Idiosyncrasy and overdominance in the structure of natural communities of arbuscular mycorrhizal fungi: is there a role for stochastic processes?

Authors


Correspondence author. E-mail: ajd121@york.ac.uk

Summary

1. Most studies of species abundance patterns focus on conspicuous macroorganisms while microbial communities remain relatively understudied. This bias is a concern given the functional importance and high diversity of microbes.

2. We determine whether a common species abundance distribution (SAD) is observed in communities of a widespread group of soil microbes, the Glomeromycota or arbuscular mycorrhizal (AM) fungi. Using molecular techniques, we intensively sampled the AM fungal community of a woodland–grassland ecotone in Yorkshire, UK. Observed species abundances were compared to theoretical models describing SADs. We also reanalysed 32 previously published data sets in a similar manner.

3. Species abundance distributions in all the AM fungal communities fitted both lognormal and broken-stick models. However, these models consistently and significantly underpredicted the abundance of the most abundant AM fungal taxon. We found that AM fungal communities are typically dominated by a single taxon; representing on average 40% of total abundance within the community. Phylogenetic analysis of the most abundant taxa across data sets showed that the dominant AM fungal type in each community was different and not a widespread generalist.

4. We conclude that a common community structure is present in AM fungal communities from different habitats. The fit to log-normal and broken-stick models suggests the influence of niche differentiation structuring these communities. However, the consistently observed overdominance indicates that local adaptation and stochastic processes may also play important roles in structuring these communities, and we propose a mechanism to explain overdominance in AM fungal communities.

5.Synthesis. This paper applies ecological models derived from studies on larger organisms to microbial communities. Results from this study suggest that a common log-normal SAD is likely to be observed across both microbial and macro taxa. However, due to the distinctive features of microbial biology, some noticeable differences, such as heavy overdominance, may lead to unique structures in microbial communities. This research not only highlights that, to a first approximation, microbial communities follow similar processes and have similar patterns to those of macroorganisms, but also the need for large-scale microbial data sets, if we are to understand the patterns and processes regulating global biodiversity.

Introduction

Understanding the processes and mechanisms that determine the species richness (how many species) and relative species abundance (number of individuals of each species) of an assemblage or community of organisms has long been a central aim of ecological research (May 1975; Pielou 1975; Tokeshi 1990; Williamson & Gaston 2005; Mcgill et al. 2007). Species abundance distributions (SADs) represent a detailed summary of a community, incorporating information on both species richness and abundance. In general, SADs show that a few species are very abundant and the majority of species are less common or rare (Mcgill et al. 2007). The universality of this observation across taxa and habitats suggests that it is one of ecology’s true universal laws (Mcgill et al. 2007). However, the universality of SADs has been inferred predominantly from studies on conspicuous macroorganisms and has rarely been examined in microbial communities and it is not certain whether microbial taxa follow this ‘universal’ pattern.

Arbuscular mycorrhizal (AM) fungi (Phylum: Glomeromycota) are a widespread and functionally important group of soil microbes with poorly resolved community ecology (Rosendahl 2008). AM fungi are obligate plant-root endosymbionts that colonize approximately two-thirds of terrestrial plant species, acquire all their carbon from the host plant and deliver to the plant a range of benefits, notably increased phosphorus uptake; they thus have profound effects on plant community dynamics and diversity (Fitter 2005; Rosendahl 2008). However, despite the importance of AM fungi within terrestrial ecosystems, we do not know whether AM fungal communities have a similar structure to that observed in higher taxa or whether similar processes are responsible for creating that structure. This lack of data on community-level processes affecting AM fungi reflects the difficulties of studying these organisms. For example, AM fungal species cannot be identified morphologically in roots beyond the genus level (Merryweather & Fitter 1998) and most isolates from natural habitats are not amenable to culture (Helgason et al. 2002). Modern DNA-based techniques have now allowed the quantification of AM fungal communities from a range of natural habitats (for a review see Opik et al. 2006). However, most studies have simply catalogued species from different habitats, and little attempt has been made to ascertain whether general patterns of community structure exist in AM fungal communities.

In this study, we examine SADs in AM fungal communities in order to investigate whether a common process structures natural AM fungal communities. First, we characterized a natural AM fungal community from a single locality and fitted the observed SADs to three commonly used theoretical models; the broken-stick, the lognormal and the geometric series. Secondly, using 32 data sets collected from 25 published studies, we examined whether the observed fit of AM fungal SADs to theoretical models is common across studies and habitats. Finally, we examined patterns of dominance in AM fungal communities to determine whether the dominant taxa in each data set are widespread generalists or locally abundant specialists.

Materials and methods

Study site

In order to fully quantify a natural AM fungal community and provide data as a baseline for comparisons of AM fungal community structure from different habitats, we extensively sampled an AM fungal community from Hetchell Wood (Leeds, UK; 53°52′ N, 1°25′ W; for a detailed site description see Dumbrell et al. in press). Field work was conducted in February, March, July and August 2007. A site of 30 × 20 m was selected covering the transition from grassland on magnesian limestone to woodland on millstone grit, with a pH range from pH 3 to 8. Six soil cores (40 mm diameter) were taken to a depth of 15 cm in February, July and August and 12 cores were taken in March 2007. Locations in which to core were chosen at random. All samples were placed in zip-lock plastic bags, returned to the laboratory within 1 h, and stored at 5 °C before they were analysed. Plant roots were separated from the soil core, washed and oven-dried at 65 °C for 5 days, and stored in sealed plastic bags ready for DNA extraction.

Molecular methods

We extracted DNA from mixed plant roots (pooled root samples from the 30 soil cores) using MoBio PowerPlant DNA isolation kits following the manufacturer’s instructions (Mo Bio Laboratories, Inc., Carlsbad, CA, USA). In order to quantify the AM fungal community from mixed plant roots, we used clone screening, sequencing and restriction fragment length polymorphism (RFLP) typing of the small subunit (SSU) region of ribosomal DNA. Assessing the relative abundance of AM fungi based on clone screening and RFLP typing provides a robust estimate of AM fungal abundance due to minimal difference in sequence structure between AM fungal taxa (see Helgason, Fitter & Young 1999).

A 550-bp partial fragment of SSU rDNA was amplified by PCR using Taq DNA polymerase (Invitrogen Co. Carlsbad, CA, USA) and the universal eukaryotic primer NS31 (Simon, Lalonde & Bruns 1992) and the primer AM1, which excludes plants and amplifies the major Glomeromycotan families (Helgason et al. 1998). PCR was carried out in a 25-μL reaction volume with 1 μL of DNA template, 2 mm dNTPs, 10 pmol of each primer (PCR conditions: 95 °C for 2 min; 30 cycles at 94 °C for 0.5 min, 63 °C for 0.5 min and 72 °C for 1 min; and 72 °C for 10 min) on a Techne TC-512 (Techne Co. Staffs, UK). To remove humic-acid-based PCR inhibitors, 0.125 μL of T4 gene 32 protein (Roche Diagnostics Ltd, W. Sussex, UK) was added to all PCR reactions. PCR products were purified using QIAquick PCR Purification Kit (Qiagen Ltd, Crawley, UK).

Purified PCR products were then ligated into pGEM-T Easy vector (Promega Co., Madison, WI, USA) and cloned in Escherichia coli (DH5α) (Invitrogen). Putative positive transformants were screened using standard SP6–T7 amplification. All positive clones were restriction digested using HinfI (Fermentas, St. Leon-Rot, Germany), Hsp92II and RsaI (Promega) enzymes following the manufacturer’s instructions. In addition, clones with distinct RFLP types were amplified using standard SP6–T7 amplification, purified using the QIA quick (Qiagen) purification kit and sequenced (Macrogen Inc., Seoul, Korea: sequencing conducted under BigDyeTM terminator cycling conditions and run using an ABI3730xl automatic sequencer).

Data analysis

Small subunit rDNA sequence chromatograms were checked using chromas. clustalx (Thompson et al. 1997) was used for multiple alignments and calculation of neighbour-joining phylogeny (Saitou & Nei 1987) using Geosiphon pyriformis (Gehrig, Schüßler & Kluge 1996) as a specific outgroup to the AM fungi as well as Corallochytrium limacisporum, a choanozoan, as a general outgroup to all fungi (Vandenkoornhuyse et al. 2002). Phylogenetic support was calculated using nonparametric bootstrapping (Felsenstein 1985), with 10 000 pseudoreplicates. DNA sequences were compared against reference sequences from cultured isolates and environmental samples (http://blast.ncbi.nlm.nih.gov/Blast.cgi) using the Blast algorithm (Altschul et al. 1990). We considered sequences with pairwise similarities < 97% to be separate sequence types and to be species for the purpose of this analysis. Sequences were considered to be new records if they were ≥ 3% different from any previously recorded sequence types collected from environmental samples. All the new sequences have been submitted to the European Molecular Biology Laboratory (EMBL) Nucleotide Sequence Data base (accession numbers FN556614FN556650). Clone numbers were used to calculate the relative abundance of each distinct sequence type and allowed the calculation of species rank abundance distributions.

The observed species rank abundance distribution was then compared to those predicted by three commonly used theoretical models describing SADs; the lognormal, broken-stick and geometric series (for model details see Magurran 2004). There has been much debate as to the best method of fitting observed SADs to theoretical models (see Bulmer 1974; Wilson 1991; Williamson & Gaston 2005; Mcgill et al. 2007) and correct choice of goodness-of-fit statistic is crucial. Throughout this study we fit observed data to theoretical models of species rank abundance plots. When working on rank abundance plots, the use of chi-squared and log-likelihood ratios can give too much importance to rare species and bias results (Mcgill et al. 2007; Poulin et al. 2008). Thus, following recommendation in Poulin et al. (2008), the goodness-of-fit of the theoretical models to the observed SADs was assessed using Kolmogorov–Smirnov tests.

Meta-analysis of previous studies

In order to examine whether observed SADs of AM fungal communities followed a similar form across habitats, we reanalysed data from previous publications. Twenty-five studies were analysed that gave a total of 32 additional data sets (for a full list see Table 1). Studies were considered to contain more than one data set if it was clear that data had been collected from two distinct habitats, for example from agricultural land and natural woodland (see Helgason et al. 1998). Criteria for inclusion of a study in this meta-analysis followed those outlined by Opik et al. (2006), who compiled an extensive list of AM fungal studies from different ecosystems across the globe. The number of distinct sequence types and their relative abundance, estimated by restriction digests of screened clones, was used to calculate species rank abundance distributions. The observed species rank abundance distribution was then compared to those predicted by the lognormal, broken-stick and geometric series, and the goodness-of-fit of these theoretical models was assessed using Kolmogorov–Smirnov tests, as described in the previous section.

Table 1.   Fit of arbuscular mycorrhizal (AM) fungal species abundance distributions from 25 different studies to theoretical species abundance models. Letters in brackets indicate more than one data set from a single study. Studies are ordered by decreasing species richness. Values in bold indicate significantly different observed species abundance distributions to those predicted by theoretical models, significance levels are denoted by asterisks (*P = 0.05; **P = 0.01; ***P = 0.001). Values highlighted in grey indicate which model best predicts the species abundance distribution of the AM fungal community sampled in each study Thumbnail image of

The total number of taxa recorded varied from five to 47 across studies. In some of the less species-rich communities the statistical power of the goodness-of-fit tests, fitting theoretical models to observed data, was reduced. In order to overcome problems associated with difference in species richness between samples, we employed a simple index of dominance to assess the numerical dominance of the most abundant species in each community (Poulin et al. 2008). We used the ratio between the overall abundance of the most abundant taxon and the second-most abundant taxon (1:2 ratio) and also between the most abundant and third-most abundant taxon (1:3 ratio). This simple index of dominance was proposed by Poulin et al. (2008) and we follow their methods.

In order to investigate whether a single, common taxon dominates AM fungal communities across habitats, we extracted the most abundant sequence type from each study from the EMBL Nucleotide Sequence Database. Twenty-four data sets used the primer pair NS31-AM1 (Helgason et al. 1998) which covers the central 500 bp of the SSU rDNA gene. These sequences were then used for phylogenetic analysis. clustalx (Thompson et al. 1997) was used for multiple alignments of abundant sequence types and calculation of neighbour-joining phylogeny (Saitou & Nei 1987) using G. pyriformis (Gehrig, Schüßler & Kluge 1996) as a specific outgroup to the AM fungi as well as C. limacisporum, a choanozoan, as a general outgroup to all fungi (Vandenkoornhuyse et al. 2002). Phylogenetic support was calculated using nonparametric bootstrapping (Felsenstein 1985), with 10 000 pseudoreplicates.

Results

Diversity and structure of the AM fungal community in Hetchell Wood

Six-hundred and seventeen clones containing the SSU rDNA gene were investigated and screened by Hinf1, Hsp92II and Rsa1 RFLP typing and 314 clones were then sequenced. Combining sequence and RFLP grouping, we recorded 37 AM fungal taxa from mixed root samples (30 soil cores) taken from Hetchell Wood. Distinct AM fungal taxa were defined as sequenced types which were ≥ 3% different from all other AM fungal sequences recorded from natural field systems. The 37 AM fungal taxa recorded in this study are given in Fig. 1 and were from Acaulospora (Acaulosporaceae), Glomus group A (Glomeraceae) and Glomus group C (Diversisporaceae) following Schüβler, Schwarzott & Walker (2001). Twenty-seven taxa had previously been sequenced from environmental samples or known isolates (Fig. 1). The remaining 10 taxa did not match any previously sequenced isolates or environmental samples with > 97% similarity (Fig. 1). Species accumulation curves were computed using rarefaction. Rarefied species accumulation curves had reached an asymptote and thus further sampling would have added few additional data and would be unlikely to have qualitatively affected the results.

Figure 1.

 Neighbour-joining phylogenetic tree showing the arbuscular mycorrhizal (AM) fungal taxa (sequence types) from mixed, pooled, root samples from Hetchell Wood. Sequences from Hetchell Wood are designated with the prefix ‘H’. Named reference sequences and closely related sequence types from environmental samples are displayed, with accession numbers, in grey. Previously unrecorded sequence types sampled from Hetchell Wood are underlined. Higher-level taxonomic classifications are shown, highlighting Glomus group A (A), Glomus group C (C), Gigasporaceae (Giga.) and Acaulosporaceae (Acau.) following Schüβler, Schwarzott & Walker (2001). The sequence types identify sequences with ≥ 3% sequence dissimilarity. Bootstrap values ≥ 75% (10 000 replicates) are shown above the branches and before the node to which they correspond. Named sequences are from studies outlined in Table 1. All the new sequences have been submitted to the EMBL Nucleotide Sequence Database (accession numbers FN556614FN556650).

Restriction fragment length polymorphism typing of screened clones was used to estimate the abundance of the 37 taxa recorded. AM fungal taxa abundance distributions were shown to fit both log-normal and broken-stick models (Fig. 2; Kolmogorov–Smirnov test; lognormal, = 0.498, =0.965; broken-stick, = 0.930, = 0.353). However, AM fungal taxa abundance distributions were significantly different to those predicted by the geometric series model (Kolmogorov–Smirnov test; = 3.255, < 0.001). Thirty-two per cent of total AM fungal taxa abundance was represented by the most abundant taxon (Fig. 1; Hetchell 22), which was notably higher than predicted by either the lognormal (19%) or broken-stick models (11%). This pattern indicates heavy dominance by the most abundant taxon in an otherwise relatively even community with a typical distribution of species abundances.

Figure 2.

 Species abundance distribution of the arbuscular mycorrhizal (AM) fungal community recorded at Hetchell Wood. Species abundance distributions are displayed as rank relative-abundance plots with relative abundance plotted on a logscale. Solid circle symbols with a solid line represent observed data, open squares show the expected distribution from a broken-stick model and open triangles the expected distribution from a log-normal model.

Structure of AM fungal communities from published studies

Thirty-two data sets, in addition to that collected in this study, were analysed from 25 studies (Table 1). Taxa abundance distributions of the AM fungal communities were shown to fit both log-normal and broken-stick models in all the data sets, but none of the analysed data sets had distributions that fitted the geometric series (Table 1). In general, taxa abundance distributions fitted both models about equally well (Table 1), although the log-normal distribution slightly outperformed the broken-stick model as indicated by an average lower Z-value (mean Kolmogorov–Smirnov test statistic; lognormal, = 0.69; broken-stick, = 0.92).

Paired comparisons of the observed proportional abundance of the most abundant AM fungal taxon (mean = 0.40, SD = 0.20) and that predicted by both log-normal (mean = 0.35, SD = 0.12) and broken-stick (mean = 0.24, SD = 0.10) models revealed that both models significantly underpredicted the abundance of the most abundant taxon across studies (paired-samples t-test; lognormal, t32 = 2.075, = 0.046; broken-stick, t32 = 5.008, < 0.001). In order to assess the numerical dominance of the most abundant taxon in AM fungal communities, ratios of abundance between the most abundant and second- (1:2) and third-most (1:3) abundant species were calculated (Fig. 3). The frequency distributions of these ratios were heavily right-skewed when data were analysed across all data sets. On average, the most abundant species was 3.5 times more abundant than the second-most abundant species and 5.6 times more abundant than the third-most abundant. However, there were some cases where the dominance of the most abundant species was far higher, shown by the tail of the distribution (Fig. 3). 1:2 and 1:3 ratios were correlated with neither species richness nor total abundance across the 32 data sets analysed (Spearman’s correlation; 1:2-richness, = −0.194, = 0.29; 1:2-abundance, = −0.091, = 0.61; 1:3-richness, = −0.299, = 0.091; 1:3-abundance, = −0.194, = 0.28) indicating a general pattern in AM fungal communities and ruling out any potential effect of sampling bias.

Figure 3.

 Frequency distributions of the ratio values of the ratio between the most abundant and second-most abundant species (top) and the most abundant and third-most abundant species (bottom) across all AM fungal studies examined (Table 1).

In order to examine whether the heavily dominant taxon in each data set belonged to a widespread generalist taxon or a locally abundant taxon, a phylogenetic analysis of the abundant sequence types was conducted (Fig. 4). Twenty-five abundant sequence types were characterized from the 33 data sets examined (32 previously published) using the commonly screened SSU rDNA region; other data sets used either the LSU or ITS genes and were not analysed due to a low sample size (Table 1). The 25 AM fungal sequences examined in this study (Fig. 4) represented taxa from Scutellospora (Gigasporaceae), Acaulospora (Acaulosporaceae) and from Glomus group A (Glomeraceae), following Schüβler, Schwarzott & Walker (2001). Of the 24 AM fungal sequences examined there were 20 distinct AM fungal taxa, indicating that there is not a small group of highly abundant widespread generalist taxa, but that individual habitats have locally dominant AM fungal taxa (Fig. 4). All of the 25 most abundant AM fungal taxa recorded in their individual habitats have been recorded elsewhere.

Figure 4.

 Neighbour-joining phylogenetic tree showing the relationship between the most abundant arbuscular mycorrhizal (AM) fungal taxon from the studies (Table 1) examined in this research. Study identities correspond to those in Table 1. Where a single study has two data sets, these are indicated by upper-case letters in brackets. Named reference sequences and closely related sequence types from environmental samples are displayed, with accession numbers, in grey. Higher level taxonomic classifications are shown, highlighting the Glomus group A (A), Glomus group C (C), Gigasporaceae (Giga.) and Acaulosporaceae (Acau.) following Schüβler, Schwarzott & Walker (2001). The sequence types identify sequences with ≥ 3% sequence dissimilarity. Bootstrap values ≥ 75% (10 000 replicates) are shown above the branches and before the node to which they correspond.

Discussion

Structure of AM fungal communities

During this study, we recorded 37 AM fungal taxa from the study site in Hetchell Wood. To date, this community of AM fungi is the richest recorded from natural ecosystems using the standard molecular methods employed in this study, even though our 30 soil cores only sampled a 0.04-m2 area of ground. Species accumulation curves were computed using rarefaction and shown to asymptote. Thus, further data collection or additional clone screening would have been unlikely to have increased our species richness estimate. In addition, this study is one of the few that explicitly sampled the AM fungal community at several times in a year. Thus, we are satisfied that we fully sampled the AM fungal community present in Hetchell Wood and that results from this study are a robust estimate for a natural AM fungal community.

Species abundance data were shown to fit both log-normal and broken-stick distributions. The log-normal distribution is a widely used null model (Magurran 2004), and species abundances from most ecological communities quantified from field studies appear to fit this distribution (Magurran 2004). However, the lognormal is a purely statistical model and both biological and statistical mechanisms have been proposed to explain this distribution (Williamson & Gaston 2005). In contrast, the broken-stick model has a similar form to the log-normal model but with an underlying biological explanation. The fit to a broken-stick model may suggest a probabilistic division of niche space between AM fungal species, where all species have approximately equal competitive abilities (Magurran 2004). The biological explanation for a log-normal SAD complements this finding, suggesting a sequential partitioning of niche space across AM fungal species (Sugihara 1980). However, both log-normal and broken-stick models heavily underpredicted the abundance of the most dominant AM fungal species, indicating additional processes are structuring these communities.

Inferring processes from pattern

Observing a particular SAD in natural communities indicates that the corresponding ecological model may explain how that community is assembled (Wilson et al. 1998). However, a good fit of species abundances to those predicted by an ecological model in a single data set is insufficient; only when this pattern has been shown to be consistent across habitats can robust conclusions about the processes structuring communities be made. The fit to log-normal and broken-stick models of AM fungal species abundances observed in the Hetchell Wood community was consistently observed across the other 32 data sets analysed. In addition, both models significantly and consistently underpredicted the abundance of the most dominant species. As with the Hetchell Wood data, a fit to the geometric series was also rejected across all data sets examined. It has been suggested that species abundances following a broken-stick distribution in species-poor communities may appear to follow a geometric series because of the low number of species (Magurran 2004). However, species richness did not affect the fit of broken-stick and log-normal models to the observed SAD in this study. The consistent results observed across data sets suggest a common process structuring AM fungal communities, independent of geographic location and habitat type, and suggest that niche differentiation and low variation in competitive ability between AM fungal species are important in structuring their communities (Sugihara 1980; Wilson et al. 1996). However, this conclusion does not explain the heavy overdominance by a single species consistently observed across data sets, nor the idiosyncratic nature of this dominance, with distinct taxa achieving very high abundance in each community.

On average, the dominant AM fungal taxon occupied 40% of the total abundance of any AM fungal community examined. This level of dominance by a single species within a community is atypical and not commonly observed in communities of higher taxa with log-normal type distributions (e.g. trees, Volkov et al. 2003; birds, Williamson & Gaston 2005; butterflies, Dumbrell & Hill 2005). Very few studies have explored SADs in natural microbial communities (Mcgill et al. 2007) and thus comparisons with microbial studies are difficult. However, a recent meta-analysis of published soil and rhizosphere bacterial communities revealed the most dominant bacterial taxon represented between 18% and 26% of the total abundance (Fierer et al. 2007), a similar level to that observed in phytoplankton communities (Venrick 1990). Although this level of dominance is greater than that typically observed in higher taxa, it is still lower than that observed in these AM fungal communities.

A similar pattern of a relatively even, log-normal type community but with a single heavily dominant species has been observed in gastro-intestinal helminth communities in marine fish (Poulin et al. 2008). Poulin et al. (2008) explain this community structure by suggesting that strong interspecific interactions between species minimize differences in abundance among the majority of species within a community, while single species may have the chance to increase their overall dominance due to high recruitment rates. Poulin et al.’s (2008) study highlights the role of stochastic events in producing a community heavily dominated by a single species that by chance had an unusually high recruitment rate. A similar stochastic process may be responsible for the observed overdominance by a single, different AM fungal taxon in each of the studies analysed in this paper, leading to the following hypothesis.

Multiple AM fungal species colonizing a network of plant roots will be subject to strong interspecific interactions, competing for limiting resources such as the plant-derived carbon on which all AM fungi depend. Niche differentiation based on the AM fungal response to the soil and host plant environment will then lead to the observed log-normal type SADs. However, not all plant roots are colonized at any given time. New roots will remain uncolonized until they contact active fungal mycelia. As many, possibly all, AM fungi are non-specific colonizers of plant roots (Smith & Read 2008), an AM fungal mycelium that can colonize previously uncolonized roots will gain substantial additional carbon. This mechanism will produce positive feedback, as the additional carbon supplied to the fungus will allow it to extend its extra-radical mycelial network and contact more uncolonized plant roots, leading to that particular species of fungus to increase in abundance within the local AM fungal community. The results from the phylogenetic analysis of the most abundant AM fungal taxon observed across studies showed that the pattern was idiosyncratic, with a different taxon achieving dominance in each case. This idiosyncratic pattern supports the conclusion that this process is largely stochastic and that the dominant AM fungal taxon is simply the species in ‘the right place at the right time’ to colonize the as yet uncolonized roots.

A potential caveat to this argument is that all data sets analysed in this study were originally compiled over short temporal scales; observed patterns are therefore snapshots of communities resulting from processes acting over longer periods of time. A striking feature of the results is that the dominant taxon of each AM fungal community analysed is different. That pattern could be explained by suggesting that local adaptation to soil chemistry and host plant communities determine the identity of the most abundant taxon. However, this suggestion does not explain why the dominant taxon is significantly more dominant than standard species abundance models predict, nor why it occupies a greater proportion of total abundance than observed in other communities, both of which may be due to the operation of stochastic processes as outlined in the hypothesis proposed above. Nevertheless, a definitive test of the hypothesis that stochastic processes play vital roles in structuring these communities will require data collected over sufficiently long timescales.

Macroecology and microbial taxa

A central aim in ecology is to examine the large-scale processes regulating species diversity (Brown & Maurer 1989) and explain the commonly observed large-scale patterns such as the species–area relationship (Harte, Kinzig & Green 1999), the hyperbolic form of SADs (Mcgill et al. 2007) and latitudinal diversity gradients (Willig, Kaufman & Stevens 2003). However, theory explaining these phenomena in microbial taxa is still relatively undeveloped (Prosser et al. 2007). In this study, we applied ecological models describing SADs to AM fungal communities and revealed a community potentially regulated by strong interspecific interactions, most likely competition for niche space, but with the potential for stochastic processes to determine the identity and abundance of the dominant taxon. However, all the studies we analysed were conducted at a similar, small spatial scale and thus the results may be scale dependent. The exact form that AM fungal SADs take at landscape or continental scales will be determined by their biogeography, species ranges sizes and an interaction between locally abundant taxa and widespread generalist species. Opik et al. (2006) showed that one Glomus type (Glo8, a closely related sequence type to Glomus intraradices) was the most frequently detected in an analysis of a number of published studies, and that several widespread generalist species were present in a range of habitats, colonizing a number of different plant species. However, these generalist species are not always the most abundant taxon in any single habitat (Fig. 4; Opik et al. 2006). This pattern of locally abundant taxa but with widespread species present in a sample is likely to generate a classic log-normal distribution at the landscape scale, due to the multiplicative effects of independent data sets and the consequences of the central-limit theorem (May 1975), but without the overdominance observed from small-scale samples of AM fungal communities. Indeed, subtle changes in the exact form of SADs when analysed at different spatial scales have been observed in both plant and animal taxa (Anderson & Mouillot 2007; Fraser et al. 2008). Thus, in order to understand fully the processes regulating the structure and diversity of microbial communities, further theory specific to microbial taxa is required (Prosser et al. 2007) and explicit examinations of large-scale biogeographical patterns and local-scale community and population dynamics are needed.

Conclusions

This study applies theoretical models describing species abundances to microbial communities. Results suggest that a common log-normal type SAD is observed for both microbial and macroorganism communities. However, due to their biology, some noticeable differences such as heavy overdominance, may lead to variations in the structure of microbial communities. In AM fungal communities from different habitats the consistent fit to log-normal and broken-stick models suggests the strong influence of niche differentiation structuring these communities. However, the consistently observed overdominance by a different taxon in each study may indicate that both local adaptation and stochastic processes play important roles in structuring these communities. This research not only highlights the importance of ecological models derived from studies of macroorganisms in studying microbial communities, but also the need for large-scale microbial data sets if we are to understand the common patterns and processes regulating global biodiversity.

Acknowledgements

We would like to thank Natural England and the Yorkshire Wildlife Trust for permission to work at Hetchell Wood, and Naveed Aziz and Celina Whalley of the University of York Genomics Facility for technical advice and assistance. We also thank Jason Hoeksema and an anonymous referee for their constructive comments and Mark Williamson for helpful discussion on species abundance distributions. This study was funded by a Natural Environment Research Council (http://www.nerc.ac.uk/) grant (NE/D01090X/1).

Ancillary

Advertisement