimates the distance over which discrete patches occur). The experimental semivariogram was obtained by averaging one-half of the difference squared of the abundances over all pairs of observations separated by a particular distance. This was done for different distance intervals. The model semivariogram is a mathematical function, such as an exponential model, fitted to the experimental semivariogram. The r2-value indicates how well the model semivariogram fits the experimental semivariogram.
The ‘best-fit’ model (Legendre & Legendre, 1998) semivariograms were used in the kriging process to interpolate the abundance of each species at points within the study area that were not directly sampled. After kriging, predicted values for plotting were obtained through back-transformation from logarithmic values (Webster & Oliver, 2001). genstat version 7.2 (VSN International, Hemel Hempstead, UK) was used for the creation and modelling of variograms and for kriging. sigmaplot 8.0 (Aspire Software International, SE Leesburg, VA, USA) was used to display the interpolated data.
Cluster analysis of the absolute abundance of all ectomycorrhizas and of ectomycorrhizas formed by individual species was conducted using Spatial Analysis by Distance IndicEs (SADIE) (Perry et al., 1996; Xu & Madden, 2003). This method is used to detect patches of below-average and above-average abundance. Randomizations of the data allow creation of critical intervals with which to compare the observed indices of aggregation (D). The maximum number of randomizations allowable in SADIE (153 blocks of 39, or 5967 in total) was applied to reduce type II errors. SADIE was also used (Perry & Dixon, 2002) to assess the degree of association between the distribution patterns of ectomycorrhizas of different fungi in the same year, and the same fungus in different years. When analysing temporal patterns only those grid points recorded in both years were used in the analysis. The method produces an overall measure of association, X (mean of local values of spatial association), which is based on the degree of overlap between spatial patterns (Winder et al., 2001). To reduce type II errors, the maximum number of pattern randomizations allowable in SADIE (10 000) was used for comparison. For correlating pairs of spatial patterns the Dutilleul adjustment to sample size was made to take account of spatial autocorrelation in the dataset (Perry & Dixon, 2002).
Finally, to examine the interactions between individual ECM fungal species and the underlying fine root distribution, and possible species’ effects on root density, we applied spatially explicit logistic regression analyses to species with frequency of > 5% in at least one year, with the response variable as the proportion of the focal species in each core and the explanatory variable the logarithm of the total tip count for that core. We fitted logit(sp_tips/total) = β0 + β1 × log(total) + error, for each species/species complex, using glmmPQL in R (R Development Core Team, 2008). The correlation between the error terms was assumed to be a function, corr(x,y), of the distance between the locations of the cores. The random effect, corr(x,y), was modelled as exponential decay with no nugget. The significance test on β1 measured the extent to which some species become over-represented (positive coefficient) or under-represented (negative coefficient) in cores of high tip abundance. The Benjamini & Hochberg false discovery rate (FDR) correction (Verhoeven et al., 2005) was applied, in which P-values were ranked from most to least significant with the strength of the correction decreasing with rank order.
The mycorrhizal community
Over 52 000 short root tips were counted in 2004, and a further 40 000 in 2005. No nonmycorrhizal short root tips were encountered. The mycorrhizas of 24 different ECM fungal species were recognized, for which relative abundance and frequency were calculated (see the Supporting Information, Table S1). Seventeen species were recorded in 2004; of these, 12 were recorded the following year, along with seven new species. The relative abundance of seven species (or species complexes) changed significantly between 2004 and 2005 (Fig. 2). Total estimated species richness derived from collector’s curves (not shown) was 17 in 2004 and 20 in 2005, suggesting that most of the species in the community that could be distinguished by morphotyping had been encountered. The most abundant mycorrhizal morphotype in both years was black, with coarse emanating hyphae. Specimens of this variable morphotype normally yielded sequences of the widely distributed ascomycete Cenococcum geophilum Fr., but a minority of specimens (< 5%) gave sequences close to Cadophora finlandica (Wang & Wilcox) Har. & McNew. Counts of mycorrhizas of these two species were combined for further analyses and referred to as ‘Black type’. In 2004, mycorrhizas formed by five different Cortinarius species were recorded. With the exception of Cortinarius semisanguineus (Fr.) Gillet mycorrhizas, it was not possible to reliably distinguish separate Cortinarius species on the basis of morphology, so the remaining four species were combined for further analyses and referred to as the ‘Cortinarius complex’. In 2005, collections of Cortinarius mycorrhizas from individual cores were counted separately and we subsequently sequenced many more individuals before combining the counts data as appropriate. Using this approach we reliably determined the location and abundance of five different species. Two of the Cortinarius spp recorded in 2004 were not encountered in 2005 (Table S1), and two new Cortinarius spp were encountered in 2005 (Table S1), although one was at extremely low abundance/frequency. For the purposes of comparing the distribution of the Cortinarius complex in 2005 with that of the previous year, the data from individual species in 2005 were combined. The decision to combine the 2005 data is supported by the positive association observed between each of these species when their data was analysed separately (see below: Table 4b).
Detecting spatial pattern
Four species/species complexes (Black type, Cortinarius complex, Clavulinaceae sp., Lactarius rufus (Scop.: Fr.) Fr.) occurred at sufficiently high frequency in 2004 to allow semivariance analyses. These and three others (C. semisanguineus, Russula sardonia Fr. and Suillus variegatus (Swartz: Fr.) O. Kuntze), could also be analysed in 2005. The total number of mycorrhizas in both years was also analysed. The proportion of variation accounted for by the fitted (exponential) semivariogram varied from 0.20 to 0.94 (Table 1). It is important to note that the abundance of all mycorrhizas taken together was spatially structured, that is, they were not regularly or randomly distributed across the sampling areas. This in itself has the potential to impose spatial structure on the mycorrhizas of individual fungal spp.. Of the individual species’ ectomycorrhizas, three (Cortinarius complex, C. semisanguineus and R. sardonia) showed both good model fit (r2 > 0.58) and strong (> 64% variance in abundance attributable to spatial autocorrelation) spatial structure. The distance over which spatial autocorrelation was detected for the mycorrhizas of these species, a measure of the ‘patch size’ in which they occur, ranged from 0.5 to 1.2 m. Other species (Clavulinaceae sp., L. rufus and Black type in 2004) showed good model fit but weaker (< 40% variance attributable to spatial autocorrelation) spatial structure.
Table 1. Semivariance analysis of the distribution of all ectomycorrhizas, and of ectomycorrhizas of six species/species complexes in a 120-yr-old Pinus sylvestris plantation (analysis was not possible for two species in 2004)
Structural variance2 (C1)
Nugget variance3 (C0)
Spatial structure4C1/(C0 + C1)
Model fit5 (r2)
1Data for 2004 are from 217 samples within a 20 m × 20 m plot: data for 2005 are from 271 samples within a 10 m × 10 m plot nested within the original 20 m × 20 m plot.
2Variance attributable to spatial autocorrelation.
3Variance not attributable to spatial autocorrelation.
4Proportion of variance due to spatial structure.
5Proportion of total variation accounted for by the fitted variogram (exponential model).
6Distance over which spatial autocorrelation was detected.
The kriged maps (Fig. 3) show the position of patches of mycorrhizas. Areas of highest mycorrhizal abundance appeared unrelated to the position of the trees (Fig. 3a) and were only weakly positively correlated with the depth of the organic matter (2004, r2 = 0.13, P <0.001; 2005, r2 =0.25, P <0.01). The mycorrhizas of individual species show patches of greater or lesser intensity (Fig. 3b–f), the most striking being those of C. semisanguineus (Fig. 3e) and the Cortinarius complex (Fig. 3f) which show discrete patches of high abundance, ranging from approx. 0.25 to 9 m2, interspersed with areas of very low or zero, abundance.
SADIE (Table 2) provided a complementary method for the detection of aggregations of higher or lower than average abundance of mycorrhizas of individual species. Usefully, it was applicable to a wider range of species than standard semivariance analysis. These indices generally confirm the results of the semivariance analyses and show aggregations in one or both years for Black type, Clavulinaceae sp, L. rufus, Cortinarius complex, C. semisanguineus, Russula emetica (Schaeff.: Fr.) Pers., R. sardonia, Tomentellopsis submollis (Svrcek) Hjortstam and an Agaricomycetes species. For all species except L. rufus, clustering was more evident in 2005, but it is not possible to tell whether this represents some real change in the spatial distribution of the species or is a consequence of the closer sampling interval employed in 2005.
Table 2. Spatial analysis by distance indices for all ectomycorrhizas, and for the mycorrhizas of 15 species/species complexes in a 120-yr-old Pinus sylvestris plantation (some species were found in only one of the study periods). Values in bold are significant (P < 0.05).
1Data for 2004 are from 217 samples within a 20 × 20 m plot: data for 2005 are from 271 samples within a 10 × 10 m plot nested within the original 20 × 20 m plot.
2Index of aggregation, where Ia > 1 implies clustering of mycorrhizas, Ia = 1 implies a random distribution and Ia < 1 implies regularity.
3Patch cluster index. For random arrangements of cores with counts of mycorrhizas > the mean, νi has an expectation of 1.
4Gap cluster index. For random arrangements of cores with counts of mycorrhizas < the mean, vj has an expectation of −1.
Elaphomyces muricatus Fr.
Taken together, the results of these two approaches show that the mycorrhizas of most of the species encountered on the site were clustered to some extent, but that the intensity of the clusters and the size of the patches differed between species. The most apparent and discrete clusters were formed by the Cortinarius complex (and its constituent species in 2005; individual species data not shown) and C. semisanguineus. In addition, the root system on which the mycorrhizas were formed was itself unevenly distributed, with patches of higher and lower than average short root abundance. The extent to which this may have determined the spatial distribution of individual mycorrhizal species is examined below.
The causes of spatial pattern
There were marked differences between species in the extent to which they were either over- or under-represented in cores with a high number of mycorrhizas. The dominant species complex, Black type mycorrhizas, was significantly under-represented in cores with high tip counts in both years (Table 3). Clavulinaceae sp., Elaphomyces muricatus Fr. and T. submollis were also under-represented, significantly so in at least one of the years. Conversely, the Cortinarius complex was over-represented in high abundance cores in both years, and L. rufus and R. emetica were over-represented in one year.
Table 3. The β1 coefficients for spatially explicit logistic regression analyses of species with frequency of > 5% in at least one year, with the response variable as the proportion of the focal species in each core and the explanatory variable the logarithm of the total tip count for that core
The significance test on β1 measures the extent to which species are over-represented (positive coefficient) or under-represented (negative coefficient) in cores of high tip abundance. Values in bold are significant (P <0.05) following Benjamini and Hochberg false discovery rate (FDR; Verhoeven et al., 2005).
To test whether associations (negative or positive) between species might account for spatial pattern in the distribution of their mycorrhizas, we calculated the correlation coefficient for pairwise comparisons between the clustering indices (‘D’ from SADIE) for each individual species/species complex in each core (Table 4). This is a more robust test than a simple χ2 test based on presence/absence of species, because it takes into account the abundance of each species in a core relative to its mean abundance (i.e. it tests the extent to which aggregations of species overlap). Over both years c. 30% of possible interactions were significant. In 2004, there were more positive than negative associations, but in 2005 the number was similar. Over the two years, all species/species complexes showed some significant interaction with another species/species complex and all showed both positive and negative associations, except for L. rufus which showed only positive associations, notably with the other widespread and abundant species/species complexes, that is, Black type and Clavulinaceae sp. The most striking outcome was the strong negative relationship between patterns of C. semisanguineus and the Cortinarius complex in 2004, and between C. semisanguineus and each of the other Cortinarius spp in 2005. By contrast, these other Cortinarius spp were all positively associated with each other, lending support to the use of the Cortinarius complex as a proxy for their distribution. In 2005, 57.6% of the significant interactions involved Cortinarius species, which made up 33% of the species examined with this method. While there was some consistency in interactions between years, as described above for L. rufus and C. semisanguineus/Cortinarius spp, there were also inconsistencies. For example, S. variegatus was positively associated with Clavulinaceae in 2004, but negatively associated in 2005.
Table . Table 4 Positive and (−) negative spatial association between mycorrhizas formed by ectomycorrhizal (ECM) fungal species/species complexes with an overall frequency > 5% in (a) 2004 and (b) 2005
In 2005 the mean number of mycorrhizas in a core (151 ± 7.5) was 27% lower than in the same area in the previous year (208 ± 25.3). There were also very marked changes in the distribution of mycorrhizas from 2004 to 2005 with a net decline in the southern half of the resampled area and a net increase in the north-west corner (Fig. 3a). In addition, there were significant changes in the relative abundance of mycorrhizas of individual species/species complexes (Fig. 2). Mycorrhizas of the Black type, L. rufus and T. submollis declined in relative abundance, while those of Clavulinaceae sp., S. variegatus, R. sardonia and R. emetica increased. The most striking change was in the abundance of L. rufus mycorrhizas, which declined from an average of 50 ± 11.6 per core in 2004 to 5 ± 1.2 per core in 2005. The position of the patches of mycorrhizas of the majority of individual species also changed (Fig. 3b–f). To test whether there was any relationship between the position of patches in 2004 and 2005 we calculated the correlation coefficient between the clustering indices for each individual species/species complex at every point that had been sampled in both years. There was a positive association for mycorrhizas in the Cortinarius complex (X = 0.38, P =0.02), and a strong dissociation for mycorrhizas of T. submollis (X = −0.639, P <0.001), but for all other species/species complexes there was no evidence of a relation between where the patches were in 2004 and where they were in 2005.
In general, the species’ relative abundance and relative frequency follow the well-documented patterns for communities of ectomycorrhizas, with a few species (Black type, L. rufus, Cortinarius complex, Clavulinaceae sp.) contributing 80–90% of the mycorrhizas, and the other species present at very low relative abundance (Horton & Bruns, 2001). Similarly, some species (e.g. C. semisanguineus, S. variegatus) were found at high relative frequency but low relative abundance, indicating that many cores contained a few mycorrhizas of these species.
Detection of spatial pattern
The scale of patchiness for individual taxa found in this study (0.5–1.2 m) is roughly similar to that detected by Lilleskov et al. (2004). Discrete patches of high mycorrhizal abundance were particularly evident for mycorrhizas formed by Cortinarius species, while mycorrhizas formed by an unknown member of the Clavulinaceae formed more diffuse patches. The smallest patches of ectomycorrhizas of any species identified in this study were c. 0.25 m2. However, other studies have shown that ectomycorrhizas are also aggregated at much finer scales. For example, Genney et al. (2006), also working at Culbin forest site, found patches of ectomycorrhizas of several of the species described here that were only a few centimetres across. Similarly Gebhardt et al. (2009) found that the ectomycorrhizas of several fungal associates of Quercus rubra L. were clustered at the centimetre scale, while other studies have shown that the distribution of ectomycorrhizas of some species can be patchy at the scale of tens of metres (e.g. Piloderma sp.; Danielson & Visser, 1989).
Are the patches of species found in this study made up of mycorrhizas formed by the same individual or genotype, that is, do they represent genets, and do multiple patches of a species represent separate genets or fragments of the same genet? Further molecular genetic investigations would be required to answer these questions. Most studies on genets of ECM fungi have been based on the molecular typing of sporocarps and suggest that genets can range from < 1 m to > 30 m in diameter depending on factors such as species identity, forest age and degree of disturbance (Gherbi et al., 1999; Sawyer et al., 1999; Anderson et al., 2001; Fiore-Donno & Martin, 2001; Bergemann & Miller, 2002; Guidot et al., 2002). Fewer studies have utilized the mycorrhizas themselves (Guidot et al., 2001; Zhou et al., 2001; Hirose et al., 2004; Kretzer et al., 2004; Lian et al., 2006) but these show a similar range in genet size. Clearly, the patches demonstrated here could equate to genets, but equally they could contain mycorrhizas from more than one genet, or indeed represent fragments of larger genets. It is possible that the difficulty in resolving the spatial distributions of some of the well-represented species in this study, for example low proportion of spatial structure detected for Black type in 2004, and both Clavulinaceae sp. and L. rufus in 2005, is caused by confounding effects such as small genet size and intraspecific competition. Clavulinaceae sp. and L. rufus both displayed high nugget values, indicating that the scale of spatial structure may have been less than the sample size of the survey (i.e. most of the spatial structure in these species may occur at distances < 5 cm). A micromapping-scaled study over a few square centimetres, coupled with molecular approaches to genet analysis, may help to resolve these patch detection issues for such species in future studies.
The causes of spatial patterns – functional morphology
The emergence of spatial structure in populations of mycorrhizas of a particular species could be related to the functional morphology (Agerer, 2001) of the fungi concerned. For example, Cortinarius mycorrhizas generally produce copious hyphal strands emanating from the sheath surface, which may help to exclude competitors around a focus of infection. L. rufus and Clavulinaceae sp. mycorrhizas, on the other hand, produce more diffuse external mycelium, possibly less suited to excluding competitors. These species occurred in less well-defined patches. By contrast, Black type mycorrhizas, predominantly formed by C. geophilum, which produce relatively little external mycelium close to the surface of the mycorrhiza, occurred throughout the host root system. Other studies have also found that C. geophilum mycorrhizas are uniformly distributed throughout root systems (Lilleskov et al., 2004; Gebhardt et al., 2009). Either this represents repeated de novo infection from mycelium or other propagules in the substrate as new roots are produced, or possibly the fungus tracks the extending root system by growing along the surface of the long roots. C. geophilum has a worldwide distribution, and is often the most abundant fungus in communities of ectomycorrhizas (LoBuglio, 1999; Dickie, 2007). However, the fungus called ‘C. geophilum’ is known to be genetically diverse (Douhan et al., 2007), even at a scale of a few centimetres (Douhan & Rizzo, 2005), so the widespread root colonization reported here may result from multiple genotypes. This, coupled with the potential for confusion with other ascomycetes forming black ectomycorrhizas, as demonstrated here, makes interpreting the spatial ecology of this important fungus particularly challenging.
The causes of spatial patterns – ecological interactions
How do patches of mycorrhizas of a particular species arise and what determines where on the root system they arise? Interspecific competition, and particularly priority effects, could lead to spatial partitioning of the fine root system between different fungi. For example Agerer et al. (2002) found that at the scale of a few centimetres there were several pairs of ECM fungi whose mycorrhizas were never found together, suggesting competitive exclusion. Moreover, a number of microcosm and field studies have shown both competitive replacement and competitive priority effects in ECM fungal interactions (Wu et al., 1999; Landeweert et al., 2003; Kennedy & Bruns, 2005; Kennedy et al., 2007a; b). However, the outcomes of competitive interactions between ECM fungi seem to be strongly dependent on local conditions (Kennedy et al., 2007b), and therefore highly stochastic in nature.
We found good evidence for positive and negative interactions between species by testing the extent to which areas of high abundance of species pairs did or did not overlap using the clustering indices from SADIE. The most striking outcome of these analyses was the consistent dissociation between C. semisanguineus and the Cortinarius complex, as well as its constituent species in 2005. It is tempting to invoke increased strength of competition between these taxa because of trait similarity (medium distance exploration types) and/or phylogenetic relatedness as an explanation (as observed for sporocarps of Russula species; Murakami, 1987). However, individual Cortinarius species other than C. semisanguineus were positively associated in 2005, and in both years L. rufus and Clavulinaceae were positively associated, despite being similar short-distance/contact exploration types. Clearly, trait similarity/phylogenetic relatedness alone cannot explain dissociation between species and other factors, such as antibiosis, may be more likely. The only previous study to look in detail at spatial interaction between the mycorrhizas of ECM fungal species (Koide et al., 2005a) also found evidence of positive and negative species interactions. Interestingly Koide et al. (2005a) were also working in a pine plantation (65-yr-old Pinus resinosa Ait) on sandy soil, and their community of ectomycorrhizas was also dominated by C. geophilum, a Clavulina sp. and a Lactarius sp. They used presence/absence data and found that 75% of the significant interactions involved these dominant species. That was not the case in our study, possibly in part because we used abundance rather than presence/absence data, and were able to avoid the multiple testing problems arising from the use of repeated χ2 tests.
Another possibility is that the development of patches is partly a consequence of the effect of a particular fungus on root system morphology. We found that certain species (Cortinarius sp 1, L. rufus and R. emetica), along with the Cortinarius complex, were over-represented when there were a high number of mycorrhizas in a core. It is well known that different ECM fungi are associated with different degrees of short root branching at the site of infection (Agerer, 2006). The most likely mechanism for this is the involvement of fungal growth regulators in mycorrhizal formation and short root morphogenesis (Barker & Tagu, 2000). However, while there is good evidence that mycorrhizal morphology is related to, for example, auxin production within species (Marmeisse et al., 2004), comparative studies between species appear to be lacking. It is possible that, given suitable conditions, some ECM fungi are capable of stimulating host fine root production, thereby enhancing the priority effect. Other possibilities are that host carbohydrate could be preferentially allocated to particular infection sites, thereby stimulating short root production (Rosling et al., 2004), or that these species are simply strong competitors in microsites which promote high mycorrhizal abundance. By contrast, Black mycorrhizas, predominantly formed by C. geophilum, were under-represented in cores with many tips. This could partly result from the simple unbranched morphology of C. geophilum mycorrhizas on pine, but also suggests that C. geophilum may be a poor competitor in some situations. Lastly, as suggested by Bruns (1995), spatial heterogeneity of resources (some of which may be caused by the activities of the fungi themselves; Aguilera et al., 1993; Agerer & Göttlein, 2003) and conditions in the soil may give rise to spatially discrete niches that favour particular ECM species. However, given the effort to minimize resource heterogeneity, this deterministic explanation seems less likely for our study site.
Previous attempts to quantify temporal changes and the extent of species turnover in communities of ectomycorrhizas have been compromised by the sampling intensity required to detect real changes when fine-scale spatial variation in species relative abundance is high (Taylor, 2002). By taking a mapping approach we were able both to quantify changes in individual species abundance in a defined area and to visualize the movement of patches of ectomycorrhizas of specific species over time. We found a marked change in the distribution of ectomycorrhizas between the two sampling events, which is not surprising given the life span of individual mycorrhizas (Downes et al., 1992) and the high rate of turnover of fine roots in temperate forest (Högberg et al., 2002; Guo et al., 2008). The patches of Cortinarius spp. were close to where they had been 12 months previously, but for most species the spatial distribution in 2005 was unrelated to that in 2004, and for T. submollis there was evidence of a negative association. The relative abundance of most species changed over the sampling interval. By contrast, at Koide’s site (described earlier) there was little change in the relative frequency of the dominant species over a 13-month period (Koide et al., 2007). The most striking change in our study was the decline in abundance of mycorrhizas of L. rufus. A similar dramatic decline has previously been shown for Hebeloma cylindrosporum (Romagnesi) (Guidot et al., 2003). It was much warmer and drier at our study site in 2005 than in 2004 (Pickles, 2007), and drought may have caused the observed overall reduction in numbers of mycorrhizas. Possibly, the strongly hydrophilic mycorrhizas of L. rufus were particularly susceptible to drought.
By applying spatial statistical analysis to the distribution of ectomycorrhizas in the surface organic horizons of a Scots pine plantation we have shown that the mycorrhizas of some, but not all, ECM fungal species were patchily distributed and that the size of patches differed between species. Patchiness was partly attributable to underlying patterns of root distribution and their changes over time, but species interactions and/or species’ effects on root density were also very important. The relative abundance of individual ECM fungal spp. and the position of patches of ectomycorrhizas changed significantly from one year to the next. These new insights into the dynamic nature of an ECM fungal community, even within a homogeneous forest stand, were made possible by using a spatially explicit sampling approach coupled with geostatistical and other spatial and temporal analyses.
We thank Pamela Parkin for her technical assistance, Marianne Pickles for her assistance with root sorting, the Forestry Commission for access to Culbin Forest, and three anonymous reviewers for their valuable comments. This work was funded by the Natural Environment Research Council (Grant: NER/A/S/2002/00861) and the Scottish Government (Rural and Environment Research and Analysis Directorate).