Phylogenetic reference data for systematics and phylotaxonomy of arbuscular mycorrhizal fungi from phylum to species level


Author for correspondence:
Arthur Schüßler
Tel: +49 89 2180 74730


  • Although the molecular phylogeny, evolution and biodiversity of arbuscular mycorrhizal fungi (AMF) are becoming clearer, phylotaxonomically reliable sequence data are still limited. To fill this gap, a data set allowing resolution and environmental tracing across all taxonomic levels is provided.
  • Two overlapping nuclear DNA regions, totalling c. 3 kb, were analysed: the small subunit (SSU) rRNA gene (up to 1800 bp) and a fragment spanning c. 250 bp of the SSU rDNA, the internal transcribed spacer (ITS) region (c. 475–520 bp) and c. 800 bp of the large subunit (LSU) rRNA gene. Both DNA regions together could be analysed for 35 described species, the SSU rDNA for c. 76 named and 18 as yet undefined species, and the ITS region or LSU rDNA, or a combination of both, for c. 91 named and 16 as yet undefined species.
  • Present phylogenetic analyses, based on the three rDNA markers, provide reliable and robust resolution from phylum to species level. Altogether, 109 named species and 27 cultures representing as yet undefined species were analysed.
  • This study provides a reference data set for molecular systematics and environmental community analyses of AMF, including analyses based on deep sequencing.


The arbuscular mycorrhizal fungi (AMF; Glomeromycota; Schüßler et al., 2001) form symbioses with most land plants, in almost any terrestrial ecosystem (Smith & Read, 2008). Despite their considerable ecological importance, the biology and ecology of these fungi are still not well understood. This is partly because of their obligate symbiotic, asexual and hidden lifestyle in soil and roots.

Regarding AMF identification, the ‘classical’ characterization based on spore structures may be flawed because of limited morphological differentiation. Some species form more than one spore morph, and several cryptic taxa were only uncovered as a result of molecular biological analyses. In many studies it is important to know the AMF identities, but even ‘model AMF’ were shown to be misclassified (Stockinger et al., 2009; Schüßler et al., 2011) and several other problems were recently addressed in taxonomic revisions (Kaonongbua et al., 2010; Morton & Msiska, 2010a; Schüßler & Walker, 2010; Oehl et al., 2011a).

Accurate identification is crucial, for example, in AMF community studies, which increasingly rely on phylotaxonomy solely based on molecular genetic data. Most commonly used is the nuclear small subunit (SSU) rRNA gene, hereafter referred to as SSU. Several SSU-targeting PCR primers (e.g. Simon et al., 1992; Helgason et al., 1998; Lee et al., 2008) that amplify fragments of c. 500–800 bp have been widely applied in ecological studies (Öpik et al., 2008; Zhang et al., 2010). However, in SSU data sets, one phylotype (often defined by a 97% sequence similarity level) may represent different species and, conversely, different phylotypes may indeed belong to just one species. This makes the resolution of closely related species impossible (Walker et al., 2007; Gamper et al., 2009) and we therefore eschew terms such as ‘virtual taxa’ (Öpik et al., 2010) and ‘species’ for phylotypes that are uncertain to represent taxa, and thus in fact are taxonomically undefined. A ‘taxon’ in mycology is clearly defined (see Article 1.1 in McNeill et al., 2006) and a more appropriate term is the ‘molecular operational taxonomic unit’ (MOTU). Standardized MOTUs are a goal for the classification of unknown fungal species from environmental samples (Hibbett et al., 2011), but care has to be taken that the units indeed are based on coherent taxonomic levels (Hawksworth et al., 2011). Standardization could also facilitate traditional biodiversity analyses (Magurran, 2004).

The more variable region covering nuclear internal transcribed spacer 1 (ITS1), the 5.8S rRNA gene and ITS2 rDNA (hereafter referred to as the ITS region) has also been used for detecting AMF (Redecker et al., 2000; Renker et al., 2003; Hempel et al., 2007), but is often inadequate for discriminating very closely related species (Stockinger et al., 2010). As a marker with intermediate sequence variability, the nuclear large subunit (LSU) rRNA gene (hereafter referred to as LSU) has proved useful for AMF detection (Gollotte et al., 2004; Pivato et al., 2007), although several PCR primers used do not amplify particular AMF lineages (Krüger et al., 2009). Other markers such as the genes for mitochondrial LSU rRNA (Börstler et al., 2010; Sýkorováet al., 2011), β-tubulin (Msiska & Morton, 2009), RNA polymerase II subunits 1 and 2 (James et al., 2006; Redecker & Raab, 2006), or H+-ATPase (Corradi et al., 2004; Sokolski et al., 2010) have been used, but either do not allow phylogenetic species identification or are as yet only known for very few species.

The nuclear rDNA region sequence data set that has been assembled over the past decade is becoming taxonomically sufficiently broad to permit molecular ecological field studies of AMF communities. However, comparison among studies is often difficult because of inconsistencies, for example regarding the coverage of the different loci. The ITS region is often used to determine fungal species (e.g. Tedersoo et al., 2008) and will be proposed as the official fungal DNA barcode (C. Schoch et al., unpublished). Unfortunately, for AMF most environmental ITS region phylotypes cannot be affiliated to species, or species-level identities are not determinable using only this short and highly variable region (Stockinger et al., 2009). Thus, neither the conserved SSU nor the highly variable ITS region alone reliably resolves closely related AMF, but this is possible using a c. 1.5-kb rDNA fragment (Stockinger et al., 2010), easily amplifiable with AMF-specific primers (Krüger et al., 2009). This SSU-ITS-LSU fragment covers c. 250 bp of the SSU, the complete ITS region and c. 800 bp of the LSU. Shorter fragments, such as the c. 400- or 800-bp reads provided by 454-sequencing, will provide information about species identities when analysed with reference to a ‘phylogenetic backbone’ based on longer sequences, such as the SSU-ITS-LSU fragment (Stockinger et al., 2010).

In this further effort to establish a reference database, we (re-)analysed nuclear rDNA regions that can be specifically and easily amplified by PCR for AMF (Krüger et al., 2009), resolve closely related species to allow DNA barcoding (Stockinger et al., 2009, 2010), and facilitate the application of deep sequencing technologies for in-field detection of AMF.

Materials and Methods

AMF material, DNA extraction, PCR, cloning and sequencing

The identities of the AMF subjected to molecular analyses were determined from morphological characters. For most of them, vouchers were deposited in the C. Walker collection and are available from the Royal Botanic Garden Edinburgh (Supporting Information Table S1).

Cleaned AMF spores were used for DNA extraction or stored as described in Schwarzott & Schüßler (2001). For some extractions, a simplified ‘PCR-buffer protocol’ was followed (Naumann et al., 2010). DNA was extracted from individual spores, except for some isolates (derived from one single spore) for which up to 10 spores were pooled. PCR amplification of the near full-length SSU was as described in Schwarzott & Schüßler (2001). Some SSU fragments, from earlier studies, were amplified with the primers AML1-AML2, NS1-NS2, NS1-Geo10 and GeoA1-ITS1Frc (ITS1F reverse complementary, 5′-TTACTTCCTCTAAATGACCAAG-3′).

For amplification of a c. 1.8-kb SSU-ITS-LSU fragment, the primers SSUmAf-LSUmAr (in some cases with LR4+2 as reverse primer; Stockinger et al., 2009) were used, mostly followed by a nested PCR with the primer SSUmCf-LSUmBr or, in some earlier attempts, SSU-Glom1-NDL22 (Krüger et al., 2009; Stockinger et al., 2010), resulting in a c. 1.5-kb amplicon covering c. 250 bp of the SSU, the whole ITS region and c. 800 bp of the LSU. PCR products were cloned and analysed as described in Krüger et al. (2009).

New sequences were deposited in the EMBL database under the accession numbers AM114274, AM713432, FR750012FR750095, FR750101FR750117, FR750126FR750127, FR750134FR750217, FR750220FR750228, FR750363FR750376, FR750526FR750544, FR772325, FR773142FR773152, FR774917, and HE610426-HE610427.

Sequence data and Glomeromycota taxonomy used

Sequences in the public databases were reviewed to establish if they were from defined cultures or environmental samples. Environmental sequences not identified to species were excluded. Defined sequences of > 650 bp and some shorter sequences were included or assembled to ‘contiguous’ sequences if they were the only ones available for a particular taxon or culture. For several database sequences it is unclear if they refer to an AMF single spore isolate, multi-spore culture, or simply a recombinant DNA Escherichia coli clone number. Our annotations follow the most recent systematics of the Glomeromycota (Schüßler & Walker, 2010; Oehl et al., 2011a), including the suggestions of Morton & Msiska (2010a) and Kaonongbua et al. (2010). Table S1 gives detailed information about sequence origin.

Phylogenetic analyses

For the SSU sequences, one strict (with variable sites coded according to IUPAC (International Union of Pure and Applied Chemistry) as degenerated bases) consensus sequence was deduced from the available (up to 10) sequence variants for each isolate or culture. The PCR primer binding sites were excluded, when known. Three different data sets were then analysed.

  • Firstly, for the phylogenetic tree computed from c. 2.7-kb sequences (Fig. 1) we concatenated the above-mentioned strict SSU consensus sequence with one strict consensus sequence made from all SSU-ITS-LSU sequence variants of the same fungus (defined by culture identifier), whereas the unalignable ITS1 and ITS2 were excluded. Such SSUfull-5.8S-LSU sequences could be assembled for 35 species from 38 cultures. As there were no corresponding SSU and SSU-ITS-LSU sequences available for a culture of Archaeospora schenckii, sequences from two different cultures (Att58-6 and Att212-4; sequences identical in the 250-bp SSU overlap) had to be concatenated to cover the genus Archaeospora. Batrachochytrium dendrobatidis (Chytridiomycota) was used as an outgroup and the following members of basal fungal lineages and Dikarya were also included: Ascomycota (Exophiala dermatitidis and Schizosaccharomyces pombe), Basidiomycota (Henningsomyces candidus and Rhodotorula hordea), Kickxellomycotina (Orphella haysii and Smittium culisetae), Mucoromycotina (Endogone pisiformis, Mortierella verticillata, Phycomyces blakesleeanus and Rhizopus oryzae) and Blastocladiales (Allomyces arbusculus and Coelomomyces stegomyiae).

  • Secondly, near full-length SSU strict consensus sequences (≤ 1.8 kb) were used to compute a SSU tree (Fig. 2) for 76 AMF species from 145 cultures (including shorter fragments of 500–1300 bp for 18 species from 26 cultures).

  • Finally, all individual SSU-ITS-LSU sequences (up to 24 variants; c. 1.5 kb) available from a culture were analysed. To ‘anchor’ phylogenetically the variable ITS and LSU sequences by the more conserved SSU, each variant was concatenated at the 5′ end with one SSU strict consensus sequence of the same culture, if available, resulting in c. 3-kb sequences. This anchoring allows a more robust resolution of deeper (above genus) topologies and avoids artificial clustering resulting from misalignment or from convergent characters resulting from mutational saturation in the highly variable regions. Subtrees at order and family level could be computed for 91 defined and 16 unnamed species (Figs 3, 4 and 5), representing all main lineages in the Glomeromycota. For the model fungus Rhizophagus irregularis DAOM197198, a reduced sequence set, still representing the breadth of rDNA variability, was used, because a detailed analysis had been already published in Stockinger et al. (2009). For Gigasporaceae, Paraglomerales and Archaeosporales, the composite data set also included short database sequences (≥ 500 bp) if their inclusion did not significantly reduce the topological support (Fig. 3). For the genera in the Glomerales (except Rhizophagus), separate analyses were conducted for long sequences (Fig. 5), and after inclusion of short sequences (Figs S1, S2).

Figure 1.

Maximum likelihood phylogenetic tree based on concatenated nuclear small subunit full (SSUfull)–5.8S–large subunit (LSU) rDNA strict consensus sequences (c. 2700 bp) of the Glomeromycota and other fungal lineages used as outgroups. Branches with < 60% bootstrap support were collapsed to polytomies, and a long branch was shortened by 50%, which is indicated with two diagonal slashes. Terminal nodes marked with ‘(consensus #)’ represent strict consensus sequences of sequences with the accession numbers listed in Supporting Information Notes S1. The scale bar indicates the number of substitutions per site. The following culture identifiers are not shown in the tree because of space limitations: Acaulospora brasiliensis (consensus 5) is derived from W4699/Att1211-0 and W5473/Att1210-5, Diversispora spurca (consensus 7) from W2396/Att246-4 and W4119/Att246-18, Diversispora aurantia (consensus 8) from W4728/Att1296-0, Glomus macrocarpum (consensus 22) from a field-collected sporocarp (W5288) and Att1495-0 (two independent samplings W5581 and W5605), Ambispora fennica (consensus 36) from W4752/Att200-23 and W3569/Att200-11, and Archaeospora schenckii from W3571/Att58-6 and W5673/Att212-4.

Figure 2.

Maximum likelihood phylogenetic tree based on concatenated nuclear small subunit (SSU) rDNA strict consensus sequences (c. 1.8 kb). Paraglomus was used as an outgroup as it represents the most basal glomeromycotan branch (see Fig. 1). Branches receiving < 60% bootstrap support were collapsed to polytomies. Terminal nodes marked with ‘(consensus #)’ represent strict consensus sequences of sequences with the accession numbers listed in Supporting Information Notes S2. The scale bar indicates the number of substitutions per site. Sequences ≤ 1300 bp in length are indicated with *. The generic type species, when included in the analysis, is shown in bold and underlined.

Figure 3.

Maximum likelihood phylogenetic trees based on individual small subunit–internal transcribed spacer–large subunit (SSU-ITS-LSU) rDNA sequence variants assembled with, when available, the corresponding SSU strict consensus sequence. Branches receiving < 60% bootstrap support were collapsed to polytomies, and long branches were shortened by 50%, which is indicated with two diagonal slashes, or by 75%, indicated with three slashes. Bootstrap values are given for branches among but not within different cultures. The scale bar indicates the number of substitutions per site. Sequences from one submission in 2010 (see text in the Results section) are marked with ◂, potential contaminant or wrongly annotated sequences are indicated with •, and the respective sequence length of all sequences < 1 kb in length is shown within the taxon labels. (a) Paraglomerales and Archaeosporales; Ascomycota and Basidiomycota were used as the outgroup. Terminal nodes marked with ‘(consensus #)’ represent strict consensus sequences of sequences with the accession numbers listed in Supporting Information Notes S3. (b) Gigasporaceae, including public database sequences of > 700 bp; Acaulospora species were used as the outgroup. Consensus 10 is a strict consensus sequence of the sequences AY635832, AY997088 and DQ273792 and consensus 11 is a strict consensus sequence of sequences AJ87127073.

Figure 4.

Maximum likelihood phylogenetic tree based on small subunit–internal transcribed spacer–large subunit (SSU-ITS-LSU) rDNA sequence variants assembled with, when available, the corresponding SSU strict consensus sequence. Branches receiving < 60% bootstrap support were collapsed to polytomies, and long branches were shortened by 50%, which is indicated with two diagonal slashes, or by 75%, indicated with three slashes. Bootstrap values are given for branches among but not within different cultures. Scale bar, number of substitutions per site. Sequences from one submission in 2010 (see text in the Results section) are marked with ◂, potential contaminant or wrongly annotated sequences are indicated with •, and the respective sequence length of all sequences < 1 kb in length is noted. (a) Acaulosporaceae, with Diversispora as the outgroup. Consensus 1 is a strict consensus sequence of sequences AJ250847, AJ242499 and FJ461802. (b) Diversisporaceae; Acaulospora species were used as the outgroup. Consensus 2 is a strict consensus sequence of sequences DQ35044853 and consensus 3 is a strict consensus sequence of sequences AM41854344.

Figure 5.

Maximum likelihood phylogenetic tree based on small subunit–internal transcribed spacer–large subunit (SSU-ITS-LSU) rDNA sequence variants of the Glomerales assembled with, when available, the corresponding SSU strict consensus sequence. Branches receiving < 60% bootstrap support were collapsed to polytomies, and long branches were shortened by 50%, which is indicated with two diagonal slashes, or by 75%, indicated with three slashes. Bootstrap values are given for branches among but not within cultures. The scale bar indicates the number of substitutions per site. Sequences from one submission in 2010 (see text in the Results section) are marked with ◂, potential contaminant or wrongly annotated sequences are indicated with •, and the respective sequence length of all sequences < 1 kb in length is shown within the taxon labels. (a) Funneliformis and Glomus; consensus 1 is a strict consensus sequence of sequences AY635833, AY997053 and DQ273793. (b) Rhizophagus and Sclerocystis; consensus 2 is a strict consensus sequence of sequences DQ322630, AY997054 and DQ273828 and consensus 3 is a strict consensus sequence of sequences AY635831, AY997052 and DQ273790. (c) Claroideoglomus; consensus 4 is a strict consensus sequence of Y17639, Z14008 and AJ239125.

All maximum likelihood phylogenetic analyses were computed through the CIPRES web-portal with RAxML ver. 7.2.8 (Stamatakis et al., 2008) using 1000 bootstraps and the GTRGAMMA model for both bootstrapping and tree inference. The alignments are freely available at


For phylogenetic analyses, a c. 1.8-kb SSU fragment and a c. 1.5-kb SSU-ITS-LSU fragment, both overlapping by c. 250 bp at the 3′ end of the SSU, and other sequences from our laboratory (Table S1) were analysed together with public database sequences. Altogether, sequences derived from 109 AMF annotated as described species and from 27 undefined putative species could be analysed.

SSUfull-5.8S-LSU phylogeny of the Glomeromycota (Fig. 1)

The phylogenetic tree was computed from 39 assembled 2.7-kb consensus sequences representing 35 species. The highly variable ITS1 and ITS2 regions were excluded because alignment is impossible among higher taxa. However, their inclusion did not alter tree topology (data not shown), demonstrating robust phylogenetic anchoring by the more conserved regions (receiving more weight in RAxML analyses). The topology of the SSUfull-5.8S-LSU tree is congruent with that of previously published rDNA trees, but supported by higher bootstrap support (BS) values. The Glomeromycota are supported as monophyletic, with the Paraglomerales as the most ancestral lineage (separated with 85% BS from all other AMF lineages). The next basal lineage, the Archaeosporales (including Geosiphonaceae, Archaeosporaceae and Ambisporaceae), resolves as monophyletic (88% BS) and the proximate lineage comprises the sister clades Diversisporales and Glomerales, which cluster together with 100% BS. The Diversisporales appears monophyletic (94% BS), with all its families well supported (except Entrophosporaceae, which had to be excluded for lack of reliable sequence data).

Members of the Glomerales (63% BS) separate into the Glomeraceae (former Glomus Group (GlGr) A) and Claroideoglomeraceae (former GlGrB). The Glomeraceae contains the four genera Funneliformis (former GlGrAa), Rhizophagus and Sclerocystis (former GlGrAb), and Glomus (former GlGrAc). Glomus is represented by the generic type species Glomus macrocarpum (epitype W5581/Att1495-0) and Funneliformis by Funneliformis mosseae, Funneliformis coronatus, Funneliformis caledonius and Funneliformis sp. WUM3. In Rhizophagus the ‘model fungus’R. irregularis DAOM197198 clusters with two other cultures of this species, GINCO4695rac11G2 (=AFTOL-ID845) and a root organ culture (ROC) annotated as DAOM212349. However, the last number is the voucher number also used for the type material of Claroideoglomus lamellosum (from a field collection) and, additionally, for an ‘isotype’ pot culture of that species. The sequences of Rhizophagus intraradices, from ex-type culture FL208, cluster as sister to Rhizophagus proliferus (DAOM226389).

SSU phylogeny of the Glomeromycota (Fig. 2)

The available sequences of 76 species (145 cultures) were analysed. For the basal lineages Archaeosporales and Paraglomerales relatively few are characterized. Sequences of the former Intraspora schenckii cluster among those of Archaeospora.

In the Diversisporales, the SSU tree shows 100% BS for the Gigasporaceae. Gigaspora appears monophyletic, but Racocetra and Scutellospora are not convincingly resolved. Scutellospora gilmorei, Scutellospora nodosa and Scutellospora pellucida cluster on a branch together with Racocetra species. Scutellospora cerradensis, Scutellospora reticulata, Scutellospora heterogama and the recently described Dentiscutata colliculosa form a monophyletic clade (80% BS). These clades cluster together with low support (62% BS) and the remaining Scutellospora species fall in a clade with the type species, Scutellospora calospora (66% BS). The family Acaulosporaceae is well supported (100% BS), but not the deeper branching order within the family. For Otospora bareae (Palenzuela et al., 2008) the concatenation of two short non-overlapping partial SSU sequences (AM400229, AM905318) clusters among Diversispora sequences, as does the only sequence (FN397100) published for Entrophospora nevadensis (Palenzuela et al., 2010). Redeckera, a genus based on data from Redecker et al. (2007), clearly separates from Diversispora. The Pacisporaceae are sister to Gigasporaceae with 79% BS.

The Glomeraceae and Claroideoglomeraceae are both supported by 100% BS. Glomus iranicum and Glomus indicum (Błaszkowski et al., 2010a,b) fall basally into a polytomy in the Glomeraceae. Funneliformis is composed of F. mosseae (nine cultures), F. coronatus (BEG28 and COG1), Funneliformis geosporus (BEG11), Funneliformis sp. DAOM225952, F. caledonius (BEG15 and BEG20), Funneliformis sp. WUM3, Funneliformis fragilistratus and Funneliformis verruculosus. Septoglomus constrictum, together with Septoglomus africanum, clusters basally. Glomus, comprising sequences of G. macrocarpum (W5293 and W5605/Att1495-0) and Glomus sp. W3347/Att565-7, clusters with low BS (61%) as sister to Funneliformis. Rhizophagus comprises R. irregularis (DAOM197198, AFTOL-ID845, W4533/Att1225-1, and the above-mentioned DAOM212349), Rhizophagus sp. W3563, Rhizophagus vesiculiferus (W2857/Att14-8; the name is used informally here, formally the species was erroneously placed in Funneliformis in Schüßler & Walker, 2010, which will be corrected soon), Rhizophagus fasciculatus (BEG53), R. intraradices (FL208), Rhizophagus clarus (W3776/Att894-7) and Rhizophagus manihotis (FL879, BR147B and W3224/Att575-9). The genus Sclerocystis is represented by two sequences, one each from Sclerocystis sinuosa (MD126) and Sclerocystis coremioides (BIORIZE), forming a lineage basal to Rhizophagus. Claroideoglomus separates into two clades, one comprising Claroideoglomus sp. W3349/Att565-11 and Viscospora viscosa BEG27 (possibly incorrectly annotated; see the Discussion section) sequences, and the other C. lamellosum (W3161/Att672-13, W3158/Att244-7 (an ex-‘isotype’ culture, corresponding to DAOM212349), W3814/Att756-1 and W3816/Att844-2), Claroideoglomus etunicatum (UT316, W3815/Att843-1 and W3808/Att367-3), Claroideoglomus luteum SA101, Claroideoglomus claroideum (BEG14, BEG23 and BEG31), and Claroideoglomus spp. (BR212, W3234/Att13-7 and DAOM215235).

SSU-ITS-LSU phylogeny of the basal AMF lineages –Paraglomerales and Archaeosporales (Fig. 3a)

Sequence data are available for all three described Paraglomus species. Paraglomus occultum sequences from four cultures cluster together with 95% BS, including two of three sequences from culture CL383. The third short CL383 sequence and one from P. occultum FL703 group with Paraglomus laccatum, but with low support. One sequence of W5141 (FR750083) and one annotated as A. schenckii (FJ461809◂) tightly group with P. laccatum. The latter must be misannotated. All sequences from the submission containing the latter sequence are herein marked with ‘◂’ (see also Figs 3b, 4, S1 and S2) for easy identification, because there were several inconsistencies found. Sequence FJ461884◂ of the INVAM (International culture collection of (vesicular) arbuscular mycorrhizal fungi) culture NI116B clusters basally to these subclades, and U81987◂ ascribed to P. occultum GR582 falls in the Paraglomusbrasilianum clade, implying a possible misannotation.

The Archaeosporales are represented by sequences from 15 Ambispora spp., five Archaeospora spp., and Geosiphon pyriformis. Archaeospora trappei was analysed using concatenated sequences for cultures AU219 (= WUM19) and NB112, respectively. Archaeospora schenckii sequences cluster with those assigned to A. trappei. For Aschenckii CL401 the two short sequences available could not be concatenated, because sequence AM743189 (3′-SSUpartial-ITS) clusters close to A. trappei NB112, but a partial LSU sequence (FJ461809◂) clusters in Paraglomus. It was meanwhile discovered that the CL401 culture also contains P. occultum (J. Morton, pers. comm.); therefore, FJ461809◂ must be considered to be derived from a contaminant. Ambispora leptoticha (85% BS), Ambispora callosa (79% BS), Ambispora fennica (100% BS), and Ambispora granatensis (Palenzuela et al., 2011; 100% BS) are well resolved, except if short NC169-3 sequences, which cluster unresolved, are included in the analysis. NC169-3 was recently named Ambispora appendicula (Kaonongbua et al., 2010) based on conspecificity with the former Acaulospora appendicula (Morton et al., 1997). The concatenated sequence of Ambispora gerdemannii AU215 clusters with Acallosa (BS 85%). Two other sequences annotated as Agerdemannii, from cultures MT106 (FJ461885◂) and n8_9 (JF439210), cluster with Afennica (BS 100%).

SSU-ITS-LSU phylogeny of the DiversisporalesGigasporaceae (Fig. 3b)

After two recent revisions (Oehl et al., 2008; Morton & Msiska, 2010a), the family Gigasporaceae currently contains Gigaspora, Scutellospora and Racocetra. Gigaspora and Racocetra are supported without conflict (99% and 96% BS, respectively). Of the nine described Gigaspora species, five could be analysed and separated into two subclades. One comprises Gigaspora rosea (DAOM194757 and BEG9) along with sequences of putatively conspecific field-collected yellowish Gigaspora spores (W2992), and one shorter sequence each of Gigaspora albida BR235◂, listed as ‘Gigaspora rosea?’ in INVAM, and Gigaspora gigantea MA401◂. The other clade comprises Gigaspora margarita BEG34 sequences from two independent cultures and shorter sequences, one from Gigaspora decipiens AU102◂, three from ‘G. gigantea isolates’ and two from G. margarita (Gigmar58 and Gigmar60).

In Scutellospora, comprising 24 described species, including ‘D. colliculosa’ and ‘Cetraspora helvetica’, sequences of 12 species are available. Scutellospora divides into three clades. One (Scutellospora sensu Oehl et al., 2008) clusters basally within the Gigasporaceae and is represented by Scutellospora spinosissima W3009/Att664-1, four Scalospora (generic type) cultures, and Scutellospora dipurpurescens WV930◂. A second clade (corresponding to Cetraspora sensu Oehl et al., 2008) clusters with high support (100% BS) as sister to Racocetra and comprises S. gilmorei (87% BS) and a clade (76% BS) with sequences of S. nodosa BEG4, S. pellucida scut1, scut2, scut3, scut4 and NC155C◂, and C. helvetica, whereas S. pellucida scut2 and scut3 cluster between C. helvetica sequences. Some short S. pellucida sequences (AY639261, AY639309, AY639313 and AY639323) are not shown in Fig. 3(b), because their inclusion decreased the BS significantly. The third clade of Scutellospora (85% BS), with sequences corresponding to Dentiscutata and Quatunica sensu Oehl et al. (2008), is sister to Gigaspora, but with low BS. It comprises sequences from several S. heterogama cultures (BR155, NY320, WV858B, SN722, FL225, CL157, BEG35 and FL654 = W5611/Att1577-4, originally determined by Schenck as Scutellospora dipapillosa), S. cerradensis MAFF520056, S. reticulata CNPAB11, and some short sequences of S. reticulata (annotated as Scutellospora nigra, but determined by C. Walker as S. reticulata from stored specimens kindly provided by J. Jansa, December 2010) and Scutellospora erythropus. Short sequences of two S.  erythropus cultures (Sen and MA453B) cluster together with reasonable support, but a third one (HA150◂) is unresolved. The well-supported genus Racocetra (96% BS) comprises sequences from six species. Racocetra fulgida (W2993) is well supported (not shown), but becomes unresolved when shorter sequences of Racocetra verrucosa, Racocetra gregaria, Racocetra persica and Racocetra coralloidea are included. Racocetra weresubiae was transferred back to Scutellospora by Morton & Msiska (2010a), but returned to Racocetra (Schüßler & Walker, 2010) because of its phylogenetic position.

SSU-ITS-LSU phylogeny of the Diversisporales Acaulosporaceae (Fig. 4a)

Presently there are sequences from 39 described Acaulospora species, 22 of which could be analysed. The phylogenetic tree clearly supports the transfer of the former Kuklospora kentinensis and Kuklospora colombiana to Acaulospora (Kaonongbua et al., 2010).

Most analysed Acaulospora species appear well resolved. Acaulospora paulinae CW4 forms a clade comprising eight sequences. Its sister clade contains three sequences from culture WUM18 and short sequences of the recently described Acaulospora sieverdingii. WUM18 is registered as A. paulinae AU103 at INVAM, but according to Oehl et al. (2011b) WUM18 corresponds to Acaulospora sieverdingii. Acaulospora cavernata BEG33 and Acaulospora denticulata cluster monophyletically with A. paulinae and A. sieverdingii (note: BEG33 was determined as Acaulospora scrobiculata when it was registered at the BEG (International Bank for the Glomeromycota) in 1986, but later shown to be A. cavernata). The sequences of Ascrobiculata AU303◂, BR224 and FO316 cluster far apart, together with Acaulospora tuberculata (VZ103E) on a clade sister to Acaulospora spinosa (W3574/Att165-9 – an ex-type culture, MN405B◂). For several short sequences the results are rather unclear, as they are only represented by one sequence or by sequences from different cultures that cluster apart from each other.

SSU-ITS-LSU phylogeny of the Diversisporales Diversisporaceae (Fig. 4b)

All data available for Pacispora have already been shown in Figs 1 and 2. For Diversispora, there are six described species (Schüßler & Walker, 2010), all characterized by rDNA sequences. The relatively short sequences of Diversispora sp. NB101 and Diversispora sp. AZ237B with stated origin from Namibia and Arizona, respectively, are very closely related. Including these decreases the BS for Diversispora celata as a monophyletic clade from 99% (not shown) to 62%. The Diversispora species are well supported, but, for both Diversispora spurca and Diversispora aurantia, two distinct clades appear in the phylogenetic analysis. One D. spurca clade is well defined by sequences from an ex-type culture (W4119/Att246-18) and contains a sequence of D. spurca WV109F◂. The second clade is composed of two sequences (FJ461848◂ and FJ461849◂) from other cultures and might represent another species. Despite the reasonable support of the D. aurantia clade, comprising sequences derived from the holotype trap culture (W4728/Att1296-0), two sequences from the same culture (EF581864 and EF581861) form a separated clade. The only sequence published for Glomus tortuosum JA306A (FJ461850◂) clusters in a basal polytomy. Three diverse ‘Diversispora trimurales’ sequences from the cultures KS101◂, FL707◂ and BR608◂ cluster at different positions throughout Diversispora and require further validation. The three species in Redeckera form a separate, well-supported clade (99% BS).

Entrophosporaceae– phylogenetically undefined

There are only two described species, Entrophospora baltica and Entrophospora infrequens (generic type), in the Entrophosporaceae. Additionally, E. nevadensis was recently described (Palenzuela et al., 2010), but its sequence clusters in the Diversispora clade (Fig. 2). Other database sequences annotated as Entrophospora species are mostly < 450 bp (e.g. AF378456523), environmental or uncharacterized, or should be annotated as Acaulospora (Kaonongbua et al., 2010). We excluded all Einfrequens sequences from the analyses as they were very short or showed high similarity with Claroideoglomus, Gigaspora or Rhizopus sequences (see Schüßler et al., 2003). Sequences from the cultures CA203◂ and IN215◂, all of which are of doubtful identity, also cluster within Claroideoglomus (not shown).

SSU-ITS-LSU phylogeny of the Glomerales Glomeraceae (Funneliformis, Septoglomus and Glomus; Fig. 5a)

Glomus in its strict sense currently comprises only G. macrocarpum (W5581&W5288/Att1495-0 and W5293/field-collected) and Glomus sp. W3347/Att565-7, which is morphologically similar to G. macrocarpum, but distinct because of a darker spore colour. One sequence attributed to Simiglomus hoi (BEG104) clusters with Glomus sp. W3347 and one of Glomus aggregatum (OR212◂) clusters basally to G. macrocarpum (Supporting Information Fig. S1). Funneliformis is well supported and represented by F. mosseae (75%), F. coronatus W3582/Att108-7 (100% BS), Funneliformis sp. WUM3 (100%) and F. caledonius BEG20 (97%). Septoglomus is represented by S. constrictum (100% BS).

When short sequences are included (Fig. S1), F. coronatus ZTL clusters with cultures W3582/Att108-7, BEG28, and IMA3. A sequence of BEG49 clusters apart, together with one from S. constrictum BEG130 and Funneliformis sp. WUM3 sequences. Funneliformis multiforus DAOM240256 is well supported; F. geosporus separates into two clades. For culture MD124 one ITS sequence annotated as Glomus geosporus (AF197918) clusters within Claroideoglomus (Fig. S2) and one LSU sequence (FJ461841◂) annotated as Gmacrocarpum clusters with F. geosporus (Fig. S1). Examination of MD124 (C. Walker W2843 in 1996 and W5729 in 2010) showed it to be Fgeosporus. Funneliformis caledonius sequences (BEG86, BEG20, DAOM234210, SC658, RMC658, RWC658 and JJ45) cluster unresolved. Several such discrepancies (e.g. for Funneliformis monosporus and Funneliformis dimorphicus) were already revealed by Stockinger et al. (2010). Sequences of Septoglomus deserticola, represented by an ex-type culture (BEG73; AJ746249), Septoglomus xanthium, and S. constrictum (NE202◂, UT188◂, 08-48-12, 08–48-17, 08-48-31) cluster in a separated clade, and a sequence from IN214A◂ forms another, basal and very long branch (Fig. S1). This also holds true for Glomus globiferum FL327B◂ and Glomus insculptum PL121◂, which were excluded from our analyses.

SSU-ITS-LSU phylogeny of the GlomeralesGlomeraceae (Rhizophagus; Fig. 5b)

For R. irregularis and R. intraradices, Stockinger et al. (2009, 2010) have already published detailed analyses. Here, we add new sequences from ‘Glomus cerebriforme’ DAOM227022 (not formally placed in Rhizophagus, because of uncertain species identification), Rhizophagus sp. MUCL46100, and several R. irregularis cultures (W4682/Att857-12, BEG195, DAOM197198, DAOM233750, MUCL46240, MUCL43205 and FTRS203). Rhizophagus irregularis, Rhizophagus sp. MUCL46100, R. intraradices (FL208 and MUCL49410), R. clarus W3776/Att894-7 and ‘G. cerebriforme’ DAOM227022, which clusters basally to all studied Rhizophagus species, are very well supported (96–100% BS). The weaker support for R. proliferus DAOM226389 (68% BS) is caused by the short sequence GQ205079 which is probably of chimeric origin. When short sequences are included, one from Glomus microaggregatum DAOM212150 clusters close to Rhizophagus sp. MUCL46100 (not shown), and one from G. microaggregatum UT216B◂ is located on a long branch within Claroideoglomus (Fig. S2). All three available Rhizophagus custos DAOM236381 sequence variants cluster among sequences of R. irregularis, as well as one ‘Glomus trimurales’ VA102A◂ sequence (not shown). One sequence of ML110◂ and two sequences annotated as ‘Glomus intraradices’ apparently are neither R. intraradices nor R. irregularis (Stockinger et al., 2009, 2010). Rhizophagus clarus sequences from 10 cultures cluster together with R. manihotis sequences in a well-resolved monophyletic clade. Sclerocystis sinuosa MD126 falls basal to Rhizophagus and Glomus achrum (FM25337981). Glomus bistratum (FM25338284) and Gindicum (GU05954449) cluster basally within Glomeraceae (formerly GlGrAb) in a polytomy (not shown).

SSU-ITS-LSU phylogeny of the GlomeralesClaroideoglomeraceae (Fig. 5c)

Claroideoglomus walkeri, Claroideoglomus drummondii and C. etunicatum are well supported, but C. claroideum is rendered paraphyletic by C. luteum SA101 sequences. The supplementary analysis including shorter sequences (Fig. S2) shows a number of sequences from additional C. etunicatum cultures (AU401, NB119, CA-OT-126-3-2, KE118, etc.) that are unresolved. Sequences of C. drummondii also form a well-supported clade. Claroideoglomus luteum, C. claroideum and a sequence annotated as G. microaggregatum UT126B◂ cluster unresolved.


By publishing > 240 further sequences produced over recent years and re-analyses of available sequences, we have established what we consider a phylogenetic basis for a natural systematics of Glomeromycota and a phylotaxonomic reference database for future environmental (deep) sequencing. For some analyses, we used consensus sequences, which are theoretical constructs and in some instances have to be interpreted with care (Lindner & Banik, 2011). However, in our analyses the use of strict SSU consensus sequences (degenerate base symbols represent all variations) anchors taxa by conserved regions and thus reduces the risks of phylogenetic attraction by shared characters at mutationally saturated sites. We analysed nuclear rDNA sequence data of c. 109 described species and c. 27 as yet unnamed AMF cultures (note that these are approximate numbers, because the species determination may not always be correct). More than 50% (120 species) of the currently c. 230 validly described AMF species are covered by sequences deposited in the public databases, but only 81 (c. 35%) are propagated in the culture collections INVAM (, BEG (Glomeromycota in vitro collection;, and GINCO (, making re-analyses of or improvements to the sequence database difficult.

Need for a solid molecular genetic basis for the systematics of Glomeromycota

SSU analyses (Schüßler et al., 2001) and the six-gene phylogeny of James et al. (2006) indicated a likely sister grouping of the Glomeromycota to Dikarya. By including basal fungal lineages as well as members of Dikarya, we again found the same sister grouping (Fig. 1). However, analyses of the mitochondrial genome of R. irregularis isolate 494 (Lee & Young, 2009) and of nucleus-encoded amino acid sequences (Liu et al., 2009) questioned this relationship and indicated a possible common ancestry of AMF with Mortierellales, although tree topologies in the latter study varied depending on taxon sampling. Therefore, we must await more data from phylogenetically basal AMF to resolve immediate sister relationships to Glomeromycota, which are nonetheless clearly monophyletic and phylogenetically basal terrestrial fungi.

The present data compilation and analyses formed part of the basis for a major taxonomic reclassification in the Glomeromycota (Schüßler & Walker, 2010), and it is expected to be important as a reference for new species descriptions. For example, the sole use of morphology for the description of Ambispora brasiliensis (Goto et al., 2008) placed an Acaulospora species incorrectly at generic, familial and even ordinal level (Krüger et al., 2011). Similar instances of problematic species descriptions only based on morphology were discussed by Morton & Msiska (2010b), who reported an albino mutant of S. heterogama WV859, which would have been considered as a new morphospecies if found in the field. Another example was the description of Glomus irregulare (Błaszkowski et al., 2008), now R. irregularis, which was mainly based on a limited analysis of intraspecific morphological plasticity. Therefore, including an accurate phylogenetic characterization should improve the quality of formal species descriptions whenever possible. Obviously, this is particularly important for species not represented by publicly available isolates.

Phylogenetically basal lineages –Paraglomerales and Archaeosporales

Only relatively few data are available for evolutionarily ancient phylogenetic lineages of Glomeromycota. Presently there are only three recognized or described species in the Paraglomerales and 11 in the Archaeosporales, but this is probably only a small proportion of all existing species. Our study is the first to yield significant branch support for Paraglomerales as the most ancient lineage of the Glomeromycota (Fig. 1). It also supports Intraspora (Sieverding & Oehl, 2006) as congeneric with Archaeospora (Schüßler & Walker, 2010).


There has been considerable nomenclatural change among the Diversisporales recently. Oehl et al. (2008) split the genus Scutellospora into three new families containing six genera (Scutellospora in the Scutellosporaceae; Racocetra and Cetraspora in the Racocetraceae; Dentiscutata, Fuscutata, and Quatunica in the Dentiscutataceae). Except for Racocetra, Morton & Msiska (2010a) rejected all these new taxa. Nevertheless, it has long been indicated that Scutellospora is nonmonophyletic (e.g. Kramadibrata et al., 2000; da Silva et al., 2006). We support the notion of Morton & Msiska (2010a) that a robust taxon sampling and phylogenetic analysis should form the basis of taxonomic changes; the phylogeny presented herein may provide support for at least some of the genera proposed by Oehl et al. (2008), but certainly not for erecting new families in this clade.

The finding of two different D. aurantia clades exemplifies problems in interpretation of data derived from trap cultures seemingly producing spores of one species (often called single species cultures). It seems possible, but cannot be proved, that the trap culture material with the spores of D. aurantia contained more than one species. For the monospecific genus Otospora (Palenzuela et al., 2008), the O. bareae sequences cluster within Diversispora. This could support the view that O. bareae is a morphologically exceptional member of the Diversisporaceae, but it is perhaps more likely to be the result of a contamination. The sequence of the recently described E. nevadensis (Palenzuela et al., 2010) also clusters unexpectedly, in terms of its morphology, among those of Diversispora. A detailed analysis of Diversisporaceae with focus on D. epigaea, often named ‘Glomus versiforme BEG47’, and including biogeographical aspects is given in Schüßler et al. (2011).

Kuklospora sensu Oehl & Sieverd. (Sieverding & Oehl, 2006) was described based solely on spore morphology. The recent transfer of all Kuklospora species to Acaulospora (Kaonongbua et al., 2010) is congruent with our analyses. In our opinion the species Acaulospora laevis and Acaulospora entreriana are morphologically indistinguishable. They could not be separated in analyses when ITS1 and ITS2 were excluded, but additional data are needed to investigate a possible conspecificity. This also holds true for cultures annotated as Acaulospora mellea, Acaulospora delicata and Acaulospora dilatata, which are not well covered by available long sequences.


A decade ago, Schwarzott et al. (2001) were already proposing that Glomus should be split into several families. These subsequently were operationally named as phyloclades Glomus Group A (GlGrA), GlGrB and GlGrC, until it became clear where the generic type of Glomus, G. macrocarpum, belongs phylogenetically (Schüßler & Walker, 2010). Now, the family Glomeraceae represents the former GlGrA, separated into six genera: Glomus (GlGrAc), Funneliformis and Septoglomus (both GlGrAa), Rhizophagus and Sclerocystis (both GlGrAb). Glomus iranicum and Gindicum sequences form a basal clade in this family, and G. bistratum and G. achrum cluster in a basal polytomy in the Glomeraceae. Robust phylogenetic placements of the last four species and of the proposed monospecific genus Simiglomus (Oehl et al., 2011a) will require additional data. The family Claroideoglomeraceae corresponds to the former GlGrB, and the Diversisporaceae to GlGrC.

For Claroideoglomus, Funneliformis and Rhizophagus, detailed analyses have already been conducted by Stockinger et al. (2010), under the previous generic name Glomus. The uncovered inconsistencies discussed in that study are also recognizable from the phylogenetic trees of the present study, but are not further discussed here. Rhizophagus irregularis was defined (Błaszkowski et al., 2008), as G. irregulare, mainly based on perceived morphological differences from G. intraradices in a former sense, which included DAOM197198. The analysis shown in Fig. 5(b) confirms that the organisms interpreted as different, based on morphology, in fact belong to the same species. Glomus irregulare (now R. irregularis) is conspecific with DAOM197198 (and other cultures of ‘G. intraradices’ in the former sense) (Stockinger et al., 2009, 2010; Sokolski et al., 2010). The molecular data suggest that R. clarus and R. manihotis are conspecific, but this possible synonymy requires further morphological work before taxonomic assessment.

Putative errors in public sequence databases

As discussed repeatedly (e.g. Schüßler et al., 2003; Bidartondo et al., 2008), annotation of sequence entries in public databases is often inadequate or incorrect. There are different types of error; some errors are based on wrong identification or undiscovered species synonymy, some on pot culture or laboratory contaminants, and others perhaps on accidental misannotation. For example, a batch of LSU sequences (FJ461790FJ461888◂) caused problems in our initial analyses because of numerous species falling into unexpected groups, until we realized that many of the contained sequences seem to be either misannotated or derived from contaminants. Sequences from ‘G. trimurales’, originally annotated as Glomus sp., fell among three different orders, in the genera Diversispora, Claroideoglomus and Rhizophagus. Several entries will be updated (J. Morton, pers. comm.). Our own past errors include the annotation of A. cavernata BEG33 as A. scrobiculata, and mixing up two samples, resulting in mistakenly naming sequences of S. spinosissima W3009/Att664-1 as S. nodosa BEG4 and vice versa. Moreover, we doubt our own annotation of a sequence (Y17652) that was recently used in the description of the monospecific genus Viscospora, now containing ‘Viscospora viscosa’ BEG27. Morphologically, affiliation to Claroideoglomus was surprising and the culture used for sequencing later was found to contain a contaminant C. claroideum-like fungus. A revived culture of BEG27 has been established and will be used to clarify this matter. A surprising issue regarded sequences (JF276401JF276423) from one database submission, which all have identical counterparts in the database, including short 88- or 211-bp sequences. They must be derived from resubmission of already existing sequences. The ‘Glomus’ SSU sequences JF276412,17,18 and JN040742 are not from members of the Glomeromycota.

An example of putative culture misannotation is DAOM212349. The original voucher number refers to both the C. lamellosum holotype (field-collected) and a pot culture from which specimens designated as ‘isotype’ (which cannot be correct, as, by definition, an isotype has to be from the original type collection) were derived (Dalpéet al., 1992). A later in vitro ROC established from this pot culture was given the same number in the GINCO database, but it contains R. irregularis. Therefore, DAOM212349 must represent either an initially mixed culture or an instance of later contamination. Obviously, the same identifier is used for fungi belonging to two distinct genera. Failure to update public database sequences, to correct errors or to implement taxonomic changes can cause confusion and impair accurate interpretation of analyses. To facilitate correct interpretation of AMF sequence data, third-party annotations are currently implemented in the PlutoF (Abarenkov et al., 2010) based fungal reference sequence database, which now also includes curated metadata for mycorrhizal fungi (Tedersoo et al., 2011).


Systematics and molecular phylogenetics influence more scientific disciplines than is often realized (Mayr, 1968). It is important to correct misclassifications of organisms, the functional, genetic, and ecological traits of which are best interpreted on phylogenetic grounds.

Besides providing a solid phylogenetic backbone, the data set presented here covers the future primary DNA barcode for fungi, namely the ITS region, and the 5′ portion of the LSU, which will be used as an extended barcode. The use of long sequences covering conserved as well as variable regions solves problems intrinsic to the use of short sequences. This will also assist detection of species in the field, but the database will have to be broadened with respect to sequence and taxon coverage. The latter relates to described species, but also to environmental MOTUs, for which species-level recognition is feasible through the use of the SSU-ITS-LSU fragment (Stockinger et al., 2010; Krüger et al., 2011). A frequent problem with AMF is a lack of well-characterized biological material from described species, as many of these biotrophic fungi have so far proved impossible or difficult to culture, even in pots together with host plants. This problem could be alleviated by contributing more isolates (single spore cultures) to public culture collections, which, however, seem to be limited by inadequate funding. This is an unfortunate situation, given the fact that AMF are integral components of nearly all terrestrial ecosystems.

To discover more about AMF–plant preferences and the functional roles of AMF, a solid systematic classification is indispensable. This study provides a reference guide for molecular species identification and phylotaxonomy that will be important for future molecular ecological studies, including the application of next-generation sequencing strategies. More sequences with sufficient lengths would, moreover, facilitate improved understanding of the biogeography and evolution of AMF. In addition, research in practical fields, such as biosafety assessments regarding the impact of genetically modified plants on AMF communities or AMF species traceability in field applications, may depend on a solid data baseline.


We thank all the people who have provided material, the anonymous reviewers for detailed comments, and Milagros Lovera for pointing out the problem with the S. nodosa and S. spinosissima sequences. Part of this work was financed by the German Research Foundation (DFG grant Schu1203/8) and the Marie Curie Early Stage Research Training Fellowship of the European Community’s Sixth framework Programme (MEST-CT-2005-021016).