Evolutionary criteria outperform operational approaches in producing ecologically relevant fungal species inventories

Molecular Ecology (2011) 20, 655–666 doi: 10.1111/j.1365-294X.2010.04964.x

Evolutionary criteria outperform operational approachesin producing ecologically relevant fungal speciesinventories

JEFF R. POWELL,* MICHAEL T. MONAGHAN,† MAARJA OPIK‡ and MATTHIAS C. RILLIG*

*Institut fur Biologie, Okologie der Pflanzen, Freie, Universitat Berlin, Altensteinstr. 6, 14195 Berlin, Germany, †Leibniz-

Institute of Freshwater Ecology and Inland Fisheries, (IGB), Muggelseedamm 301, 12587 Berlin, Germany, ‡Department of

Botany, Institute of Ecology and Earth, Sciences, University of Tartu, 40 Lai St., 51005 Tartu, Estonia

Corresponde

E-mail: jeffpo

� 2010 Black

Abstract

Analyses of the structure and function of microbial communities are highly constrained

by the diversity of organisms present within most environmental samples. A common

approach is to rely almost entirely on DNA sequence data for estimates of microbial

diversity, but to date there is no objective method of clustering sequences into groups

that is grounded in evolutionary theory of what constitutes a biological lineage. The

general mixed Yule-coalescent (GMYC) model uses a likelihood-based approach to

distinguish population-level processes within lineages from processes associated with

speciation and extinction, thus identifying a distinct point where extant lineages became

independent. Using two independent surveys of DNA sequences associated with a group

of ubiquitous plant-symbiotic fungi, we compared estimates of species richness derived

using the GMYC model to those based on operational taxonomic units (OTUs) defined

by fixed levels of sequence similarity. The model predicted lower species richness in

these surveys than did traditional methods of sequence similarity. Here, we show for the

first time that groups delineated by the GMYC model better explained variation in the

distribution of fungi in relation to putative niche-based variables associated with host

species identity, edaphic factors, and aspects of how the sampled ecosystems were

managed. Our results suggest the coalescent-based GMYC model successfully groups

environmental sequences of fungi into clusters that are ecologically more meaningful

than more arbitrary approaches for estimating species richness.

Keywords: arbuscular mycorrhizal fungi, biodiversity, coalescent, GMYC model, operational

taxonomic unit

Received 2 June 2010; revision received 23 September 2010; 8 November 2010; accepted 14 November 2010

Introduction

Species form the systematic basis of biodiversity and

delineating species remains a foundational issue in biol-

ogy. The task is particularly challenging for microbial

ecologists because of the enormous diversity present

within most environmental samples, the difficulty in-

readily assigning cryptic morphological characters to

species-level differences, and the more fundamental

lack of a widely applicable concept of microbial species.

nce: Jeff R. Powell, Fax: ++49 (0)30 838 53886;

[email protected]

well Publishing Ltd

The vast majority of microbial biodiversity cannot be

cultured in the laboratory (Liesack & Stackebrandt

1992), and a common approach to dealing with this

problem is to rely almost entirely on DNA sequence

data for biodiversity estimates, where variation at one

or a few genetic loci is used to characterize whole

microbial communities in culture and in the environ-

ment. Analogs of species richness are estimated by

assigning sequences to operational taxonomic units

(OTUs) based on some level of sequence similarity, a

methodology that arose from early empirical studies by

bacterial taxonomists linking percentage sequence

similarity to DNA–DNA hybridization percentages

656 J . R . POWELL ET AL.

(Stackebrandt & Goebel 1994; Hanage et al. 2006) and

that was then extrapolated to other microbial taxa.

Criteria based on sequence similarity can have lim-

ited meaning where rates of sequence divergence are

nonuniform across clades and where inaccurate a priori

groups have been used to calibrate thresholds (e.g.

97%) of sequence divergence (Gevers et al. 2005;

Hanage et al. 2006; Vogler & Monaghan 2007; Koeppel

et al. 2008; Fraser et al. 2009). Taxonomists and evolu-

tionary biologists have recently begun to apply phylog-

enetics and population genetics theories to the

delineation of species boundaries using DNA

sequences, but the integration of these tools into envi-

ronmental surveys of microbial biodiversity has been

limited. Using data sets that combine both population-

and species-level genetic polymorphism, the general

mixed Yule-coalescent (GMYC) model (Pons et al. 2006;

Fontaneto et al. 2007) identifies evolutionarily indepen-

dent clusters of DNA sequences by differentiating

branches within clusters, associated with neutral coales-

cent (merging of lineages within a population) pro-

cesses (Kingman 1982), from branches between clusters,

associated with speciation events (Yule 1925). The point

of highest likelihood of this mixed model estimates the

switch from speciation to coalescent processes and can

thus be interpreted as the species boundary. A more

recent version of the GMYC model allows for multiple

lineages to each have their own transition threshold

within a single phylogenetic tree (Monaghan et al.

2009). The approach has been proposed as a means to

delineate species in a variety of metazoan (Pons et al.

2006; Ahrens et al. 2007; Fontaneto et al. 2007; Jousselin

et al. 2009; Monaghan et al. 2009; Papadopoulou et al.

2009; Bode et al. 2010), protozoan (Leliaert et al. 2009),

and bacterial (Barraclough et al. 2009; Jousselin et al.

2009) groups.

The GMYC method may be a more appropriate tool for

estimation of diversity from environmental DNA

sequence data due to its grounding in evolutionary the-

ory (Pons et al. 2006; Fontaneto et al. 2007). However, a

direct comparison of this and more arbitrary approaches

to predict biologically-relevant patterns is lacking. For

instance, the GMYC method may be expected to perform

better than more arbitrary approaches in its ability to pre-

dict biologically meaningful groups that differ in ecologi-

cal function or behaviour, especially involving traits that

are likely to have been derived recently (i.e. between clo-

sely related species). Such traits may include colonization

of divergent ecosystem types within a landscape (Koep-

pel et al. 2008) or interaction with an obligate symbiont

(Jousselin et al. 2009).

The arbuscular mycorrhizal (AM) fungi make up a

group of microorganisms for which an objective method

for clustering sequences into putative species is critical,

because no unambiguous method for quantifying diver-

sity from DNA sequence data has emerged. Obligate

symbionts of plants, the group associates with species in

70–90% of terrestrial plant families (Smith & Read 2008).

Their demonstrated importance for promoting plant

diversity and productivity in some systems (van der

Heijden et al. 1998; Klironomos et al. 2000) has triggered

several attempts to characterize AM fungal diversity in a

variety of natural and managed systems (Opik et al.

2006, 2010). Functional variation among AM fungi exists

at various taxonomic and phylogenetic levels (Hart &

Reader 2002; Koch et al. 2006; Maherali & Klironomos

2007; Powell et al. 2009; Sikes et al. 2009; Wehner et al.

2010), suggesting that evolutionary and ecological trade-

offs have contributed to the diversification of this group.

In addition, there is clear evidence that the structure and

function of AM fungal communities vary in response to

the chemical and biotic characteristics of their environ-

ment (Johnson et al. 2005; Johnson 2010).

Earlier surveys of AM fungal diversity in environ-

mental samples focused on extraction and identification

of spores from environmental samples (e.g. Johnson

et al. 1992). Given that species descriptions are based

primarily on spore morphology (Schenck & Smith

1982), this approach was well suited to utilize the exist-

ing AM fungal taxonomic framework. However, sur-

veys have increasingly relied on DNA sequence-based

approaches to estimate AM fungal diversity in environ-

mental root and soil samples, and there are several

issues that prevent the use of a strictly taxonomic

approach to estimate diversity. Most importantly, taxo-

nomic descriptions of AM fungi are likely biased

toward those species that produce abundant and/or

long-lived spores. Compounding this issue is the fact

that studied isolates of AM fungi are usually obtained

from environmental samples by culturing with one of

only a few host plant species, resulting in a significant

cultivation bias (Sykorova et al. 2007). In addition, there

have been clades of AM fungi detected in environmen-

tal surveys for which no comparable DNA sequence

from a described species exists, and many of these

fungi have not been and may never be observed (Hibb-

ett et al. 2009). As a result, AM fungal biodiversity sur-

veys from environmental DNA samples have largely

relied on methods based on phenetics (i.e. sequence

similarity) and are likely to benefit from more objective

and biologically meaningful clustering approaches.

In this study, we evaluated the ability of the GMYC

model to detect evolutionarily informed clusters (here-

after referred to as GMYC groups) from AM fungal 18S

rDNA sequence data collected during two regional

surveys: (i) northern boreal forest and dry meadow

plots in Estonia and (ii) neotropical primary forest and

plantation plots in Ecuador. We compared estimates of

� 2010 Blackwell Publishing Ltd

SPECIES DELINEATI ON IN FUNGAL INVENTORIES 657

species richness derived from these GMYC groups to

those derived by assigning sequences to OTUs based on

traditional limits in sequence similarity. Then, we com-

pared the ecological relevance of the approach by test-

ing whether groups could account for variation in

putative niche-based parameters associated with the

spatial distributions. Our analyses suggest that discrete

shifts in branching rates in single-locus phylogenetic

trees can be used to obtain evolutionarily and ecologi-

cally relevant estimates of fungal richness in environ-

mental surveys. In addition, they suggest that species

groups predicted by the GMYC model can have greater

ecological relevance than more arbitrary distance-based

approaches, which will help researchers addressing

pressing questions in fungal biodiversity studies. For

instance, our results provide insight into how host-plant

specialization and ecosystem management influence

AM fungal species distributions.

Materials and methods

Obtaining DNA sequences

To address whether the GMYC model could be used to

delineate lineages of AM fungi, we used 18S rRNA

gene sequences obtained from environmental surveys in

Estonia (Opik et al. 2003, 2008) and Ecuador (Haug

et al. 2010). These studies were chosen because they

had sufficient sampling intensity across the breadth of

much of the phylum, and they provided estimates of

environmental parameters at the sample level. Briefly,

DNA was extracted from roots of various host species

associated with the sampling locations. Seven sites in

Estonia were sampled. The sites encompassed three dif-

ferent ecosystem types (boreal forest, forest border, dry

meadow) and boreal forest sites were subjected to two

levels of management intensity. Selective harvesting in

low-intensity plots resembled mature, old-growth for-

est. High-intensity management was represented by

clear-cutting <30 years prior to sampling and replanting

with Norway spruce. The original authors used the

NS31/AM1 primer set (Helgason et al. 1998) to amplify

a 550-bp central fragment of the 18S rRNA gene with a

high level of specificity to AM fungal DNA. Samples

from Ecuador were taken from 15 neotropical primary

forest plots, three reforested pasture sites, and seven

nursery plants (Kottke et al. 2008; Haug et al. 2010).

The authors used a series of primer sets, each of which

amplifies 18S rDNA from a different AM fungal taxon

in a nested set of reactions (Glomus Group A: SSU128/

SSU1536IH then SSU300/GLOM1310rc; Glomus Group

B: SSU817/NS8 then SSU817/LETC1670rc; Acaulospora-

ceae: SSU817/NS8 then SSU817/ACAU1660rc; Archaeo-

sporales: SSU817/SSU1536IH then SSU817/ARCH

� 2010 Blackwell Publishing Ltd

1375rc). Amplicons are of variable length, but all con-

tain an overlapping region of approximately 530 bp at

the 5’-end of the 18S rRNA gene (based on the

sequences we analyzed). This amplicon lies down-

stream of the NS31/AM1 amplicon and does not

overlap with it. Further details on the isolation, amplifi-

cation, and sequencing of AM fungal DNA are

described in the original publications.

Species delineation by sequence clustering

For our analyses, sequences were aligned using MUSCLE

v.3.6 (Edgar 2004). Prior to phylogenetic reconstruction,

the most appropriate models of evolution were esti-

mated from the alignments using ModelTest 3.7

(Posada & Crandall 1998); a GTR nucleotide substitu-

tion model with gamma-distributed substitution rates

was selected in each case; we included an estimated

proportion of invariant sites for the Estonia data. Dupli-

cate haplotypes were removed using the ‘unique.seqs’

command in mothur (Schloss et al. 2009), after which

tree topology and branch lengths were estimated using

a Bayesian search and a strict molecular clock as imple-

mented in BEAST version 1.5.3 (Drummond & Rambaut

2007). In a study of five insect taxa, Monaghan et al.

(2009) observed little variation in the number of groups

predicted by the GMYC model when using a relaxed or

strict molecular clock and a coalescent prior; however,

the use of a Yule prior resulted in large differences

compared with a coalescent prior. A coalescent prior

was used in our search to allow comparision with the

GMYC null model in which all sequences are derived

from a single population coalescent. Demographic his-

tory is likely to vary among lineages, and we assumed

a constant population size as the most conservative esti-

mate across all lineages. Model parameters were esti-

mated from the data, and a UPGMA tree was used to

start the chain. We reconstructed phylogenetic trees

from all unique sequences obtained in each of the Esto-

nia (577 aligned bp) and Ecuador (532 aligned bp) sur-

veys separately. In addition, to compare our richness

estimate to that of Haug et al. (2010), which was only

estimated for Glomus group A, we reconstructed the

phylogeny of 134 unique Glomus A sequences (1113

aligned bp). The MCMC chain was run for 3 · 107 gen-

erations for the Estonia tree (395 unique sequences) and

for 1 · 107 generations for the Ecuador trees (137 and

134 unique sequences); chain length was chosen with

regard to the number of sequences in each data set.

Trees were sampled every 1000 generations from the

MCMC chains, and the final 10 000 (Estonia) or 9000

(Ecuador) trees of each chain (each burn-in was chosen

to ensure sufficient effective sample size for the esti-

mated parameters and to remove initial trees of low

658 J . R . POWELL ET AL.

likelihood from the chain) were analyzed with TREEAN-

NOTATOR (Drummond & Rambaut 2007), using a poster-

ior probability limit of 0.5 and targeting the maximum

clade credibility tree while retaining node heights.

The GMYC model was optimized to these clock-

constrained trees using the GMYC script, available as

a tool within the ‘SPLITS’ package (Ezard et al. 2009)

for R (R Development Core Team 2009). The proce-

dure uses functions from the ‘ape’ library (Paradis

et al. 2004). Briefly, the procedure examines the distri-

bution of waiting times between diversification events

using a likelihood-based approach. The model has two

components (the neutral coalescent model and the

Yule model) for estimating the likelihood of waiting

times (Pons et al. 2006; Fontaneto et al. 2007). Each

component has two parameters that are estimated

from the data (a branching parameter and a scaling

parameter); the fifth parameter indicates the threshold

transition point between the two-model components.

A modified version of the model permits the age of

the transition point to vary among lineages (multiple

thresholds; Monaghan et al. 2009). For the Ecuador

data, we performed the analysis without removing the

outgroup from the tree; we did this to improve the

power of the algorithm to estimate parameters associ-

ated with the Yule portion of the model, as the num-

ber of early branching events (i.e. divergences between

species and higher order taxa) was relatively low for

this tree. Likelihood ratio (LR) tests were used to eval-

uate the significance of the single-threshold model to

the null model, and the multiple-threshold model to

the single-threshold model.

To compare diversity estimated by the GMYC

approach to that estimated by OTU approaches, we used

the approaches for OTU delineation that were used in the

original analyses (Opik et al. 2003; Haug et al. 2010). For

the Estonia data, sequences were grouped into OTUs on

the basis of bootstrap support and sequence similarity of

‡97% based on a neighbor-joining tree. We used the

actual OTU definitions originally determined in Opik

et al. (2003) to facilitate comparison of the approaches.

For the Ecuador data, OTUs (‘sequence types’ in the ori-

ginal publication) were obtained by forming groups of

sequences based on 99% (as in the original publication)

or 97% similarity, using the furthest-neighbor method in

mothur (Schloss et al. 2009). Species accumulation curves

were derived from rarefaction using the ‘vegan’ package

(Oksanen et al. 2010) in R.

Explaining variation in environmental axes

We used redundancy analysis to estimate the amount

of variation in parameters associated with putative

niches that could be explained by the identity of

genetic clusters. This approach is analogous to the par-

titioning of variance between- and within-groups in

standard analysis of variance. Analyses were per-

formed using the ‘ade4’ (Dray & Dufour 2007) and

‘vegan’ packages in R. For the Estonia survey, a mix

of qualitative (host plant species and stand age cate-

gory) and quantitative variables (soil pH, N, P, organic

C) made up the columns of the table; further details

are available in Opik et al. (2008). Individuals (rows)

in the table were limited to 866 sequences obtained

from the study published in Opik et al. (2008), as the

relevant environmental data had been collected for the

samples from which these sequences were obtained.

We used a combination of normed principal compo-

nents analysis and multiple correspondence analysis

(MCA) in the AM fungal ordination (Hill & Smith

1976; Dray & Dufour 2007). For the Ecuador survey,

the 197 sequences were annotated with information on

the host species and ecosystem type (neotropical pri-

mary forest, plantation, nursery) from which they were

obtained. We used MCA to transform the ecosystem

type classes into quantitative variables prior to redun-

dancy analysis; we did not include the host data in

the analysis because of the small number of sequences

associated with each host species (median ¼ 3). For

each model, we calculated the Akaike information cri-

terion corrected for small sample size (AICc; McQuar-

rie & Tsai 1998) using a multivariate analogue of

residual sums of squares, calculated as the sum of all

eigenvalues minus the sum of the canonical eigen-

values of Y (the matrix of dependent variables) on X

(the matrix of explanatory variables), described by

Legendre & Anderson (1999).

Results

Detection of GMYC groups

In the gene trees from the Estonia and Ecuador surveys,

we observed evidence of a discrete threshold at which

the time between branching events changes (Fig. 1a,d),

identifying a possible species boundary. A single-

threshold GMYC model provided a better fit to the data

than a null model of uniform coalescent branching

across the entire tree in each case (Estonia: LR ¼ 24.2,

d.f. ¼ 3, P < 0.001; Ecuador: LR ¼ 7.3, d.f. ¼ 3,

P ¼ 0.06). Allowing the Yule-coalescent threshold to

vary among lineages (Monaghan et al. 2009) did not

significantly improve the fit of the model when com-

pared to the single-threshold fit (Estonia: v2 ¼ 10.8,

d.f. ¼ 12, P ¼ 0.54; Ecuador: v2 ¼ 6.8, d.f. ¼ 9, P ¼0.66), suggesting that the age of the species boundary

did not vary significantly among the lineages repre-

sented in these gene trees.

� 2010 Blackwell Publishing Ltd

12

51020

50100200

500

Num

ber

of li

neag

es

(a)

–0.7 –0.5 –0.3 –0.1

3560

3570

3580

3590

3600

3610

Time

Like

lihoo

d

(b)

Glomus Ab

Glomus Aa

Gigasporaceae Acaulosporaceae Diversisporaceae

(c)

1

2

5

10

20

50

100

Num

ber

of li

neag

es

(d)

–1.0 –0.6 –0.2

1126

1128

1130

1132

1134

1136

Time

Like

lihoo

d

(e)

Glomus B Gigasporaceae

Acaulosporaceae

Paraglomeraceae

Archaeosporales

Glomus A

(f)

Fig. 1 Lineage-through-time (LTT) plots (a,d) and likelihood profiles of the single-threshold general mixed Yule-coalescent (GMYC)

model (b,e) fit to the evolutionary reconstruction of arbuscular mycorrhizal fungi detected from field surveys in Estonia (c) and Ecua-

dor (f). The LTT plots indicate an increase in branching rate (¼ reduction in waiting times between branching events) near the tips.

The solid vertical line in each plot indicates the threshold time for the maximum likelihood GMYC model transition, which suggests

a shift from phylogenetic Yule processes to coalescent population processes. The dashed vertical lines indicated the confidence inter-

val for this estimate (within two log-likelihood units of the maximum). Time is scaled from )1 (divergence of the ingroup and

outgroup) to 0 (time of sampling).

SPECIES DELINEATI ON IN FUNGAL INVENTORIES 659

� 2010 Blackwell Publishing Ltd

660 J . R . POWELL ET AL.

When we limited the analysis in the Ecuador survey

to only the Glomus A clade, as was done in the original

study (Haug et al. 2010), the GMYC model was not a

significantly better fit to the data than the null model of

a single coalescent population (single-threshold: LR ¼0.6, d.f. ¼ 3, P ¼ 0.90; multiple-threshold: LR ¼ 5.8,

d.f. ¼ 15, P ¼ 0.57), even though the maximum likeli-

hood model predicted the same number of groups

within this clade (four) as did the GMYC model fit to

the entire gene tree.

Comparison of GMYC and OTU richness

The species accumulation curve suggests that the sam-

pled GMYC group richness approached the asymptotic

level of richness in both surveys (Fig. S1, Supporting

information).

Application of the model to the Estonia survey

resulted in 40 GMYC groups. This was fewer than the

45 OTUs estimated using a 97% similarity criterion,

although the confidence interval (within two log-likeli-

hood units of the maximum likelihood; Fig. 1b) over-

lapped this estimate (Table 1). The large majority of

differences were attributed to the estimated time of the

transition from between- to within-species divergences

predicted by the maximum likelihood model, although

a few differences in the grouping of sequences resulted

from differences in tree topology between the two

approaches (Fig. S2, Supporting information).

The GMYC model predicted a relatively low level of

species richness in the Ecuador survey (11 groups,

Table 1 Predicted species richness in arbuscular mycorrhizal fung

accounting for sequence similarity (OTU) or fitting a coalescent-sp

between branching events (GMYC)

Location Clade K* GMYC

Estonia Glomus Aa 157 20

Glomus Ab 201 9

Acaulosporaceae 21 6

Diversisporaceae 9 2

Gigasporaceae 6 3

Total 394 40

Ecuador Glomus A 91 4

Glomus B 5 1

Acaulosporaceae 20 1

Archaeosporales 15 3

Gigasporaceae 3 1

Paraglomeraceae 2 1

Total 136 11

nd, not determined; GMYC, general mixed Yule-coalescent; OTU, ope

*Number of unique sequences associated with each taxon.

†Confidence interval on richness estimate (within two log-likelihood u

‡OTU methods use either a 97% or 99% cutoff in sequence similarity

Table 1, Fig. S3, Supporting information), especially

when compared to the 99% similarity threshold used in

the original analysis (33 groups, Table 1). Across almost

all taxa (identified from the determined affiliations in

the source publications), the estimated number of

GMYC groups was more similar to the number of

OTUs estimated using a 97% similarity criterion and

significantly fewer than OTU estimates at 99% similar-

ity. The only exception was the Glomus A clade, for

which the maximum likelihood GMYC estimate agreed

with the OTU estimate at 97% similarity but also had a

large confidence interval, overlapping the estimate at

99% similarity.

Ability of clustering methods to account for nichevariation

Classifying groups of sequences according to the GMYC

model better accounted for variation in putative niche-

based parameters than using groups based on OTUs

(Table 2). For the Estonia survey, classes associated

with the maximum likelihood GMYC model accounted

for a similar proportion of variation in host species,

edaphic factors, and forest stand age as did OTUs

(‡97% similarity) but using fewer parameters (classes

of sequences) and, therefore, performed much better at

partitioning variation among classes according to the

model selection information criterion (dAICc ¼ 13.1).

For the Ecuador survey, assigning sequences that

share ‡99% similarity to common OTUs accounted for

the most variation in the ecosystem type they were

al communities, sampled during two environmental surveys,

eciation model to the temporal distribution of waiting times

(CI†) OTU (97%)‡ OTU (99%)‡

(16–26) 25 nd

(7–14) 8 nd

(5–6) 7 nd

(1–3) 2 nd

(1–3) 3 nd

(30–52) 45 nd

(4–14) 4 11

(1–1) 2 2

(1–3) 2 7

(2–6) 3 9

(1–1) 1 2

(1–1) 2 2

(10–26) 14 33

rational taxonomic unit.

nits of maximum likelihood model).

.

� 2010 Blackwell Publishing Ltd

Table 2. Ability of species clustering approaches to account for variation in putative-niche variables in two environmental surveys

of arbuscular mycorrhizal fungi, expressed as a proportion of the total variation

Location Niche axes Model Number of entities Variance explained† AICc‡

Estonia Edaphic factors, host species, forest stand age GMYC 28 0.106** 3170.1

OTU (97%) 34 0.105** 3184.0

Ecuador Ecosystem type GMYC 11 0.141* 576.3

OTU (97%) 14 0.115ns 588.9

OTU (99%) 33 0.275ns 599.2

Significance of including classes as a predictor in the model (**P £ 0.01, *P £ 0.05, nsnot significant).

GMYC, general mixed Yule-coalescent; OTU, operational taxonomic unit.

†Calculated as the constrained inertia divided by the total inertia, a multivariate analog to the coefficient of determination (R2), with

no correction for the number of groups.

‡Akaike Information Criterion corrected for small sample size (McQuarrie & Tsai 1998); model with the lowest value (indicated in

bold) has the best fit to the data.

SPECIES DELINEATI ON IN FUNGAL INVENTORIES 661

sampled from, but at a high cost associated with the

large number of classes. Using classes associated with

the maximum likelihood GMYC model most efficiently

accounted for this variation (dAICc ¼ 12.6 against OTUs

at 97%; dAICc ¼ 22.9 against OTUs at 99%).

Discussion

These analyses for the first time highlight the potential

utility of the GMYC approach to provide ecologically,

in addition to evolutionarily, meaningful differentiation

between AM fungal sequence types in environmental

surveys. We demonstrated that classes identified using

the analysis of branching patterns isolated differences

in the ecological distributions of the groups better than

a priori thresholds of divergence traditionally used.

This was the case for estimates of the niche associated

with the symbiotic host, edaphic factors, and aspects of

how the ecosystems were managed.

Although the groups predicted by the GMYC model

provided the best fit to the environmental data associ-

ated with the samples, the variance explained by clus-

tering sequences into GMYC groups represented a

small proportion (<15%) of the total variance in both

surveys. While it is possible that the environmental fac-

tors that are ultimately responsible for defining niche

breadth in AM fungi were not measured in these stud-

ies, two additional hypotheses based on our knowledge

of AM fungal ecology may explain this observation.

First, at least some of the isolates we detected may

belong to generalist taxa with regard to their ecological

niches. For instance, individual GMYC groups con-

tained isolates that were associated with multiple plant

species; this was also the case for the other clustering

approaches (Opik et al. 2003,2008; Haug et al. 2010).

While the discussion of niche breadth in AM fungi usu-

ally focuses on the degree of functional host specificity

� 2010 Blackwell Publishing Ltd

in AM fungi (Hart et al. 2003; Opik et al. 2009), it has

also been addressed in relation to environmental vari-

ables driving AM fungal distributions and community

structure from local (Scheublin et al. 2004; Mummey

et al. 2005) to global (Rosendahl et al. 2009; Opik et al.

2010) scales. These studies have shown that at least

some AM fungal taxa have wide geographic distribu-

tions and exhibit broad environmental tolerances.

Second, studies of certain morphospecies have dem-

onstrated considerable intraspecific variation in toler-

ance to environmental factors and in symbiotic function

(as opposed to behaviours and tolerances that are

shared by many or all members of a generalist species).

Within three Glomus species, intraspecific variation in

AM fungal growth (extraradical hyphal production)

and symbiotic functioning (phosphorus uptake and host

plant growth parameters) was observed by Munkvold

et al. (2004). Functional variation in nutrient uptake and

response to heavy metal contamination was observed

among isolates of Glomus mosseae (Biro & Takacs 2007;

Biro et al. 2009). As these responses were evaluated in

‘common garden’ studies, differences among isolates

may have been attributable to the intrinsic fitness and

symbiotic capacity of each isolate and/or variable

responses to the environment in which they were

grown (in terms of the host plant or the abiotic condi-

tions). Both genetic and environmental contributions

were observed in a series of studies using isolates from

a single population of Glomus intraradices; investigators

observed variation among isolates in functional parame-

ters, including production of hyphae and spores, effects

on plant growth, responses to changing nutrient condi-

tions and host species, and observed frequency coloniz-

ing different host species in the field (Koch et al.

2004,2006; Croll et al. 2008; Ehinger et al. 2009). While

the isolates used in each of these studies are likely to

represent a biased sample (Sykorova et al. 2007), they

662 J . R . POWELL ET AL.

provide clear evidence for ecological differences within

certain lineages of AM fungi that contribute to niche

breadth observed at the species level.

There are some peculiar aspects of genetic systems in

the Glomeromycota that may contribute to this large

proportion of environmental variation that remains

unexplained, but that may also bias the GMYC method

(as well as similarity-based approaches using DNA for

species delineation). Most studies have found evidence

for clonal reproduction within populations of AM fungi

(e.g. Rosendahl & Taylor 1997), although recent studies

suggest that recombination may occur to a certain

extent (Croll et al. 2009; Croll & Sanders 2009). Transfer

of nuclei has been observed via hyphal anastomoses

between genetically distinct individuals of the same

species, but not different species and at reduced fre-

quencies relative to within-isolate transfers (Giovannetti

et al. 1999; Croll et al. 2009). This should facilitate the

formation of genetic clusters within species and, to a

certain extent, within populations. Irregardless, the

GMYC approach is not limited to sexually reproducing

clades; Fontaneto et al. (2007) used this approach to

identify genetic clusters within asexual rotifers. Another

matter to consider is that Jany & Pawlowska (2010)

observed nuclear dynamics in Glomus etunicatum in

which spores were populated by multiple nuclei

derived from the surrounding mycelium as opposed to

copies of a single founder nucleus. Markers may also

be polymorphic within individual nuclei (Pawlowska &

Taylor 2004). These processes contribute to polymorphic

rDNA within individual fungal isolates that could

result in an individual being classified into multiple

genetic clusters (Clapp et al. 1999, 2001) and overesti-

mation of species richness. How to deal with this intra-

individual polymorphism in taxonomic studies and

environmental surveys is an open question (Rodriguez

et al. 2004).

In any case, the relevant conclusion to draw from

these data is that the identity of groups to which the

isolated DNA belongs explained a statistically signifi-

cant proportion of the variation in their environmental

distributions (i.e. their niches) and that evolutionarily

informed clustering performed best among the

approaches tested here. We do not expect this to be the

case for all axes of the niche. For example, AM fungi

exhibit strong phylogenetic niche conservatism with

regard to colonization patterns within a host plant, with

less variation in colonization strategy among members

of the same family, and sometimes the same order

(Powell et al. 2009). Indeed, circularity will remain a

problem for any such comparison of methodological

‘accuracy’. More fine-scaled criteria can be employed in

systems where the specific drivers of species distribu-

tions and associations are better understood [e.g. host-

bacterial co-evolution in aphids (Jousselin et al. 2009)],

but the coalescent-speciation approach employed here

is more likely to be applied to broad-scale environmen-

tal studies and, in the absence of other data, provides a

more evolutionarily meaningful approach to clustering

sequences than similarity-based proxies.

Some preliminary information on the richness and

evenness of species within an ecosystem will be helpful

to design an appropriate sampling strategy. The GMYC

model requires a sample of sequences at both popula-

tion and phylogenetic levels of variation to detect speci-

ation-coalescent shifts in branching rates and, while the

necessary sample sizes for each level are difficult to

determine, our results provide some insight when

viewed in comparison with previous studies. As dis-

cussed elsewhere (Monaghan et al. 2009), 7–10 geno-

types appears to be an empirically and theoretically

appropriate sample size for capturing coalescent-level

variation, although the model performs consistently

well with smaller sample sizes (Pons et al. 2006; Ahrens

et al. 2007; Fontaneto et al. 2007; Barraclough et al.

2009; Jousselin et al. 2009; Leliaert et al. 2009; Mona-

ghan et al. 2009; Bode et al. 2010). At deeper nodes (i.e.

speciation branching), our results suggest that a mini-

mum number of coalescent groups may be needed in a

data set for the model to delineate groups that agree

with other criteria of independence. Limiting the analy-

sis of sequences from Ecuador to only the Glomus A

clade (the 1113-bp alignment), as was done in the origi-

nal study (Haug et al. 2010), the GMYC model pre-

dicted the same number of groups within this clade as

the larger analysis but was not a significantly better fit

to the data than the null model of a single coalescent

population, probably because of a lack of analytical

power. Limiting the sampling strategy to the Glomus A

clade was therefore not entirely suitable for GMYC

analysis, which could have important consequences for

study design and data interpretation. Further work,

including simulation studies, should lead to appropriate

sampling strategies.

This subsequent analysis did present an opportunity

to test whether the high similarity threshold used to

delimit OTUs in some studies was suitable for these

data. Adjusting the speciation-coalescent threshold (T)

to yield 43 GMYC groups (i.e. the number of Glomus A

OTUs using a ‡99% similarity threshold in this subse-

quent analysis; data not shown) decreased the model

likelihood by 2.2 log-likelihood units. This difference is

greater than the confidence interval and indicates a

poor fit to the observed branching pattern relative to

the best-fit model. Therefore, studies using such high

thresholds of sequence divergence to delineate OTUs

are likely overestimating AM fungal species richness.

On the other hand, our results suggest that a ‡97%

� 2010 Blackwell Publishing Ltd

SPECIES DELINEATI ON IN FUNGAL INVENTORIES 663

similarity threshold can provide a useful preliminary

assessment of AM fungal diversity based on 18S rRNA

gene sequences with which to plan further sampling

efforts using evolutionarily informed clustering.

We employed this evolutionarily informed approach

under circumstances where the species-level taxonomic

identities are largely unknown and the context of a

phylogenetic species is impractical; we envision this as

the primary use of this tool by microbial ecologists.

However, many environmental microbiologists are

interested in the detection of particular species from

environmental samples, such as agriculturally or medi-

cally important pathogens and related species that may

evolve to become important. Here, it is necessary to

establish a clear taxonomy in these environmental sam-

ples and approaches based on phenetics are not always

useful. Bacterial taxonomists rely primarily on DNA-

based approaches, such as DNA–DNA hybridization

and multilocus sequence analysis (Hanage et al. 2006),

to delineate species; this often involves but is not lim-

ited to phenetics-based approaches. Fungal taxonomists

rely to a greater extent on the identification of fixed dis-

tances in morphological, reproductive, and molecular

characters (e.g. Morton & Bentivenga 1994), and phenet-

ics-based approaches with fixed cutoffs are clearly not

suitable in this context (Seifert et al. 2007; Nilsson et al.

2008). The peculiarities of AM fungal genetics, as

described above, highlight the pitfalls of employing an

entirely DNA-based approach to resolving an AM fun-

gal taxonomy. However, there are many fungi that have

not been cultured from, or even observed in, their natu-

ral environment and DNA-based approaches are the

only methods available for classifying these individuals

into putative species (Hibbett et al. 2009). Therefore, it

is possible that the GMYC method, given its lack of reli-

ance on fixed and somewhat arbitrary cutoffs, may be

of value from a taxonomic perspective for those organ-

isms where the primary taxonomic method is based on

clustering of similar DNA sequences, especially when

the organisms themselves are not available for study.

Environmental scientists are frequently called upon to

evaluate the community-level responses of AM fungi

and other microorganisms to environmental change as

a proxy for ecosystem-level impacts. Advances in me-

tagenomics and bioinformatics ensure that DNA-based

approaches will continue to dominate surveys that eval-

uate environmental impacts on microbial communities.

The GMYC approach represents an objective method of

clustering sequences into evolutionarily independent

groups and estimating diversity in the presence of cryp-

tic species and intraspecific morphological variation.

Here, we have shown that this approach can also better

identify ecologically meaningful groups than using arbi-

trary levels of sequence similarity; this result in particu-

� 2010 Blackwell Publishing Ltd

lar lends support to its use within DNA-based

environmental surveys. To enhance its utility for

quantifying microbial diversity, especially using next-

generation sequencing, priority should be given to the

accurate estimation of branch lengths in large phylo-

genetic trees.

Acknowledgements

We thank Tancredi Caruso for statistical advice and Benjamin

Sikes, Stefan Hempel, Ingeborg Haug, Dirk Redecker, and

three anonymous reviewers for helpful comments on this

manuscript. This work was supported by funding from the

7th European Community Framework Programme (PIIF-GA-

2009-236798, PERG03-GA-2008-231034), the Alexander von

Humboldt Foundation, the Natural Sciences and Engineering

Research Council of Canada, the German Research Council

(DFG) through subproject Rillig of FOR 816, the Estonian Sci-

ence Foundation (grants 7738, SF0180098s08), and the Euro-

pean Regional Development Fund (Centre of Excellence

FIBIR).

References

Ahrens D, Monaghan MT, Vogler AP (2007) DNA-based

taxonomy for associating adults and larvae in multi-species

assemblages of chafers (Coleoptera: Scarabaeidae). Molecular

Phylogenetics and Evolution, 44, 436–449.

Barraclough TG, Hughes M, Ashford-Hodges N, Fujisawa T

(2009) Inferring evolutionarily significant units of bacterial

diversity from broad environmental surveys of single-locus

data. Biology Letters, 5, 425–428.

Biro I, Takacs T (2007) Effects of Glomus mosseae strains of

different origin on plant macro and micronutrient uptake in

Cd-polluted and unpolluted soils. Acta Agronomica

Hungarica, 55, 183–192.

Biro I, Nemetha T, Takacs T (2009) Changes of parameters of

infectivity and efficiency of different Glomus mosseae

arbuscular mycorrhizal fungi strains in cadmium-loaded soils.

Communications in Soil Science and Plant Analysis, 40, 227–239.

Bode S, Adolfsson S, Lamatsch D, et al. (2010) Exceptional

cryptic diversity and multiple origins of parthenogenesis in

a freshwater ostracod. Molecular Phylogenetics and Evolution,

54, 542–552.

Clapp JP, Fitter AH, Young JPW (1999) Ribosomal small

subunit sequence variation within 19 spores of an arbuscular

mycorrhizal fungus, Scutellospora sp. Molecular Ecology, 8,

915–921.

Clapp JP, Rodriguez A, Dodd JC (2001) Inter- and intra-isolate

rRNA large subunit variation in Glomus coronatum spores.

New Phytologist, 149, 539–554.

Croll D, Sanders, IR (2009) Recombination in Glomus

intraradices, a supposed ancient asexual arbuscular

mycorrhizal fungus. BMC Evolutionary Biology, 9, 13.

Croll D, Wille L, Gamper HA, et al. (2008). Genetic diversity

and host plant preferences revealed by simple sequence

repeat and mitochondrial markers in a population of the

arbuscular mycorrhizal fungus Glomus intraradices. New

Phytologist, 178, 672–687.

664 J . R . POWELL ET AL.

Croll D, Giovannetti M, Koch AM, et al. (2009) Nonself

vegetative fusion and genetic exchange in the arbuscular

mycorrhizal fungus Glomus intraradices. New Phytologist, 181,

924–937.

Dray S, Dufour AB (2007) The ade4 package: implementing the

duality diagram for ecologists. Journal of Statistical Software,

22, 1–22.

Drummond A, Rambaut A (2007) BEAST: Bayesian evolutionary

analysis by sampling trees. BMC Evolutionary Biology, 7, 214.

Edgar RC (2004) MUSCLE: multiple sequence alignment with

high accuracy and high through put. Nucleic Acids Research,

32, 1792–1797.

Ehinger M, Koch AM, Sanders IR (2009) Changes in arbuscular

mycorrhizal fungal phenotypes and genotypes in response to

plant species identity and phosphorus concentration. New

Phytologist, 184, 412–423.

Ezard T, Fujisawa T, Barraclough T (2009) splits: SPecies’ LImits

by Threshold Statistics. Rpackage version 1.0-11/r29.

Available from http://R-Forge.R-project.org/projects/splits/.

Fontaneto D, Herniou EA, Boschetti C, et al. (2007)

Independently evolving species in asexual bdelloid rotifers.

PLoS Biology, 5, e87.

Fraser C, Alm EJ, Polz MF, Spratt BG, Hanage WP (2009) The

bacterial species challenge: making sense of genetic and

ecological diversity. Science, 323, 741–746.

Gevers D, Cohan FM, Lawrence JG, et al. (2005) Opinion: re-

evaluating prokaryotic species. Nature Reviews Microbiology,

3, 733–739.

Giovannetti M, Azzolini D, Citernesi AS (1999) Anastomosis

formation and nuclear and proto plasmic exchange in

arbuscular mycorrhizal fungi. Applied and Environmental

Microbiology, 65, 5571–5575.

Hanage WP, Fraser C, Spratt BG (2006) Sequences, sequence

clusters and bacterial species. Philosophical Transactions of the

Royal Society B: Biological Sciences, 361, 1917–1927.

Hart MM, Reader RJ (2002) Taxonomic basis for variation in

the colonization strategy of arbuscular mycorrhizal fungi.

New Phytologist, 153, 335-344.

Hart M, Reader R, Klironomos J (2003) Plant coexistence

mediated by arbuscular mycorrhizal fungi. Trends in Ecology

& Evolution, 18, 418–423.

Haug I, Wubet T, Weiß M, et al. (2010) Species-rich but

distinct arbuscular mycorrhizal communities in reforestation

plots on degraded pastures and in neighboring pristine

tropical mountain rain forest. Tropical Ecology, 51, 125–148.

van der Heijden MGA, Klironomos JN, Ursic M, et al. (1998)

Mycorrhizal fungal diversity determines plant biodiversity,

ecosystem variability and productivity. Nature, 396, 69–72.

Helgason T, Daniell TJ, Husband R, Fitter AH, Young JPW

(1998) Ploughing up the wood-wide web? Nature, 394, 431.

Hibbett DS, Ohman A, Kirk PM (2009) Fungal ecology catches

fire. New Phytologist, 184, 279–282.

Hill MO, Smith AJE (1976) Principal component analysis of

taxonomic data with multi-state discrete characters. Taxon,

25, 249–255.

Jany J, Pawlowska TE (2010) Multinucleate spores contribute to

evolutionary longevity of asexual Glomeromycota. The

American Naturalist, 175, 424–435.

Johnson NC (2010) Resource stoichiometry elucidates the

structure and function of arbuscular mycorrhizas across

scales. New Phytologist, 185, 631–647.

Johnson NC, Tilman D, Wedin D (1992) Plant and soil controls

on mycorrhizal fungal communities. Ecology, 73, 2034–2042.

Johnson D, IJdo M, Genney DR, Anderson IC, Alexander IJ

(2005) How do plants regulate the function, community

structure, and diversity of mycorrhizal fungi? Journal of

Experimental Botany, 56, 1751–1760.

Jousselin E, Desdevises Y, Coeur d’acier A (2009) Fine-scale

cospeciation between Brachycaudus and Buchnera aphidicola:

bacterial genome helps define species and evolutionary

relationships in aphids. Proceedings of the Royal Society B:

Biological Sciences, 276, 187–196.

Kingman JFC (1982) On the genealogy of large populations.

Journal of Applied Probability, 19, 27–43.

Klironomos JN, McCune J, Hart M, Neville J (2000) The

influence of arbuscular mycorrhizae on the relationship

between plant diversity and productivity. Ecology Letters, 3,

137–141.

Koch AM, Kuhn G, Fontanillas P, Fumagalli L, Goudet J,

Sanders IR (2004) High genetic variability and low local

diversity in a population of arbuscular mycorrhizal fungi.

Proceedings of the National Academy of Sciences of the United

States of America, 101, 2369–2374.

Koch AM, Croll D, Sanders, IR (2006) Genetic variability in a

population of arbuscular mycorrhizal fungi causes variation

in plant growth. Ecology Letters, 9, 103–110.

Koeppel A, Perry EB, Sikorski J, et al. (2008) Identifying the

fundamental units of bacterial diversity: a paradigm shift to

incorporate ecology into bacterial systematics. Proceedings of

the National Academy of Sciences of the United States of America,

105, 2504–2509.

Kottke I, Haug I, Setaro S, et al. (2008) Guilds of mycorrhizal

fungi and their relation to trees, ericads, orchids and

liverworts in a neotropical mountain rain forest. Basic and

Applied Ecology, 9, 13–23.

Legendre P, Anderson MJ (1999) Distance-based redundancy

analysis: testing multispecies esponses in multifactorial

ecological experiments. Ecological Monographs, 69, 1–24.

Leliaert F, Verbruggen H, Wysor B, De Clerck O (2009) DNA

taxonomy in morphologically plastic taxa: algorithmic

species delimitation in the Boodlea complex (Chlorophyta:

Cladophorales). Molecular Phylogenetics and Evolution, 53,

122–133.

Liesack W, Stackebrandt E (1992) Unculturable microbes

detected by molecular sequences and probes. Biodiversity and

Conservation, 1, 250–262.

Maherali H, Klironomos JN (2007) Influence of phylogeny on

fungal community assembly and ecosystem functioning.

Science, 316, 1746–1748.

McQuarrie ADR, Tsai C (1998) Regression and Time Series Model

Selection. World Scientific, Singapore.

Monaghan MT, Wild R, Elliot M, et al. (2009) Accelerated

species inventory on madagascar using coalescent-based

models of species delineation. Systematic Biology, 58, 298–311.

Morton JB, Bentivenga SP (1994) Levels of diverstity in

endomycorrhizal fungi (Glomales, Zygomycetes) and their

role in defining taxonomic and non-taxonomic groups. Plant

and Soil, 159, 47–59.

Mummey DL, Rillig MC, Holben WE (2005) Neighboring plant

influences on arbuscular my corrhizal fungal community

composition as assessed by T-RFLP analysis. Plant and Soil,

271, 83–90.

� 2010 Blackwell Publishing Ltd

SPECIES DELINEATI ON IN FUNGAL INVENTORIES 665

Munkvold L, Kjøller R, Vestberg M, Rosendahl S, Jakobsen I

(2004) High functional diversity within species of arbuscular

mycorrhizal fungi. New Phytologist, 164, 357–364.

Nilsson RH, Kristiansson E, Ryberg M, Hallenberg N, Larsson

KH (2008) Intraspecific ITS variability in the Kingdom Fungi

as expressed in the international sequence databases and its

implications for molecular species identification. Evolutionary

Bioinformatics, 4, 193–201.

Oksanen J, Blanchet FG, Kindt R, et al. (2010) vegan:

Community Ecology Package. R package version 1.17-0.

Available from http://CRAN.R-project.org/package¼vegan.

Opik M, Moora M, Liira J, Koljalg U, Zobel M, Sen R (2003)

Divergent arbuscular mycorrhizal fungal communities

colonize roots of Pulsatilla spp. in boreal scots pine forest

and grassland soils. New Phytologist, 160, 581–593.

Opik M, Moora M, Liira J, Zobel M (2006) Composition of

root-colonizing arbuscular mycorrhizal fungal communities

in different ecosystems around the globe. Journal of Ecology,

94, 778–790.

Opik M, Moora M, Zobel M, et al. (2008) High diversity of

arbuscular mycorrhizal fungi in a boreal herb-rich coniferous

forest. New Phytologist, 179, 867–876.

Opik M, Metsis M, Daniell TJ, Zobel M, Moora M (2009)

Large-scale parallel 454 sequencing reveals host ecological

group specificity of arbuscular mycorrhizal fungi in a

boreonemoral forest. New Phytologist, 184, 424–437.

Opik M, Vanatoa A, Vanatoa E, et al. (2010) The online

database MaarjAM reveals global and ecosystemic

distribution patterns in arbuscular mycorrhizal fungi

(Glomeromycota). New Phytologist, 188, 223–241.

Papadopoulou A, Anastasiou I, Keskin B, Vogler AP (2009)

Comparative phylogeography of tenebrionid beetles in the

Aegean archipelago: the effect of dispersal ability and

habitat preference. Molecular Ecology, 18, 2503–2517.

Paradis E, Claude J, Strimmer K (2004) APE: analyses of

phylogenetics and evolution in R language. Bioinformatics, 20,

289–290.

Pawlowska TE, Taylor JW (2004) Organization of genetic

variation in individuals of arbuscular mycorrhizal fungi.

Nature, 427, 733–737.

Pons J, Barraclough TG, Gomez-Zurita J, et al. (2006) Sequence-

based species delimitation for the DNA taxonomy of

undescribed insects. Systematic Biology, 55, 595–609.

Posada D, Crandall K (1998) MODELTEST: testing the model

of DNA substitution. Bioinformatics, 14, 817–818.

Powell JR, Parrent JL, Hart MM, Klironomos JN, Rillig MC,

Maherali H (2009) Phylogenetic trait conservatism and the

evolution of functional trade-offs in arbuscular mycorrhizal

fungi. Proceedings of the Royal Society B: Biological Sciences,

276, 4237–4245.

R Development Core Team (2009) R: A Language and

Environment for Statistical Computing. R Foundation for

Statistical Computing, Vienna, Austria. Available from

http://www.R-project.org.

Rodriguez A, Clap, JP, Dodd JC (2004) Ribosomal RNA gene

sequence diversity in arbuscular mycorrhizal fungi

(Glomeromycota). Journal of Ecology, 92, 986–989.

Rosendahl S, Taylor JW (1997) Development of multiple

genetic markers for studies of genetic variation in

arbuscular mycorrhizal fungi using AFLP. Molecular

Ecology, 6, 821–829.

� 2010 Blackwell Publishing Ltd

Rosendahl S, McGee P, Morton JB (2009). Lack of global

population genetic differentiation in the arbuscular

mycorrhizal fungus Glomus mosseae suggests a recent range

expansion which may have coincided with the spread of

agriculture. Molecular Ecology, 18, 4316–4329.

Sykorova Z, Ineichen K, Wiemken A, Redecker D (2007) The

cultivation bias: different communities of arbuscular

mycorrhizal fungi detected in roots from the field, from bait

plants transplanted to the field, and from a greenhouse trap

experiment. Mycorrhiza, 18, 1–14.

Schenck NC, Smith GS (1982) Additional new and unreported

species of mycorrhizal fungi (Endogonaceae) from Florida.

Mycologia, 74, 77–92.

Scheublin TR, Ridgway KP, Young JPW, van der Heijden

MGA (2004) Nonlegumes, legumes, and root nodules harbor

different arbuscular mycorrhizal fungal communities. Applied

and Environmental Microbiology, 70, 6240–6246.

Schloss PD, Westcott SL, Ryabin T, et al. (2009) Introducing

mothur: open-source, platform independent, community-

supported software for describing and comparing microbial

communities. Applied and Environmental Microbiology, 75,

7537–7541.

Seifert KA, Samson RA, deWaard JR, et al. (2007) Prospects for

fungus identification using CO1 DNA barcodes, with

Penicillium as a test case. Proceedings of the National Academy

of Sciences of the United States of America, 104, 3901–3906.

Sikes BA, Cottenie K, Klironomos JN (2009) Plant and fungal

identity determines pathogen protection of plant roots by

arbuscular mycorrhizas. Journal of Ecology, 97, 1274–1280.

Smith SE, Read DJ (2008) Mycorrhizal Symbiosis, 3rd edn.

Academic Press, London.

Stackebrandt E, Goebel BM (1994) Taxonomic note: a place for

DNA–DNA reassociation and 16S rRNA sequence analysis

in the present species definition in bacteriology. International

Journal of Systematic and Evolutionary Bacteriology, 44, 846–

849.

Vogler AP, Monaghan MT (2007) Recent advances in DNA

taxonomy. Journal of Zoological Systematics and Evolutionary

Research, 45, 1–10.

Wehner J, Antunes PM, Powell JR, Mazukatow J, Rillig MC

(2010) Plant pathogen protection by arbuscular mycorrhizas:

a role for fungal diversity? Pedobiologia, 53, 197–201.

Yule GU (1925) A mathematical theory of evolution, based on

the conclusions of Dr. J. C. Willis, F.R.S. Philosophical

Transactions of the Royal Society of London. Series B, Containing

Papers of a Biological Character, 213, 21–87.

J.R.P. is interested in the development and use of tools for

describing the ecological structure of microbial communities

and in identifying the factors driving this structure. These

analyses were performed while he was a postdoctoral research

associate in the lab of M.C.R., who studies various aspects of

plant-fungal symbioses that are relevant for the functioning of

ecosystems. M.T.M. studies biodiversity using evolutionary

and phylogenetic methods, primarily in aquatic invertebrates

and microorganisms. M.O. studies the structure and dynamics

of arbuscular mycorrhizal fungi from local to global spatial

scales and curates a database of virtual AM fungal taxa (Maar-

jAM) for these purposes.

666 J . R . POWELL ET AL.

Supporting information

Additional supporting information may be found in the online

version of this article:

Fig. S1 Species accumulation curves indicating the relationship

between AM fungal species richness and sampling effort in the

two surveys.

Fig. S2 Relationship between AM fungal genetic clusters pre-

dicted by the GMYC model and OTUs predicted based on a

sequence similarity cutoff (97%) for AM fungal sequences

obtained from the Estonia survey.

Fig. S3 AM fungal genetic clusters from the Ecuador survey

predicted by the GMYC model.

Please note: Wiley-Blackwell are not responsible for the content

or functionality of any supporting information supplied by the

authors. Any queries (other than missing material) should be

directed to the corresponding author for the article.

� 2010 Blackwell Publishing Ltd