Detection of a novel intracellular microbiome hosted in arbuscular mycorrhizal fungi

Abstract

Arbuscular mycorrhizal fungi (AMF) are important members of the plant microbiome. They are obligate biotrophs that colonize the roots of most land plants and enhance host nutrient acquisition. Many AMF themselves harbor endobacteria in their hyphae and spores. Two types of endobacteria are known in Glomeromycota: rod-shaped Gram-negative Candidatus Glomeribacter gigasporarum, CaGg, limited in distribution to members of the Gigasporaceae family, and coccoid Mollicutes-related endobacteria, Mre, widely distributed across different lineages of AMF. The goal of the present study is to investigate the patterns of distribution and coexistence of the two endosymbionts, CaGg and Mre, in spore samples of several strains of Gigaspora margarita. Based on previous observations, we hypothesized that some AMF could host populations of both endobacteria. To test this hypothesis, we performed an extensive investigation of both endosymbionts in G. margarita spores sampled from Cameroonian soils as well as in the Japanese G. margarita MAFF520054 isolate using different approaches (molecular phylotyping, electron microscopy, fluorescence in situ hybridization and quantitative real-time PCR). We found that a single AMF host can harbour both types of endobacteria, with Mre population being more abundant, variable and prone to recombination than the CaGg one. Both endosymbionts seem to retain their genetic and lifestyle peculiarities regardless of whether they colonize the host alone or together. These findings show for the first time that fungi support an intracellular bacterial microbiome, in which distinct types of endobacteria coexist in a single cell.

Introduction

The discovery that the human body can be described as a complex ecosystem where human cells interact with trillions of bacteria and other microbes has represented a scientific revolution. The human microbiome, that is, the microbial communities and the genetic information they contain, cooperate with the human genome to regulate crucial physiological processes ranging from digestion to obesity and immunity (Methé et al., 2012). Similarly, plants rely on microorganisms living both in their tissues and in the rhizosphere, creating a network of mutual relationships (Porras-Alfaro and Bayman, 2011; Berendsen et al., 2012; Bulgarelli et al., 2012: Lundberg et al., 2012). To date, most of the work on plant-associated microbes focused almost exclusively on bacteria (Bulgarelli et al., 2012; Lundberg et al., 2012), even though eukaryotes such as fungi are also crucial components of the plant microbiome. They not only thrive in the rhizosphere, but also colonize plant tissues, exhibiting a range of lifestyles, including mutualism, parasitism and commensalism (Porras-Alfaro and Bayman, 2011).

Among plant-associated microbiota, arbuscular mycorrhizal fungi (AMF) are the most widespread: they belong to an ancient monophyletic phylum, the Glomeromycota (Schüßler et al., 2001), and have a key role in nutrient cycling and plant health due to their capacity for improving the mineral nutrition of plants (Smith and Read, 2008). AMF display many unusual biological features. In addition to their obligate biotrophy (Bonfante and Genre, 2010), many of them harbor endobacteria in their cytoplasm (Bonfante and Anca, 2009). Bacterial endosymbionts are widespread among animals (Wernegreen, 2012; McFall-Ngai et al., 2013) and in particular the ones living in insect tissues have been investigated in depth (Ferrari and Vavre, 2011). In contrast, examples of endobacteria living inside the fungal cells are much more limited (Bianciotto et al., 2003; Partida-Martinez and Hertweck, 2005; Lackner et al., 2009; Naumann et al., 2010; Kai et al., 2012).

The endobacteria of Glomeromycota are the most thoroughly investigated bacterial endosymbionts of fungi, having been discovered in the early 1970s on the basis of electron microscope observations (Mosse, 1970). Two types of endosymbionts are known in AMF: (i) a rod-shaped, Gram-negative beta-proteobacterium (Bonfante et al., 1994), Candidatus Glomeribacter gigasporarum (CaGg), common in several species of the family Gigasporaceae (Bianciotto et al., 2003; Mondo et al., 2012), and (ii) a coccoid bacterium displaying a homogeneous Gram-positive-like wall structure (MacDonald et al., 1982; Scannerini and Bonfante, 1991), which represents a currently undescribed taxon of Mollicutes-related endobacteria (Mre) with a wide distribution across Glomeromycota (Naumann et al., 2010).

The CaGg genome sequence (Ghignone et al., 2012) revealed that Glomeribacter endobacteria are nutritionally dependent on the fungal host and have a possible role in providing the fungus with essential factors like vitamin B12 (Ghignone et al., 2012). Phenotypic consequences of CaGg removal from the host include important morphological changes as well as reduced proliferation of host presymbiotic hyphae. Yet, the host is not obligately dependent on the bacteria (Lumini et al., 2007; Mondo et al., 2012). These features suggest that Glomeribacter endobacteria are mutualistic associates of AMF (Lumini et al., 2007). Comparisons of host and symbiont phylogenies indicate that, while CaGg is a heritable endosymbiont (Bianciotto et al., 2004), it also engages in recombination and host switching, which play an important role in stabilizing this 400-million-year-old association (Mondo et al., 2012). In contrast, information on the coccoid Mre is much more limited. Based on the 16S rRNA gene sequences, this novel lineage is sister to a clade encompassing the Mycoplasmatales and Entomoplasmatales (Naumann et al., 2010). The Mre have been detected in 17 out of 28 investigated AMF samples from culture collections, including members of Archaeosporales, Diversisporales, Glomerales (Naumann et al., 2010), as well as in mycorrhizal thalli of liverworts (Desirò et al., 2013). In most of the AMF hosts and irrespectively of the AMF identity, these endobacteria displayed a conspicuous variability in their 16S rRNA gene sequence. Collectively, these observations indicate that CaGg is a stable associate of Gigasporaceae, whereas the lifestyle of the Mre and the nature of their association with Glomeromycota are uncertain. Furthermore, the interaction between the two endosymbionts remains unclear, that is, it is not known whether the presence of one endosymbiont in the host leads to the exclusion of the other one.

The goal of the present study is to investigate the patterns of distribution and coexistence of the two endosymbionts, CaGg and Mre, in isolates of one host species, Gigaspora margarita WN Becker & IR Hall. Previous electron microscopy observations revealed that the strain of G. margarita MAFF520054 harbored a Gram-positive-like endobacterium (Kuga et al., 2008), whereas molecular analyses indicated the presence of CaGg (E Lumini, personal communication, ref. seq. AM886455; Long et al., 2009). Based on these observations, we hypothesize that some AMF could host populations of both endobacteria. To test this hypothesis, we performed an extensive investigation of both endosymbionts in G. margarita spores sampled from Cameroonian soils as well as in G. margarita MAFF520054 from Japan using different approaches. We found that a single AMF host can harbor both types of endobacteria, with Mre populations being more abundant, variable and prone to recombination than the CaGg ones. These findings show for the first time that fungi support an intracellular bacterial microbiota, in which distinct types of endobacteria coexist in a single cell.

Materials and methods

All the details of the experimental procedures are available in the Supplementary Text S1.

Sampling and sample preparation

Twelve soil samples were collected from three locations in Cameroon (Table 1). Trap cultures with Sorghum and Vigna were established using autoclaved sand mixed with the sampled soils. The Japanese isolate G. margarita MAFF520054 was provided by NIAS Genebank and propagated in pot cultures with Trifolium.

Table 1. List of the spore samples studied in this work.

Sample	Origin	AM species	Endobacteria
CM2	Cameroon: Nkolbisson, Yaoundé	Gigaspora margarita	Mre+CaGg
CM3	Cameroon: Nkolbisson, Yaoundé	Gigaspora margarita	Mre
CM9	Cameroon: Nkolbisson, Yaoundé	Gigaspora margarita	Mre+CaGg
CM21	Cameroon: Nkolbisson, Yaoundé	Gigaspora margarita	Mre+CaGg
CM23	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre+CaGg
CM27	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre+CaGg
CM46	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre+CaGg
CM47	Cameroon: Maroua	Gigaspora margarita	Mre+CaGg
CM49	Cameroon: Nkolbisson, Yaoundé	Gigaspora margarita	Mre+CaGg
CM50	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre+CaGg
CM51	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre+CaGg
CM52	Cameroon: Nkoemvone, Ebolowa	Gigaspora margarita	Mre
MAFF520054	Japan: Saitama	Gigaspora margarita	Mre+CaGg

Collection sites, fungal species and endobacteria typology are shown for each sample.

The spores were recovered from pot cultures by wet sieving (Gerdemann and Nicolson, 1963) and surface sterilized (Lumini et al., 2007). The spore samples were morphologically identified as Gigaspora margarita following Bentivenga and Morton (1995).

DNA extraction, amplification and clone library analysis

DNA extractions were performed by crushing either individual spores or groups of five or ten spores according to Lumini et al. (2007). Three fragments of the fungal ribosomal gene cluster, namely 18S, ITS and 28S, were amplified.

The CaGg 16S rRNA gene was specifically amplified with the newly designed primers CaGgADf (5′-AGATTGAACGCTGGCGGCAT-3′) and CaGgADr (5′-ATGCGTCCTACCGTGGCCATC-3′), while the Mre 16S rRNA gene was amplified as described in Desirò et al. (2013). Fungal and bacterial PCR amplicons were transformed and the obtained clone libraries were analyzed.

Bioinformatics

Sequences were assembled and curated in Mega v. 5.2 (Tamura et al., 2011), aligned with MAFFT (Katoh et al., 2002) or MUSCLE (Edgar, 2004) and then examined for chimerism. Sequence similarity/divergence was evaluated using MOTHUR (Schloss et al., 2009). Nucleotide diversity (π) was calculated in DNAsp v. 5.10.01 (Librado and Rozas, 2009). The CaGg and Mre 16S rRNA gene sequences were grouped into operational taxonomic units at the cut-off of 0.03 genetic distance value using MOTHUR. Phylogenetic analyses were conducted using one representative sequence for each OTU. The Genetic Algorithm for Recombination Detection (Kosakovsky Pond et al., 2006), was used to identify recombination breakpoints in 16S rRNA genes of CaGg and Mre.

Ultrastructural analysis

Single G. margarita spores from CM23 and CM47 samples were processed by using high-pressure freezing followed by freeze substitution. Single spores floating in water were transferred in the cavity of an aluminum carrier with a pipette. Excess of water was drawn off with filter paper and the space was filled with 1-Hexadecene. The sandwich was completed with a flat specimen carrier and frozen in a HPM 100 high-pressure freezing machine (Leica Microsystems, Wetzlar, Germany) (McDonald et al., 2010). Samples were then freeze substituted, resin embedded and processed for transmission electron microscopy.

FISH experiments and confocal microscopy

Sterilized spores of the samples CM21, CM23, CM47, CM52 and G. margarita BEG34 were fixed as described in Naumann et al. (2010). The Mre-specific probe BLOsADf2 (Desirò et al., 2013), together with a newly designed CaGg-specific 16S rRNA probe (CaGgADf1 5′-CTATCCCCCTCTACAGGAYAC-3′), were used to label the endobacteria. In addition, the eubacterial probe EUB338 (Amann et al., 1990) and the Buchnera-specific probe ApisP2a (Koga et al., 2003) were used. Spores were observed using a Leica TCS-SP2 confocal microscope (Leica Microsystems).

Quantification of the bacterial populations

The sample CM23 (containing both Mre and CaGg) was selected for the relative quantification of the two bacterial populations by real-time quantitative PCR (qPCR). Briefly, the 16S rRNA gene sequences obtained for both CaGg and Mre were used to design two distinct qPCR primer pairs. Template plasmids containing the target DNA sequences were constructed to generate a standard curve as an external standard. The number of target DNA sequences present in each PCR mixture was calculated by comparing the crossing points of the samples with those of the standards.

Results

Identity of AMF

To confirm the morphological identification of AMF originating from Cameroon and Japan as Gigaspora margarita, we analyzed their 18S, ITS and 28S rRNA gene regions. These analyses revealed that all the fungi could be identified as G. margarita (Figure 1 and Supplementary Figure S3). As expected, the 18S rRNA gene analysis led to an unresolved, polytomic phylogeny (not shown), whereas a better resolution was provided by the ITS region (Figure 1) and the 28S rRNA gene (Supplementary Figure S3). DNA sequences are available in GenBank (KF378653-KF378691).

Figure 1.

Phylogenetic placement of Cameroonian and Japanese spore samples inside the Gigasporaceae tree. The fungal phylogeny was inferred from ITS1, 5.8S rRNA gene and ITS2 sequences. The DNA sequences retrieved in this work are in bold. All the thirteen spore samples are located inside the Gigaspora clade, close to Gigaspora margarita. Supported values are from Bayesian/maximum likelihood/maximum parsimony analyses. The partitioned Bayesian analysis was performed with JC, K80+G, and HKY+G nucleotide substitution models for ITS1, 5.8S and ITS2 regions, respectively. The maximum likelihood analysis was performed with GTR+CAT nucleotide substitution model. Dashes instead numbers imply that the topology was not supported in the respective analysis.

Identity of endobacteria

Bacterial 16S rRNA gene sequences were PCR-amplified from single AMF spores using primers specific for CaGg and Mre (Naumann et al., 2010) to detect endosymbiont presence. Most samples harbored both types of endobacteria with the exception of the G. margarita samples CM3 and CM52, which contained only Mre (Table 1). The absence of CaGg in the samples CM3 and CM52 was confirmed by real-time qPCR (data not shown), which can detect up to 10 bacterial cells (Salvioli et al., 2008).

In order to faithfully describe the microbiome contained inside the AMF spores and to capture all of the bacterial biodiversity, a more extensive analysis was performed on pools of 10 spores from four Cameroonian samples (CM21, CM23, CM47, CM50) and from the Japanese isolate.

The RFLP analysis of CaGg 16S rRNA gene sequences revealed a single RFLP profile for each 10-spore sample, suggesting a limited intra-sample variability that was further confirmed by sequence analysis. The obtained sequences (∼1460 bp) were grouped into operational taxonomic units at 97% of sequence similarity and, as expected, a single OTU for each sample was obtained (Table 2). Phylogenetic analyses of CaGg sequences retrieved from spore samples showed that they clustered with other CaGg sequences available in GenBank (Figure 2).

Table 2. Candidatus Glomeribacter gigasporarum sequences generated from the selected Gigaspora margarita spore samples.

G. margaritaa	Retrievedb	OTUc	Sequences ind
sample	sequences	number	OTU1
CM21	3	1	3
CM23	4	1	4
CM47	5	1	5
CM50	4	1	4
MAFF520054	4	1	4

^aGigaspora margarita spore sample.

^bNumber of Candidatus Glomeribacter gigasporarum generated sequences.

^cNumber of OTUs and

^dtheir related sequences.

Figure 2.

Phylogenetic placement of representative Candidatus Glomeribacter gigasporarum partial 16S rRNA gene sequences retrieved from spores of AMF. The DNA sequences retrieved in this work are in bold. The tree encompasses several CaGg groups. Sequences from Gigaspora margarita sample CM47 and CM50 cluster in a group sister to the one (with thickened branches) including CaGg from G. margarita BEG34 isolate (highlighted in gray) and from the Cameroonian CM21 and CM23 samples. The 16S rRNA gene sequences from the Japanese sample MAFF520054 are located in a different and more basal position inside the tree, together with other CaGg sequences retrieved from worldwide G. margarita isolates. Cameroonian isolates showed 97–100% sequence similarity with Gigasporaceae isolates (that is, Gigaspora decipiens, Gigaspora gigantea, G. margarita, including the isolate BEG34, Gigaspora rosea, Racocetra castanea and Racocetra verrucosa), which are located in the upper part of the tree. By contrast, CaGg sequence similarity, in particular of the samples CM47 and CM50, decreased to 96% relative to CaGg sequences retrieved from other worldwide isolates of Cetraspora pellucida and G. margarita, including the G. margarita isolate MAFF520054. Supported values are from maximum likelihood/Bayesian/maximum parsimony analyses. The maximum likelihood and Bayesian analyses were performed with GTR+G and TIM3+G nucleotide substitution models, respectively. Dashes instead numbers imply that the topology was not supported in the respective analysis.

Sequencing of the Mre 16S rRNA gene clones generated a total of 118 sequences (Table 3). To eliminate potential PCR artefacts expected in amplifications from complex templates such as Mre populations (Naumann et al., 2010), the obtained sequences (1049–1087 bp) were submitted to a rigorous chimera screen, which reduced the total amount to 52 sequences (Table 3). They were grouped into operational taxonomic units at 97% sequence similarity (Table 3). Most of the sequences (48 out of 52) showed sequence similarity values lower than 97% when compared with the Mre sequences obtained from GenBank, suggesting the presence of novel phylotypes (Table 3).

Table 3. Mollicutes-related endobacteria sequences generated from the selected Gigaspora margarita spore samples.

G. margaritaa sample	Retrievedb sequences	Sequences afterc chimera screen	OTUd number	Sequences ine			Nucleotide diversity (π)f up to	Sequenceg divergence (%)
				OTU1	OTU2	OTU3
CM21	14	14	2	13	1	—	0.1764	20
CM23	38	7	3	3	2	2	0.1882	20
CM47	31	12	3	5	4	3	0.1608	17
CM50	10	10	1	10	—	—	—	—
MAFF520054	25	9	3	4	4	1	0.1590	17

^aGigaspora margarita spore sample.

^bNumber of Mollicutes-related endobacteria generated sequences.

^cNumber of sequences after the chimera removal.

^dNumber of OTUs and

^etheir related sequences.

^fThe highest values of nucleotide diversity and

^gsequence divergence between two representative sequences of different OTUs of the same sample.

Despite the high variability, all retrieved Mre sequences clustered together with those obtained in previous studies (Naumann et al., 2010; Desirò et al., 2013) (Figure 3). Moreover, because the resulting phylogenies presented here are better supported and resolved than those constructed in previous works (Naumann et al., 2010; Desirò et al., 2013), we conclude that there are at least two distinct and well-supported Mre clades, identified as Mre group A and group B (Figure 3), and that the level of sequence divergence among sequences clustering in the same Mre group reached up to 15 and 16% in Mre group A and B, respectively. Overall, in all the samples, with the only exception of CM50, CaGg showed a high level of intra-host sequence similarity, whereas Mre revealed high levels of intra-host sequence diversity.

Figure 3.

Phylogenetic placement of representative Mollicutes-related endobacteria partial 16S rRNA gene sequences retrieved from AM spores within the Mollicutes clade. The DNA sequences retrieved in this work are in bold. The tree encompasses at least two main and well-supported groups (Mre group A and B), which also include sequences retrieved in previous experiments from AM spore collection (Naumann et al., 2010) and AMF liverworts-associated (Desirò et al., 2013). The number of sequences included in each OTU is in brackets. Supported values are from Bayesian/maximum likelihood/maximum parsimony analyses. The Bayesian and maximum analyses were performed with GTR+G and GTR+CAT nucleotide substitution models, respectively. Dashes instead of numbers imply that the topology was not supported in the respective analysis.

Representative DNA sequences are available in GenBank (KF378648-KF378652, KF378692-KF378705).

Recombination detection

To explore the underlying causes of differences in sequence evolution patterns between CaGg and Mre, we used the Genetic Algorithm for Recombination Detection (Kosakovsky Pond et al., 2006) to look for evidence of recombination in 16S rRNA genes of the two endosymbionts associated with AMF from Cameroon and Japan. No evidence of recombination was detected in the CaGg sequences. In contrast, in the Mre data set, we found that the AIC_C score of 8529.9 for the best-fitting model allowing for different topologies of the alignment segments defined by recombination breakpoints was lower than the AIC_C score of 8819.4 for the model that assumed the same topology for all segments, indicating that a multiple tree model is preferable over a single tree model. Using the KH test, one breakpoint at the alignment position 479 was identified as resulting in significant topological incongruence between segments (P<0.001, Supplementary Figure S4).

Localization of the two bacterial morphotypes in AMF cells: high pressure/freeze substitution and transmission electron microscopy

We used electron microscopy to confirm the cytoplasmic location of both types of endobacteria. To ensure proper preservation of endosymbiont cells and fungal organelles, which could be jeopardized by the very thick fungal cell wall (12–16 μm, Lumini et al., 2007), we used high pressure and freeze substitution specimen preparation. On the basis of the previous molecular analyses, two isolates of G. margarita (CM23 and CM47) were selected for this experiment. When inspected under the transmission electron microscope, CM23 and CM47 presented both the rod-shaped and coccoid bacteria in the same area of their cytoplasm (Figure 4). The rod-shaped CaGg were 330–550 × 960–1050 nm in size, with a layered, Gram-negative type cell wall (Figures 4a and b) and were located inside a vacuole-like organelle (Figure 4a), consistent with reports from earlier studies (Bianciotto et al., 1996, 2003). The vacuole revealed an electron-dense matrix, which was identified as of protein origin (Bonfante et al., 1994) (Figure 4a). In other cases, the matrix was reduced in size and the bacterium was more closely surrounded by a membrane of fungal origin (Figure 4b). In contrast, the coccoid Mre were directly embedded in the fungal cytoplasm (Figures 4a and c). They were consistently smaller, 300–600 nm in size, with a homogeneous, Gram-positive-like cell wall (Figure 4c).

Figure 4.

Electron microscopy of Gigaspora margarita sample CM23. (a) The two bacterial types, Candidatus Glomeribacter gigasporarum (arrow) and Mollicutes-related endobacteria (arrowhead) are present in the same district of the sporal fungal cytoplasm (fc). The rod-shaped type is constantly located inside a vacuole-like organelle (v). The vacuole reveals an electron-dense matrix (m), identified as of protein origin. (b) Sometimes CaGg (here cut in a transversal section) is more closely surrounded by a membrane of fungal origin (arrow). (c) The Mre is directly embedded in the fungal cytoplasm. Scale bars, (a) 1.5 μm; (b) 0.26 μm; (c) 0.17 μm.

Localization of the two endosymbionts in AMF spores: FISH

To further validate our molecular and morphological observations of the CaGg and Mre coexistence in G. margarita, we performed fluorescence in situ hybridization (FISH) experiments in samples CM21, CM23, CM47 and CM52. G. margarita BEG34 was used as negative control, as Mre have never been found in this isolate (Naumann et al., 2010). We used two probes: CaGgADf1, which was designed to specifically detect CaGg, and BLOsADf2 (Desirò et al., 2013), which targeted entire Mre variability contained in our spore samples. In agreement with PCR results, we did not observe any CaGg signal in CM52, where CaGg have never been detected by PCR amplification of 16S rRNA gene. Similarly, we did not observe any Mre signal in BEG34. On the contrary, the two specific probes produced simultaneous FISH signals in the spores where the presence of both bacterial types was expected (Figure 5). Image analysis on 16 confocal microscope images from spore samples containing the two bacterial populations revealed that Mre were 1.62–3.15 times more abundant than CaGg (Supplementary Table S1).The fluorescent signals were located in the fungal cytoplasm and never on the spore surface. Importantly, the fluorescent signal of the probes BLOsADf2 (Desirò et al., 2013) and CaGgADf1 was always co-localized with the fluorescence given by the universal bacterial probe EUB338 (Amann et al., 1990) (Figures 5b–e). No fluorescent signal was detected with the negative control probe ApisP2a (Koga et al., 2003) (Figure 5i). Pre-treatment with RNase, as well as control hybridization with nonsense probes, did not provide any FISH signal. A weak autofluorescence of the fungal cytoplasm, probably derived from the use of aldehydic fixatives, was visible in all spore samples. Hence, FISH experiments, validating the PCR results, confirmed the simultaneous presence of Mre and CaGg in some G. margarita samples.

Figure 5.

FISH on crushed spores of Gigaspora margarita samples CM21 (a-e) and CM23 (f-i). (a) Bright-field image of the fungal cytoplasm (fc) trapped in a drop of agarose. (b) Triple labelling of the endobacteria with the Mollicutes-related endobacteria-specific probe BLOsADf2 (red), the Candidatus Glomeribacter gigasporarum-specific probe CaGgADf1 (blue) and the universal bacterial probe EUB338 (green); bacteria are seen as coccoid or rod-shaped fluorescent spots (arrowheads); in this image, where red and green or blue and green channels are overlaid, bacteria are visualized as fluorescent orange or light blue spots inside the brown cytoplasm. The corresponding red, blue and green channels are shown in (c-e). (f) Triple labelling of the endobacteria with the Mre-specific probe BLOsADf2 (red), the CaGg-specific probe CaGgADf1 (blue) and the Buchnera-specific probe ApisP2a (green) used as negative control; bacteria are seen as coccoid or rod-shaped fluorescent spots (arrowheads). The corresponding red and blue channels are shown in (g, h). (i) No presence of non-specific fluorescent signal is detected. The insets show the magnification of some Mre and CaGg cells surrounded by the fungal cytoplasm. Scale bars: 12 μm, 3 μm in the insets.

Mre and CaGg abundance in AMF cells: real-time qPCR

To further examine differences in Mre and CaGg abundance suggested by FISH experiments, we used real-time qPCR to quantify the bacterial populations present in the G. margarita sample CM23 that was previously shown to contain both Mre and CaGg endobacteria. The 16S rRNA gene was used as a target gene, but while in the CaGg genome the 16S rRNA gene is present in a single copy (Ghignone et al., 2012), in Mre one or at most two rRNA gene copies are expected based on the comparison with the closest microbes already sequenced (Fraser et al., 1995; Glass et al., 2000; Jaffe et al., 2004; Minion et al., 2004; Vasconcelos et al., 2005; Bai et al., 2006).

The accuracy of qPCR primers of CaGg and Mre was confirmed by assessing the melting profile generated by each primer pair (Supplementary Figure S2). Subsequently, we quantified the relative abundance of the two bacterial endosymbionts on the basis of the 16S rRNA gene sequences. In G. margarita CM23, we found that Mre were always more abundant than CaGg, and the bacterial ratio was maintained fairly constant irrespective of the size of the batches considered (that is, one, five or ten spores) (Table 4).

Table 4. Quantification of Mollicutes-related endobacteria and Candidatus Glomeribacter gigasporarum detected with real-time qPCR.

Batch	Biological replicates	Mre average	Mre standard deviation	CaGg average	CaGg standard deviation	Ratio
A
1	7	927	729	179	140	5.17
5	6	3963	1608	648	400	6.12
10	5	10752	3266	1897	1121	5.67
B
1	7	463	364	179	140	2.59
5	6	1982	804	648	400	3.06
10	5	5376	1633	1897	1121	2.83

The quantification was performed for batches of 1, 5 and 10 spores considering at least five biological replicates. The ratio is obtained by dividing the number of Mollicutes-related endobacteria for that of Candidatus Glomeribacter gigasporarum. We observe high variability in the quantification of 16S rRNA gene sequences of both types of endobacteria when single spores are analyzed, with this variation being reduced when batches of multiple spores are considered. This pattern is consistent with previous observations that CaGg abundance in individual spores can vary greatly (Jargeat et al., 2004). (A) The qPCR analysis revealed that Mre are 5.17–6.12 times more abundant than CaGg in the Gigaspora margarita CM23 spores, assuming that a single copy of the 16S rRNA gene is present in the Mre genome. (B) The value should be reduced to 2.59–3.06 times if two copies of the 16S rRNA gene are present in Mre genomes.

The qPCR analysis of the bacterial 16S rRNA gene sequences revealed that Mre are 5.17–6.12 times more abundant than CaGg in the G. margarita CM23 spores, assuming that a single copy of the 16S rRNA gene is present in the Mre genome. This value should be reduced to 2.59–3.06 times if two copies of the 16S rRNA are present in Mre genome instead (Table 4). This finding is consistent with our FISH observations, which suggested that Mre were more abundant than CaGg in G. margarita spores.

Discussion

A combination of morphological, molecular and phylogenetic analyses demonstrates that Gigaspora margarita spores host a complex microbiome consisting of rod-shaped and coccoid bacteria. The two bacterial groups are very distinct not only in their phylogenetic placement, that is, CaGg is closely related to Burkholderiaceae, whereas the coccoid endobacteria are related to the Gram-positive Mollicutes, but also in their genetic features.

Sharing the same host and revealing intra-host diversity

Notwithstanding the endobacteria share the same fungal host, a relevant difference in genetic diversity patterns between them was revealed. While CaGg shows a high level of intra-host sequence similarity, the Mre are characterized by high levels of intra-host sequence diversity. One of the underlying causes of differences in sequence evolution patterns between CaGg and Mre may be differences in their lifestyle. For example, in Mre, we found evidence of recombination, which was not apparent in CaGg. This finding was supported by some genomic features of CaGg genome: notwithstanding its high repetitive DNA (15%), CaGg contains a low number of active insertion sequences, which are considered important determinants for recombination (Ghignone et al., 2012). Indeed, a recent study of CaGg, using a set of four marker genes, revealed that recombination is not entirely absent from the CaGg evolutionary history and, together with host switching, may have an important role in evolutionary stability of CaGg association with Glomeromycota (Mondo et al., 2012). Detecting evidence of recombination in a single gene of Mre sampled in the present study may suggest that Mre engage in more frequent recombination than CaGg. Interestingly, cryptic prophage remnants have been detected in the genome of the Mre-related phytoplasma, leading to the suggestions that these genetic elements may have had important roles in generating phytoplasma genetic diversity (Wei et al., 2008).

Phylogenetic divergence patterns of the co-existing endobacteria

The extensive phylogenetic analyses performed on the endobacteria thriving in the cytoplasm of five spore samples and their comparison with data from previous investigations (Bianciotto et al., 1996, 2000, 2003; Mondo et al., 2012) confirmed that the 16S rRNA gene sequences of CaGg were relatively conserved, irrespective of the geographic origin of the fungal host. However, our careful analyses showed that the sequence similarity between CaGg from G. margarita MAFF520054 isolate and the already sequenced CaGg from G. margarita BEG34 was below the critical level of 97%. In fact, although this distinction is controversial (Rossello-Mora, 2003), it is generally accepted that sequences with similarity greater than 97% are typically assigned to the same species and those with similarity greater than 95% to the same genus (Stackebrandt and Goebel, 1994; Everett et al., 1999; Gevers et al., 2005). Consequently, further work is needed to resolve whether CaGg from G. margarita MAFF520054 and G. margarita BEG34, which show sequence similarity lower than 97% and a different location inside the CaGg phylogenetic tree, represent distinct taxa.

In contrast to CaGg and despite the stringent removal of chimeric sequences, the 16S rRNA gene sequences of Mre turned out to be highly variable inside at least four out of five spore samples. Moreover, in only 8% of the sequences generated in this study (4 out of 52), the similarity with sequences from GenBank was above 97% the remaining 92% of the sequences showed sequence similarity lower than 97%. Despite such high-sequence dissimilarity levels, all Mre sequences obtained in this study clustered together with the ones previously retrieved from Glomeromycota spore collection and liverworts-associated AMF. It is additionally possible that the stringent chimera removal excluded some non-chimeric sequences. However, this allowed us to enhance our phylogenetic resolution beyond what was presented in previous studies (Naumann et al., 2010; Desirò et al., 2013). As a result, we could recognize at least two distinct well-supported Mre clades, here identified as Mre group A and Mre group B. However, due to high level of sequence divergence between Mre sequences clustering in the same Mre group, we hypothesize that these newly described groups can mask other still hidden clades.

Genetic and lifestyle features of endobacteria are not affected by their co-occurrence

Our present study is the first one to describe in a single fungal host the coexistence of two distinct bacterial endosymbionts. Until now, these two symbionts have been studied in isolation from each other. We found that the morphological characteristics of the two co-existing bacterial endosymbionts did not differ from those described previously in the samples where only one bacterial symbiont was present. For example, even when sharing the same cell volume, CaGg remained enclosed in a vacuole-like structure, whereas Mre were embedded directly in the cytoplasm.

Interestingly, the spore samples that we investigated showed different patterns of intersymbiont dynamics. For example, in the sample CM50 a single Mre phylotype leading to a single OTU was detected together with the homogenous CaGg population. In contrast, in the remaining samples, Mre showed high levels of nucleotide diversity and sequence divergence. It would be useful to explore which of these two scenarios is more recent and which is more evolutionarily stable.

Irrespective of the dynamic levels of Mre sequence similarity in different samples, FISH and molecular quantitative analysis revealed that Mre were unambiguously more abundant than CaGg. The stronger presence of the Mre, together with their high variability, may indicate that they are stronger colonizers of AMF. On the basis of their 16S rDNA phylogeny, Mre have been described as related to Mollicutes (Naumann et al., 2010), a bacterial group that clusters with microbes (that is, Mycoplasma) thriving inside many eukaryotic hosts and manipulating host development, thanks to the release of effector proteins (Sugio et al., 2011). Due to their capacity to interact with many AMF host genotypes, we hypothesize that Mre have been one of the factors shaping AMF evolution and/or their ecological success.

Similarities between endosymbionts of insects and AMF

The wealth of natural history and molecular evolution data available for heritable endosymbionts of insects make them into an excellent model for understanding symbiotic associations that involve vertically transmitted endobacteria. In addition to essential endosymbionts, insects can support complex communities of bacteria that include non-essential endosymbionts as well as reproductive manipulators (Moran et al., 2008). Essential endosymbionts show strict vertical transmission and functional complementation with their hosts resulting from millions of years of reciprocal selection (McCutcheon and Moran, 2010). The genomes of essential endosymbionts are usually highly reduced (McCutcheon and Moran, 2010; McFall-Ngai et al., 2013). In this context, Buchnera aphidicola is a paradigm for primary endosymbionts. Buchnera's association with aphids is ancient, being approximately 200 million years old and revolves around the endosymbiont's capacity to synthesize essential amino acids for its host (van Ham et al., 2003). Due to their pleiotropic effects on their hosts, the situation is not so clear-cut for the non-essential (secondary) endosymbionts, as their transmission may be both vertical and horizontal and the ratio between cost and benefits strictly depends on environmental conditions (Ferrari and Vavre, 2011).

Given our observations that a single cell (a spore) of a fungus can host endosymbionts with distinct characteristics, it is worth considering whether the biological features of these fungal endobacteria are comparable to those of endosymbionts of insects.

In the case of CaGg, one of its hosts, Gigaspora margarita, can survive and multiply in the absence of the endobacterium (Lumini et al., 2007), and there are natural CaGg-free isolates of Gigasporaceae (Mondo et al., 2012), demonstrating that this symbiosis is facultative for the host. However, the fungal fitness can be strongly reduced by removal of the endobacteria (Lumini et al., 2007; P Bonfante and M Novero, personal communication, 2013). In addition, by using codiverging partner pairs, Mondo et al. (2012) demonstrated that this fungal/bacterial association is ancient (at least 400 million years old) and evolutionarily stable. Analysis of the 1.72 Mb CaGg genome (Ghignone et al., 2012) revealed that it is reduced when compared with the free-living related Burkholderia species, and that the metabolic profile of CaGg unambiguously clusters with insect endobacteria, including essential endosymbionts like Buchnera and Wigglesworthia (Moran et al., 2008). These data suggest that CaGg has undergone functional convergent evolution with phylogenetically distant endobacteria. However, genome annotation also shows functional similarities with the secondary non-essential symbionts (for example H. defensa). On the basis of these considerations, we concluded that CaGg is an obligate intracellular symbiont, characterized by a genetic mosaic where determinants for different nutritional strategies are integrated in a reduced genome (Ghignone et al., 2012). Collectively, its life history features (that is, a strict vertical transmission) as well as molecular evolution and genomic features seem to share patterns from both essential and non-essential endosymbionts of insects.

While the knowledge of the Mre biology is too limited to advance any hypothesis concerning their impact on the host biology, Mre relatedness to Mycoplasma and Phytoplasma, which are widespread parasites of animals and plants, might explain the colonization capacities of Mre, irrespectively of their role in the fungal hosts. On the other hand, it cannot be excluded that they are beneficial associates of fungi, akin to Spiroplasma endosymbionts that protect their insect hosts from the parasitoid pressure (Xie et al., 2010). Consequently, taking into consideration the limited available empirical evidence, we conclude that classifying Mre into categories established for bacterial associates of insects is not yet possible.

Are endobacteria favoured by coenocytic hyphae?

In the rapidly evolving taxonomic classification of Glomeromycota (Redecker et al., 2013), the taxon named Gigasporaceae identifies a group of AMF with distinct features of spore morphology (size, wall layering, bulbous base, germination shield) and host root colonization patterns (lack of intraradical vesicles and formation of auxiliary cells). In addition, this lineage of Glomeromycota turns out to be a preferential niche for endobacteria. Our present results confirm previous analyses (Bianciotto et al., 1996, 2000, 2003; Mondo et al., 2012) that demonstrated a strict association of CaGg with the Gigasporaceae. In contrast, Mre are widespread; they have been found in both basal and more recently evolved Glomeromycota taxa (Naumann et al., 2010). This differential distribution pattern is one of the key distinctions between the two groups of endosymbionts.

Our present results clearly demonstrate that Gigaspora margarita can harbor both endosymbionts, CaGg and Mre, and this is probably true also for other Gigasporaceae taxa (A Desirò and GA da Silva, personal communication, 2013). The underlying mechanisms responsible for the propensity of Gigasporaceae to host endobacteria are unknown. However, the genome sequence of the CaGg (Ghignone et al., 2012) shows that this bacterium is metabolically dependent on its fungal host. Perhaps only Gigasporaceae with their relatively large spores, which are rich in reserves of glycogen, fats and proteins (Bonfante et al., 1994), can support the energetic cost of complex bacterial communities, which thrive inside a protected niche.

There is, however, increasing evidence that Mortierella species (Mucoromycotina) host endobacteria that are related to CaGg (Sato et al., 2010; Kai et al., 2012; Bonito et al., 2013). These data open a novel interesting scenario: fungal endobacteria might prefer coenocytic hyphae. The absence of transverse septa may facilitate bacterial movement across the fungal mycelium, as observed in Rhizopus microsporus (Mucoromycotina) (Partida-Martinez and Hertweck, 2005). Recently, mitochondrial (Lee and Young, 2009; Pelin et al., 2012) and nuclear (Martin, 2012) genome analyses suggested that Mucoromycotina, instead of Dikarya, is the sister group of Glomeromycota. In this context, our data provide an additional evidence of the relationship between the two fungal lineages. The pattern of endosymbiont distribution across lineages of closely related fungal hosts raises questions about the role of symbiosis in the evolution and diversification of these fungal taxa and their associated endobacteria.

Conclusion

Our investigation has revealed for the first time that a single spore of an AMF can harbor multiple bacterial endosymbionts that represent phylogenetically diverse groups and show distinct patterns of sequence evolution. Both endosymbionts seem to retain their genetic and lifestyle peculiarities regardless of whether they colonize the host alone or together. Mre population consistently appears to be more abundant, variable and prone to recombination events than the CaGg one, suggesting that the same niche (the fungal spore) exerts a different selection pressure on its dwellers.

Our findings showing that a single fungal cell can harbor an intracellular bacterial microbiome, raise novel questions concerning molecular, cellular and metabolic interactions resulting from such complex inter-domain relationships.

The authors declare no conflict of interest.