RIBOSOMAL RNA GENE DIVERSITY, EFFECTIVE POPULATION SIZE, AND EVOLUTIONARY LONGEVITY IN ASEXUAL GLOMEROMYCOTA

rRNA GENE SAMPLING, DIVERSITY, AND EVOLUTIONARY HISTORY

We characterized rRNA gene diversity in individual spores (isolates) collected from nine fungal populations representing the Claroideoglomus lineage (Table 1) according to Daniels and Skipper (1982). Total DNA was amplified from individual spores with the illustraTM GenomiPhiTM Version 2 whole genome amplification kit (GE LifeSciences, Piscataway, NJ). GenomiPhiTM products were diluted 1:20 in ddH₂O, and 1 μL of this diluted product was used as template in 25 μL polymerase chain reactions (PCRs) consisting of 1.25 U PfuTurbo® HotStart polymerase (Agilent Technologies, Santa Clara, CA), the supplied buffer, 0.4 mM total dNTPs, and 0.2 μM each primer, ITS3 5′-GCATCGATGAAGAACGCAGC-3′ (White et al. 1990) and NDL22 5′-TGGTCCGTGTTTCAAGACG-3′ (van Tuinen et al. 1998) (Fig. 1). Cycling conditions were 95°C for 2 min followed by 13 cycles of 30 s at 95°C, 30 s at 58°C, and 1.5 min at 72°C. Primer and dNTP concentrations as well as cycling conditions (13 cycles) were designed to maximize recovery of low-frequency rRNA gene variants and minimize PCR artifacts (Cline et al. 1996). PCR products were cloned using the TOPO® TA kit for sequencing (Invitrogen, Carlsbad, CA). Clones were randomly picked for plasmid amplification with the illustraTM TempliPhiTM kit (GE LifeSciences), and 1.5 μL of the products used directly in 12 μL sequencing reactions using the BigDye® v3.1 cycle sequencing kit (Applied Biosystems, Foster City, CA). Sequencing reactions were analyzed at the Cornell University Life Sciences Core Laboratory Center on an ABI 3730xl sequencer. We sequenced at least 15 clones from each of 25 spores, resulting in 533 sequences (GenBank accessions HQ856918-HQ857097).

rRNA gene sequences were edited in Sequencher 4.9 (GeneCodes, Ann Arbor, MI) and aligned using MUSCLE (Edgar 2004). All polymorphisms were verified in the original chromatograms. We applied clone correction to sequences recovered from a single spore. Only sequences that differed from others by more than one change, which was attributed to PCR error, were retained in the final alignment. Nucleotide diversity, that is, the average number of nucleotide differences per site between two randomly chosen sequences, π, and insertion/deletion (indel) diversity, π(i), in rRNA gene sequences were estimated using DnaSPv5 (Librado and Rozas 2009). To retain phylogenetic information from indels in the rRNA genes, gaps in the alignment were coded using simple indel coding (Simmons and Ochoterena 2000) implemented by the IndelCoder module of SeqState (Müller 2005; Müller 2006). To determine the relationships of the rRNA gene sequences recovered from Claroideoglomus isolates, we reconstructed their phylogeny using both Bayesian and maximum likelihood (ML) approaches. Bayesian phylogeny reconstructions were carried out in MrBayes 3.1.2 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003) on the CIPRES portal (Miller et al. 2010) using a general time-reversible model with Γ-distributed rate heterogeneity across sites (GTR+Γ) (Tavaré 1986) for the nucleotide partition. We allowed MrBayes to estimate nucleotide frequencies and substitution rate parameters (Table S1). Rate variation across sites was set to equal and coding was set to variable for the binary (coded gap) partition (Ronquist and Huelsenbeck 2003). All likelihood settings were unlinked between the two partitions (Ronquist and Huelsenbeck 2003). The analysis included two runs of 10,000,000 generations with the burn-in of 2,500,000 generations. We used jModelTest 0.1 (Posada 2008) to determine the appropriate nucleotide substitution model for ML searches (GTR+Γ; Table S1). ML searches and 1000 bootstrap replicates were executed in RAxML 7.2.6 (Stamatakis 2006; Stamatakis et al. 2008).

We reconstructed phylogenetic relationship between the Claroideoglomus rRNA genes and those in other major clades of Glomeromycota using the 5′ end of the 28S rRNA gene (Fig. 1). All characters in the alignment that contained gaps in multiple sequences were removed. Reference taxa were from Stockinger et al. (2009) and Stockinger et al. (2010) (Table S2). Phylogenies were inferred using MrBayes 3.1.2 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003) in two runs of 6,000,000 generations with the burn-in of 1,500,000 generations, and RAxML 7.2.6 with 1000 bootstrap replicates (Stamatakis 2006; Stamatakis et al. 2008).

We examined the evolution rates of Claroideoglomus rRNA gene variants using the likelihood ratio (LR) relative rates test implemented in r8s version 1.71, rrlike (Sanderson 2003; O’Meara et al. 2006), and Tajima's 1D relative rates tests (Tajima 1993) in MEGA version 4.0 (Tamura et al. 2007). The rrlike algorithm tests the hypothesis that a model with one constant rate of sequence evolution for a clade, which descended from a given most recent common ancestor (MRCA), fits the data as well as a model in which each of the subclades (two or more) that descended from the MRCA have different but constant rates (Sanderson 2003). Tajima's 1D test explores the differences in the rates of evolution of two ingroup sequences by evaluating the number of substitutions at segregating sites between a reference outgroup sequence and the two ingroup sequences (Tajima 1993).

MUTATION ACCUMULATION AND rRNA THERMODYNAMIC STABILITY

Because rRNA secondary and tertiary structure is crucial for rRNA function, we assessed how mutations affected the thermodynamic stability of the structures formed by internal transcribed spacer 2 (ITS2; Fig. 1) and the 5.8S/5′-end of the 28S rRNA regions. We estimated Gibb's free energy change, ΔG, of structures using mfold 2.7 (Walter et al. 1994; Zuker 2003). ΔG is the change in free energy of a folded rRNA molecule compared with the free energy of an unfolded rRNA strand; more negative ΔG indicates a more stable structure. For a given RNA sequence, mfold uses thermodynamic models to identify the secondary structure(s) that minimize ΔG. To analyze folding conformations of the ITS2, sequences were trimmed at the base of the ITS2-proximal stem (van Nues et al. 1995) (Fig. S1) before folding. To analyze the 5.8S and 5′-end of the 28S regions, the 5.8S-ITS2–28S alignment was trimmed to contain only the 5.8S and 28S regions (Fig. S2). We aligned the sequences to Saccharomyces cerevisiae U53879, which allowed us to use the published yeast thermodynamic model (Fields and Gutell 1996) to guide the folding of representative Claroideoglomus sequences. Claroideoglomus sequences and the yeast sequence showed an average divergence, K, (Nei 1987) of 0.22 nucleotide substitutions per site. This level of nucleotide divergence enabled us to finely guide folding of much of the 5.8S and 5′∼340 bp of the 28S regions by dividing the regions predicted to fold into stem loops, folding individual sections, and adding sections back together in VARNA version 3–4 (Darty et al. 2009).

To determine the pattern of accumulation of specific types of mutations and evaluate the selection pressure against deleterious mutations, we identified mismatch, compensatory, and complementary mutations occurring in the stem (double-stranded) regions of rRNA secondary structures from the combination of rRNA secondary structure models and phylogenetic comparisons of the sequence data. The substitutions were mapped onto the secondary structure predictions to determine if they occurred in a stem region or within a loop. For substitutions that occurred in a stem, we compared the effect that the mutation had on the base pairing with its partner and determined whether there had been a mismatch mutation, a compensatory mutation, or a complementary mutation by comparisons with the ancestral Claroideoglomus sequence, which was inferred using the ML method of Yang et al. (1995) implemented in PAML 4.4 (Yang 2007).

Because rRNA gene sequences in the Claroideoglomus lineages existed in the form of two major variants, S and L (Fig. 1), we used a generalized linear mixed model approach to assess the effect of molecule type (S or L) on 28S rRNA secondary structure stability and accumulation of mutations. This allowed us to take into account any random effects on the characteristics of a sequence due to its presence in a particular spore (isolate). We used a normal distribution to model the effect of molecule type on secondary structure stability and a Poisson distribution to model the effect of molecule type on mutation accumulation; a Poisson distribution more accurately describes patterns of mutation accumulation in DNA sequences (Luria and Delbrück 1943). All statistical analyses were done in R 2.12.1 (http://www.r-project.org).

EXPRESSION OF rRNA GENE VARIANTS

Total RNA was extracted from C. etunicatum CA-OT135–4-2 using the RNaqueous Micro Kit (Ambion, Austin, TX). Claroideoglomus etunicatum CA-OT135–4-2 was a single spore in vitro culture established and maintained in a root organ culture according to the protocol of Pawlowska et al. (1999), but with a layer of cellophane (LabSource, Romeoville, IL) separating the roots and the fungus from the medium. This system generates large amounts of genetically homogeneous material (Pawlowska and Taylor 2004). Roots were removed before harvesting the fungus for total RNA extractions. Pre-rRNA transcripts were amplified using the MasterAmpTM RT-PCR Kit for High Sensitivity (Epicentre Biotechnologies, Madison, WI) using either primers ITS3 and NDL22 or ITS3 and 862r 5′-GACACATCGCCACATCGC-3′, a primer specific to S variant rRNA genes.

DISTRIBUTION OF rRNA GENE VARIANTS IN NUCLEI

Glomeromycota spores contain hundreds of nuclei. To assess the distribution of rRNA gene variants within and between nuclei we conducted fluorescence in situ hybridization (FISH) experiments in C. etunicatum CA-OT135–4-2. Details of these experimental procedures are presented in Text S1.

EFFECTIVE POPULATION SIZE IN C. ETUNICATUM

We measured nucleotide diversity, π, and the diversity from the number of segregating sites, θ, at seven anonymous noncoding loci, a1, a2, a3, a7, 4a, 4j, 5f, sampled from 20 isolates representing the Euro-American population of C. etunicatum (AZ201C-1, BR220–1, CA301–1, CA-GT24–1, CA-OT140–31, CL372–1, CU127–41, FL705A-4, IA205–1, KS887–1, MA104–1, MD127–2, MX916B-2, NE102–1, PL118B-1, SP108C-1, TN101B-1, TX104–1, VZ102–2, and WV579A-1). The loci and the isolates were characterized in detail in our earlier study (den Bakker et al. 2010). Briefly, to ascertain that markers are not variable within spore individuals, an important consideration given multinucleate structure of Glomeromycota spores, we cloned their PCR amplicons, sequenced, and analyzed clones from isolates representing several geographic locations (Table S1 in den Bakker et al. 2010). We found no evidence of intraindividual polymorphism. The Euro-American population was defined through an exact test for population differentiation (Raymond and Rousset 1995) involving pairwise comparisons between populations representing different continents (den Bakker et al. 2010). We rejected the null hypothesis that alleles at different loci in this population are in linkage equilibrium by calculating the classical index of association I_A (Maynard Smith et al. 1993) and the standardized index of association I^S_A (Haubold and Hudson 2000). Moreover, using the genetic algorithm recombination detection method (Kosakovsky Pond et al. 2006), which accounts for nucleotide substitution rate heterogeneity that could confound recombination signal, we found no evidence of recombination breakpoints in the alignment of concatenated marker sequences (den Bakker et al. 2010). Each of the Euro-American isolates sampled in the present study represented a subpopulation from a different geographic location (Table 1 in den Bakker et al. 2010). This scattered sampling approach was chosen to best approximate the assumptions of the coalescent model (Lessard and Wakeley 2004; Cutter 2006). DnaSPv5 (Librado and Rozas 2009) was used for diversity estimates, π and θ, and to conduct tests of neutrality, including Tajima's D (Tajima 1989), and Fu and Li's D* and F* tests (Fu and Li 1993) as well as the analysis of the distribution of pairwise genetic differences for concatenated sequence data (Rogers and Harpending 1992). The neutrality tests, in addition to departures from neutrality caused by selection, are expected to detect departures from demographic equilibrium caused by recent population expansion or decline (Simonsen et al. 1995). Episodes of population expansion or decline are also predicted to generate characteristic unimodal distribution of pairwise genetic differences (Rogers and Harpending 1992).

To measure the mutation rate per site per year in C. etunicatum, we estimated: (1) nucleotide divergence K (Nei 1987) at four anonymous noncoding loci (a1, a2, a3, 5f) shared between C. etunicatum and its sister C. claroideum/luteum, and (2) divergence times between species of Glomeromycota. (1) To measure K, we used DnaSPv5 (Librado and Rozas 2009), and included the Euro-American isolates of C. etunicatum together with isolates AU401A-1, KE118–13, MR102A-1, and NB119–17, characterized in den Bakker et al. (2010) as well as C. claroideum/luteum isolates MT109–2, SA112–1, and SW202–1 (GenBank accessions EF627536-EF627538, EF627634-EF627636, EF627722-EF627724, EF641304, and EF641307). (2) Divergence times were determined using BEAST version 1.6.1 (Drummond and Rambaut 2007) from published 18S rRNA gene sequences after aligning them in MUSCLE (Edgar 2004). We used LogCombiner version 1.6.1 to merge four independent MCMC runs of 10,000,000 steps with the burn-in of 1,000,000 steps each under the uncorrelated lognormal relaxed clock model (Drummond et al. 2006), the GTR+I+Γ nucleotide substitution model (Tavaré 1986), and with two calibration points.

As synonymous substitution rates at protein-coding loci are expected to approximate neutral mutation rates (Kasuga et al. 2002), we compared the mutation rate at noncoding anonymous loci of C. etunicatum with synonymous mutation rates among other species of Glomeromycota. We computed the numbers of synonymous and nonsynonymous sites in four published protein-coding gene sequences (alpha-tubulin, beta-tubulin, elongation factor 1-alpha, RNA polymerase II subunit RPB1) using the F3×4 model, and we estimated the synonymous and nonsynonymous substitution rates using the ML method (Goldman and Yang 1994) implemented in the codonml module of PAML version 4.4 (Yang 2007).

To estimate N_e, we sampled the posterior distribution of the coalescent parameter N_eτ in a demographic model of the Euro-American population, where τ is the generation length in units of time, using BEAST version 1.6.1 (Drummond and Rambaut 2007). With LogCombiner version 1.6.1 we merged four independent MCMC analyses of 10,000,000 steps with the burn-in of 1,000,000 steps each run under the GTR+I+Γ nucleotide substitution model (Tavaré 1986) and the strict molecular clock. We also obtained point estimates of N_e from π=θ= 2N_eμ, where μ is the mutation rate per site per generation.

PATTERNS OF rRNA GENE POLYMORPHISM AND DIVERGENCE IN CLAROIDEOGLOMUS

Although intraindividual variation in rRNA gene sequences has been reported in virtually all species of Glomeromycota, its extent has not been explored or catalogued. To understand the patterns of rRNA gene variation in Glomeromycota, we examined the rRNA gene repeat segment comprising the 5.8S rRNA gene, ITS2, and 5′-end of the 28S rRNA gene in the Claroideoglomus lineage of Glomeromycota (Fig. 1, Table 1). This fragment was PCR amplified from individual spores, cloned, and sequenced. We found 130 unique 5.8S-ITS2–28S rRNA gene sequences in 25 spores from nine globally distributed populations of C. etunicatum, C. claroideum, and C. luteum (Table 1, Fig. 2). Between 4 and 12 unique sequences were present in each individual spore. Each spore contained two distinct rRNA gene types, a long (L) type (∼1100 bp) and a short (S) type (∼1000 bp), which were primarily differentiated by a ∼100 bp indel at the 3′ end of ITS2 (Figs. 1 and 2). Phylogenetic reconstructions showed that the L and S sequences formed two distinct clades; each of these two clades contained sequences from all three species (Fig. 2B, C, Table S1). In both, the L and S clades, rRNA genes of C. etunicatum clustered away from C. claroideum and C. luteum. Sequences from C. claroideum and C. luteum could not be separated from each other with confidence and were therefore examined jointly.

Claroideoglomus etunicatum and C. claroideum/luteum rRNA genes show further differentiation and clustering within the L and S clades (Fig. 2B, C). For example, in C. etunicatum, the rRNA genes from the Australian population (AU401A), which is highly divergent from the remaining globally distributed populations (den Bakker et al. 2010), group away from all other C. etunicatum rRNA genes and are differentiated into five distinct subclades, Au-1, Au-2, Au-3, Au-4, and Au-5 (Fig. 2B, C). In contrast, the rRNA genes from the California (CA-OT135), Colombia (CL372), and Morocco (MR202A) populations cluster together while still differentiating into several subclades, including CaClMr-1, CaClMr-2, CaClMr-3, and CaClMr-4, with some of the Moroccan sequences (Mr-1) separating from the Californian and Colombian sequences (Fig. 2B, C). Clustering of sequences from the American and African isolates is unexpected because the American population is significantly differentiated from the African population (den Bakker et al. 2010). A similar pattern of multiple rRNA gene polymorphisms present in individual isolates and often shared by geographically distinct populations is also noticeable in the C. claroideum/luteum clade (Fig. 2B, C). Such sharing of polymorphisms by geographically distinct populations suggests that the divergent sequence variants are genetically linked.

To infer the history of rRNA genes in the Claroideoglomus lineage in the context of evolution of all Glomeromycota, we included in our phylogeny GenBank sequences representing intraisolate rRNA gene variation in other major Glomeromycota lineages (Stockinger et al. 2009; Stockinger et al. 2010). This reconstruction revealed that rRNA genes sampled from the Claroideoglomus lineage are specific to this lineage (Fig. 2A, Table S1). Such pattern of separation is consistent with the birth of the 5.8S-ITS2–28S rRNA gene length polymorphism in the MRCA of the Claroideoglomus lineage. We next tested whether the L and S 28S rRNA genes have evolved at similar rates since their divergence using LR (Sanderson 2003; O’Meara et al. 2006) and Tajima's 1D (Tajima 1993) relative rates tests. The LR test showed a significant difference in the evolution rate of the L and S clades (LR = 407.02, P < 0.001, df = 3). Representative Tajima's 1D tests corroborated this result (Table 2). In contrast, tests between sequences from the same clade were not significant (Table 2). Altogether, these results indicate that L and S gene clusters experienced distinct evolutionary histories characterized by different rates of nucleotide substitutions.

MUTATION ACCUMULATION PATTERNS IN rRNA GENES

To understand how mutations differentiating the L and S gene variants affect function of the rRNA molecules, we explored the patterns of mutation accumulation in different functional regions of these two sequence types using several approaches.

Levels of nucleotide conservation along the 28S rRNA gene sequence

Comparative studies of eukaryotic 28S rRNA gene sequences revealed that the level of sequence conservation varies along the molecule (Schnare et al. 1996). Regions that map to the ribosome functional sites display high degree of primary sequence and secondary structure conservation whereas regions that map to the surface of the ribosome show variability in both size and structure. In addition to divergent ITS2 regions, Claroideoglomus L and S variants are differentiated by multiple nucleotide polymorphisms and small indels along the 28S rRNA gene. To assess whether these polymorphisms are contained in specific variable regions or evenly distributed along the molecules, we computed nucleotide diversity, π, separately in L and S variants of C. etunicatum and C. claroideum/luteum using a sliding window approach (Fig. 3). We found that both L and S variants exhibit low π values in the same sequence regions. These regions are also among the most highly conserved regions across eukaryotes (Schnare et al. 1996). The π data suggest that functional constraints modulate mutation accumulation in both L and S variants, which, in turn, indicates that the S variants are not pseudogenes that accrue mutations in a random fashion.

EXPRESSION OF rRNA GENE VARIANTS

The patterns of mutation accumulation in the L and S variant rRNA genes throughout the Claroideoglomus lineage suggested that both variants are functional. To provide more evidence for this claim, we examined expression of the rRNA genes in the C. etunicatum isolate CA-OT135–4-2. Reverse transcriptase-PCR of both L and S variant pre-rRNA transcripts followed by cloning and sequencing of the products showed that both variants are transcribed and their transcripts are stable. We confirmed this result by targeted amplification of the S type transcripts using primers specific to S variant pre-rRNAs (Fig. S3).

DISTRIBUTION OF rRNA GENE VARIANTS IN NUCLEI

Our phylogenetic reconstructions and transcript analyses suggested that both the L and S variants existed in the genome in high copy number. To examine how these multiple copies are organized among and within nuclei, we performed FISH experiments using probes designed to differentiate between the L and the S 28S rRNA genes in C. etunicatum CA-OT135–4-2. Spores, which were crushed to release nuclei, and intact hyphal fragments were embedded in polyacrylamide pads to preserve three-dimensional structures of the nuclei. Both probes hybridized to chromatin in every examined nucleus (n = 310; Fig. 4). In spores, which contain spherical nuclei, the probes localized to a centrally located area of the nucleus (Fig. 4D–F). The probes also showed a pattern of distinct co-localization in hyphae nuclei, which assume various shapes (Fig. 4G–I). Three-dimensional image reconstructions revealed that the L and S variant signals were close to each other but were spatially separated (Movie S1). Overall, the FISH data suggest that the L and S rRNA gene arrays, while associated with the same nucleoli, have a disjunct organization. Such organization is likely to impair unequal crossing over and gene conversion, which are normally responsible for rRNA gene homogenization (Seperack et al. 1988).

EFFECTIVE POPULATION SIZE IN C. ETUNICATUM

Effective population size, N_e, influences the degree of identity among the rRNA gene copies (Ohta 2000) and reflects the efficacy of selection relative to drift (Charlesworth 2009). To understand forces that shape rRNA gene diversity in Glomeromycota, we estimated N_e in the Euro-American population of C. etunicatum (den Bakker et al. 2010). We measured π= 0.00192 and θ= 0.00235 in seven randomly selected anonymous noncoding single-copy genomic fragments of a total length of 2538 bp sampled from 20 isolates (Fig. 5, Table S1) characterized in detail in our earlier study (den Bakker et al. 2010). Tests of neutrality, including Tajima's D (Tajima 1989) as well as the Fu and Li's D* and F* (Fu and Li 1993), did not reveal any significant departures from the neutral equilibrium model (Table 3). Similarly, the distribution of observed pairwise differences did not show a signature wave pattern expected in a population with a history of expansion or decline (Rogers and Harpending 1992; Rosendahl et al. 2009). Instead, the empirical mismatch distribution was characterized by multiple peaks expected in equilibrium populations (Fig. 6).

Because the mutation rate in C. etunicatum was not known, we determined the neutral mutation rate by measuring the nucleotide divergence K between C. etunicatum and C. claroideum/luteum at four anonymous noncoding genomic loci that could be sampled from both taxa (Table 4), and by estimating the between-species divergence time from the 18S rRNA gene phylogeny (Fig. 7, Table S1). The 18S rRNA gene phylogeny was reconstructed under the uncorrelated lognormal clock model (Drummond et al. 2006) with two calibration points: the 396 ± 12 million year old fossils marking the Scutellospora radiation (Dotzler et al. 2006), and the divergence of moss and vascular plants 700 ± 35 million years ago marking the upper bound of the Glomeromycota radiation (Heckman et al. 2001). Given the average divergence K= 0.12351 (Table 4) and the divergence time of 25.9 million years between C. etunicatum and C. claroideum (Fig. 7), the average neutral mutation rate is 2.38 × 10⁻⁹± standard error (SE) of 0.85 × 10⁻⁹ per site per year. This value is comparable to the synonymous substitution rates estimated at four independent protein-coding loci (alpha-tubulin, beta-tubulin, elongation factor 1-alpha, RNA polymerase II subunit RPB1) sampled across Glomeromycota (Table S5). The synonymous substitution rates in Glomeromycota ranged from 0.24 × 10⁻⁹ to 2.34 × 10⁻⁹ per site per year and were similar to the ones measured in other fungi (0.9 × 10⁻⁹–16.7 × 10⁻⁹; Kasuga et al. 2002).

Given the neutral mutation rate of 2.38 × 10⁻⁹ per site per year and a generation time of 1 year, the coalescent demographic model of the Euro-American population of C. etunicatum constructed under the assumption of constant population size (Fig. 5) generated the mean value of N_e posterior distribution equal to 428,000 with the lower bound of the 95% highest posterior density (HPD) interval at 167,000 and the upper bound of the 95% HPD interval at 750,000. The diversity estimates π and θ (Table 3) yielded N_e(π) = 402,000 with a confidence interval of 296,000–625,000 based on SE of the neutral mutation rate, and N_e(θ) = 493,000 (363,000–767,000).

The N_e estimates suggest that, while smaller than populations of many fungi (Fisher 2007; Ellison et al. 2011), the C. etunicatum population is relatively large and similar in size to populations of Daphnia magna, a partially asexual crustacean with N_e of ∼500,000 (Haag et al. 2009), Mus musculus castaneus, a wild mouse with N_e of ∼580,000 (Halligan et al. 2010), and Capsella grandiflora, an obligately outcrossing cruciferous plant with N_e of ∼500,000 (Slotte et al. 2010). In contrast, populations of many sexual fungi can be considerably larger, reaching N_e of ∼10⁷ (Fisher 2007).

PATTERNS OF rRNA GENE EVOLUTION IN GLOMEROMYCOTA

Most eukaryotic species, other than interspecific hybrids (Hillis et al. 1991), harbor uniform sets of rRNA genes (Hillis and Dixon 1991). Glomeromycota are unique in that they exhibit extensive intraindividual rRNA gene sequence variation (Stockinger et al. 2010). Species in the Claroideoglomus lineage show three distinct patterns of rRNA gene variation: (1) ancestral polymorphism, (2) concerted evolution, and (3) imperfect homogenization of minor intraspecific polymorphisms. Our data indicate that several different processes are collectively responsible for generation of these patterns.

Imperfect homogenization

The existence of minor intraindividual/intraspecific rRNA gene polymorphisms suggests that rRNA gene homogenization is highly localized and limited to neighboring gene copies. This pattern can be best explained by the combination of a large N_e and a low rate of homogenization relative to the mutation rate (Ohta 2000) (Fig. 8). An alternative scenario of minor polymorphisms being distributed among different nuclei (Kuhn et al. 2001; Hijri and Sanders 2005) is unlikely because these polymorphisms are often shared by geographically differentiated populations (e.g., C. etunicatum CaClMr-1 and CaClMr-2 variants; Fig. 2), which is consistent with genetic linkage that can occur between gene copies but would not be expected between different nuclei. In the clonal population of C. etunicatum (den Bakker et al. 2010), the apparent absence of meiotic crossing over and chromosome reassortment eliminates the potential for population-level homogenization (Nagylaki 1984). If the rate of homogenization is not sufficient to outpace the rate of mutation, a pattern of neutral rRNA gene sequence polymorphisms is expected to emerge in a population due to genetic drift (Nagylaki 1984; Seperack et al. 1988), which is precisely what we observe in C. etunicatum (Fig. 8).

EFFECTIVE POPULATION SIZE IN C. ETUNICATUM IS A PRODUCT OF MULTILEVEL SELECTION

Our N_e estimates suggest that the C. etunicatum population is relatively large and similar in size to obligately outcrossing populations of the wild mouse M. m. castaneus (Halligan et al. 2010) and the crucifer C. grandiflora (Slotte et al. 2010). Both, M. m. castaneus and C. grandiflora show evidence of efficient positive and purifying selection (Halligan et al. 2010; Slotte et al. 2010). Given the clonal structure of the C. etunicatum population (den Bakker et al. 2010), N_e of such magnitude suggests that, in the absence of sexual recombination, this organism must have other mechanisms that reduce accumulation of slightly and moderately deleterious mutations (Fig. 8). For example, Glomeromycota possess a unique mode of asexual sporogenesis that could be responsible for the improved efficacy of selection against deleterious mutations in their asexual populations (Jany and Pawlowska 2010). Glomeromycota spores, unlike propagules of most other eukaryotes, harbor hundreds of nuclei. In other eukaryotes, propagules are uninucleate, which is believed to facilitate resolution of within-organismal conflicts generated by mutations that accumulate during a lifetime of a multicellular organism (Bell and Koufopanou 1991; Roze and Michod 2001). In Glomeromycota, live microscopy observations revealed that spore primorida are populated by influx of a stream of random nuclei from the surrounding aseptate mycelium (Jany and Pawlowska 2010). Moreover, the nuclei with damaged DNA are selectively eliminated from the mycelium by a regulated process consistent with programmed death in response to deleterious mutations. Quantification of the collective effects of these nuclear dynamics on the population mutation load revealed that for uniformly deleterious mutations, that is, mutations that reduce both the mutant nucleus replication rate and the organismal fitness, multinucleate propagules are expected to surpass uninucleate propagules in the ability to eliminate mutations (Jany and Pawlowska 2010). Consequently, we believe that the mode of sporogenesis combined with selection acting at the level of individual nuclei (Jany and Pawlowska 2010) may be a key factor increasing the efficacy of selection against deleterious mutations and contributing to the relatively large N_e of C. etunicatum (Fig. 8).

IS MULLER'S RATCHET A THREAT TO GLOMEROMYCOTA?

There are no known sexual taxa in Glomeromycota (Smith and Read 2008). Fossil records indicate that Glomeromycota existed as a diversified clade as early as 400 million years ago (Redecker et al. 2000; Dotzler et al. 2006; Dotzler et al. 2009). Therefore, it is unlikely that these organisms lost sexual reproduction during their recent evolutionary history and Muller's ratchet is only starting to operate in their populations. Instead, we believe that Glomeromycota have been asexual for an extended evolutionary time and their vulnerability to extinction is reduced thanks to relatively large effective population sizes.

Muller's ratchet is believed to be a problem even in moderately sized populations such as self-fertilizing hermaphrodite worms Caenorhabditis elegans, which have N_e of ∼80,000 and experience impaired selection against mildly deleterious mutations (Loewe and Cutter 2008). The magnitude of Muller's ratchet threat in the relatively large population of C. etunicatum is uncertain. Evidence of functional constrains in mutation accumulation in rRNA gene sequences suggests that C. etunicatum experiences rather effective elimination of deleterious mutations. Moreover, compensatory mutations, which restore fitness losses inflicted by past mutations, are apparent in C. etunicatum's rRNA gene sequences. This evidence of compensatory evolution is important for understanding the fate of AM fungal populations because both theoretical (Wagner and Gabriel 1990; Poon and Otto 2000) and empirical studies (Burch and Chao 1999; Estes and Lynch 2003; Estes et al. 2011) indicate that compensatory mutations can decelerate the advance of Muller's ratchet.

Like other filamentous fungi, Glomeromycota occasionally engage in vegetative hyphal fusions that can mediate gene exchanges (Croll et al. 2009). It is also possible that a yet undiscovered sexual process operates in Glomeromycota and is controlled by the meiotic genes present in these fungi (Halary et al. 2011). Nevertheless, with few exceptions (Vandenkoornhuyse et al. 2001; Croll and Sanders 2009), populations of Glomeromycota surveyed to date display significant linkage disequilibrium (Rosendahl and Taylor 1997; Stukenbrock and Rosendahl 2005; Rosendahl and Matzen 2008; Rosendahl et al. 2009; den Bakker et al. 2010). Evidence of linkage disequilibrium is not inconsistent with occurrence of rare recombination events in the population history (Maynard Smith et al. 1993). For example, the global population of C. etunicatum is overwhelmingly clonal although a pattern of homoplasy consistent with history of rare recombination is discernible in the phylogeny of the C. etunicatum lineage (den Bakker et al. 2010). Theory predicts that Muller's ratchet can be arrested by recreating mutation-free genotypes through recombination if the product of the recombination rate and population size achieves a numerical value of about 10⁶ (Bell 1988). Consequently, small populations require recombination rates associated with nearly obligate sexual reproduction to ensure their long-term survival whereas in large populations, rare recombination events may be sufficient to halt the ratchet (Bell 1988; Charlesworth et al. 1993). Given this prediction, rare cryptic recombination may contribute to arresting the ratchet in the sizable population of C. etunicatum.

Our model species, C. etunicatum, is likely not the only AM fungus with a relatively large N_e. A recent study in Funneliformis mosseae, another globally distributed AM fungus with a clonal population structure, revealed levels of diversity (Rosendahl et al. 2009) that are also consistent with a sizeable N_e.