Intercontinental divergence in the Populus-associated ectomycorrhizal fungus, Tricholoma populinum

Authors

  • Lisa C. Grubisha,

    1. Institute of Arctic Biology, 902 N. Koyukuk Drive, 311 Irving 1 Building, University of Alaska Fairbanks, Fairbanks, AK 99775-7000, USA
    2. Present address: Biology Program, Centre College, Danville, KY 40422, USA
    Search for more papers by this author
  • Nicholas Levsen,

    1. Department of Biological Sciences, Texas Tech University, Lubbock, TX 79409, USA
    Search for more papers by this author
  • Matthew S. Olson,

    1. Institute of Arctic Biology, 902 N. Koyukuk Drive, 311 Irving 1 Building, University of Alaska Fairbanks, Fairbanks, AK 99775-7000, USA
    2. Department of Biological Sciences, Texas Tech University, Lubbock, TX 79409, USA
    Search for more papers by this author
  • D. Lee Taylor

    1. Institute of Arctic Biology, 902 N. Koyukuk Drive, 311 Irving 1 Building, University of Alaska Fairbanks, Fairbanks, AK 99775-7000, USA
    Search for more papers by this author

Author for correspondence:
Lisa C. Grubisha
Tel: +1 859 238 5370
Email: lisa.grubisha@centre.edu

Summary

  • The ectomycorrhizal fungus Tricholoma populinum is host-specific with Populus species. T. populinum has wind-dispersed progagules and may be capable of long-distance dispersal. In this study, we tested the hypothesis of a panmictic population between Scandinavia and North America.
  • DNA sequences from five nuclear loci were used to assess phylogeographic structure and nucleotide divergence between continents.
  • Tricholoma populinum was composed of Scandinavian and North American lineages with complete absence of shared haplotypes and only one shared nucleotide mutation. Divergence of these lineages was estimated at approx. 1.7–1.0 million yr ago (Ma), which occurred after the estimated divergence of host species Populus tremula and Populus balsamifera/Populus trichocarpa at 5 Ma. Phylogeographic structure was not observed within Scandinavian or North American lineages of T. populinum.
  • Intercontinental divergence appears to have resulted from either allopatric isolation; a recent, rare long-distance dispersal founding event followed by genetic drift; or the response in an obligate mycorrhizal fungus with a narrow host range to contractions and expansion of host distribution during glacial and interglacial episodes within continents. Understanding present genetic variation in populations is important for predicting how obligate symbiotic fungi will adapt to present and future changing climatic conditions.

Introduction

Mycorrhizal symbioses are ubiquitous in nature, with symbiotic mycorrhizal plant species found in c. 90% of all plant families (Trappe, 1987). These symbioses date back over 400 million yr (Redecker et al., 2000), and are integral to ecosystem function, affecting both nutrient cycling and the establishment and growth of plants (Smith & Read, 2008). Despite the importance of the mycorrhizal symbiotic relationship, phylogeographic and taxonomic relationships between host and fungal symbionts remain understudied. The problem is particularly acute for the mycorrhizal fungal partner because morphological traits to discriminate species boundaries and population structure are often limited.

For many years, the ubiquitous dispersal hypothesis (Finlay, 2002) has been accepted by many researchers. This hypothesis contends that microbes are capable of long-distance dispersal owing to their small size and the large number of propagules (Finlay, 2002), and therefore population structure across vast geographic areas with similar habitat should be minimal. This principle is supported by global panmictic population structure in some fungi, including several human pathogens (Pringle et al., 2005; Rydholm et al., 2006) and root endophytes (Queloz et al., 2011). Other studies, however, have come to different conclusions, with several examples in the fungal literature of morphologically similar isolates traditionally placed in the same species that, in fact, have deep phylogenetic and phylogeographic structure when analyzed with molecular methods (Taylor et al., 2006; Douhan et al., 2011). Even at fine spatial scales, ectomycorrhizal fungi have been shown to have phylogenetic structure resulting in sympatric cryptic species as in Rhizopogon (Kretzer et al., 2003), Cenococcum (Douhan & Rizzo, 2005) and the Amanita muscaria complex (Geml et al., 2006, 2008).

Intercontinental geographic distribution for any species may result from vicariance or long-distance dispersal. Intercontinental vicariance is, by definition, ancient, but long-distance dispersal may have occurred at any point since continental separation. Thus, recent long-distance disperal and vicariance should have very different genetic signatures, while ancient long-distance dispersal may be difficult to distinguish from vicariance. In the case of mycorrhizal fungi that are obligate symbionts, suitable host species need to be present for intercontinental gene flow to be successful, a characteristic that is more constraining for fungal species with a narrow host range. Studies of some ectomycorrhizal fungi with widespread distributions have revealed distinct phylogenetic lineages with biogeographic and host boundaries, for example, Pisolithus (Martin et al., 2002), Auritella (Matheny & Bougher, 2006), A. muscaria (Geml et al., 2006, 2008) and Tylopilus ballouii (Halling et al., 2008). Although a large number of ectomycorrhizal fungi have Holarctic distributions (Tedersoo et al., 2010), there are too few studies of genetic variation from species with a Holarctic biogeographic range to elucidate general trends (Douhan et al., 2011).

Tricholoma populinum Lange (Basidiomycota, Agaricales) is a mushroom-forming ectomycorrhizal fungus that is host-specific with species in the genus Populus (Salicaceae), whereas Populus associate with many ectomycorrhizal fungi (Cripps & Miller, 1993; Krpata et al., 2008; Bahram et al., 2010). Populus (poplars, cottonwoods, aspens) is composed of c. 29 upland and riparian species with largely Northern Hemisphere distributions (Eckenwalder, 1996). Populus species have either a Palearctic or Nearctic distribution; however, no individual Populus species has a native Holarctic range (Dickmann & Kuzovkina, 2008). The European poplars comprise a polyphyletic group of three to four species (Hamzeh & Dayanandan, 2004; Cervera et al., 2005), whereas 12 species are found in North America (Eckenwalder, 1996). Except for Populus tremula L. (European aspen), which is sister to the North American Populus tremuloides Michx (quaking aspen), the European poplar species are most closely related to Asian species (Cervera et al., 2005). The most well-resolved phylogenetic reconstructions of the entire genus are based on amplified fragment length polymorphism (AFLP) and intersimple sequence repeat (ISSR) markers for which phylogenetic accuracy above the species-level is low (Simmons et al., 2007; García-Pereira et al., 2010) and models of evolution are not well developed (Ehrich et al., 2009). Divergence dates have been estimated for only the basal node (Tuskan et al., 2006), based on fossil data, and the very recent split (c. 75 000 yr ago) between Populus trichocarpa and Populus balsamifera in North America (Levsen et al., 2012). T. populinum co-occurs with P. tremula, Populus nigra, and Populus alba, in Europe and with P. balsamifera L. (balsam poplar) and P. trichocarpa Torr. & Gray (black cottonwood) in North America, but it is not known whether this species comprises a panmictic intercontinental fungal population or if there are separate fungal lineages that have diverged as a result of allopatric isolation or the evolution of host specificity. In this paper, we used analyses of five nuclear loci to test the hypothesis that T. populinum constitutes a single panmictic population through examination of genetic variation of T. populinum populations from Scandinavia and North America that are associated with different Populus host species, P. tremula in Scandinavia and P. balsamifera and P. trichocarpa in North America.

Materials and Methods

Study area and taxon sampling

Specimens used in this study are detailed in Supporting Information, Table S1, including name, collection location, host, and GenBank accession numbers for each locus sequenced. Collections were from herbaria or collected for this project. A total of 48 T. populinum collections were used in this study: 36 from North America and 12 from Norway and Sweden. In North America, T. populinum was collected associated with two hosts, P. balsamifera and P. trichocarpa. Host association could not be verified from herbaria collections from Colorado, Michigan and Ontario. In Norway and Sweden, the collections were associated with P. tremula. The maximum distance within North America between host populations of P. balsamifera (interior Alaska) and P. trichocarpa (Pacific Northwest, USA) was approx. 2500 km (Table S1, Fig. 1). The maximum distance between Scandinavian populations was approx. 400 km.

Figure 1.

Approximate collection locations for Tricholoma populinum isolates used in this study. The maximum distance of 2500 km from the Pacific Northwest (PNW) to Alaska Interior (AI) encompassed collecting sites from Populus balsamifera and Populus trichocarpa hosts. Sites from herbarium collections in Norway and Sweden represent a maximum distance of approx. 400 km from Uppsala, Sweden, to Oslo, Norway. AI, Alaska Interior, USA; AC, Alaska Coast, USA; BC, British Columbia, Canada; CC, Central Canada; PNW, Pacific Northwest, Canada and USA; CO, Colorado, USA; NM, Northern Michigan, USA; ON, Ontario, Canada; S, Sweden; N, Norway. Symbols indicate which Populus host species was present: closed circles, P. tremula; open circles, P. balsamifera; open squares, P. trichocarpa; closed squares, P. balsamifera may be found in these areas, but the exact Populus host species is not known.

DNA Isolation, PCR amplification, and sequencing

Dried fungal tissue from the pileus (mushroom cap) or stipe (stalk) was homogenized in a bead beater. DNA was isolated using either Qiagen DNeasy Plant Mini Kit (Qiagen) or E-Z 96® Fungal DNA Kit (Omega Bio-tek, Norcross, Georgia, USA) following the manufacturer’s instructions.

In this study six loci were sequenced that included one mitochondrial locus, the sixth subunit of the adenosine triphosphatase gene (ATP6); and five nuclear loci: the nuclear ribosomal internal transcribed spacer region (ITS) that included both ITS1 and ITS2 plus the 5.8S ribosomal subunit gene, the largest subunit of the RNA polymerase II binding protein (RPB1), the second largest subunit of the RNA polymerase II binding protein (RPB2), translation elongation factor 1-alpha gene (TEF1-α) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH).

Loci were PCR-amplified in a 20 μl final volume of the following mixture: 0.05 U μl−1 JumpStart REDTaq DNA polymerase (Sigma), 1× PCR buffer (100 mM Tris-HCL (pH 8.3 at 25ºC), 500 mM KCL, 15 mM MgCl2, 0.01% (w/v) gelatin), 0.3 μM of each primer, 2.5 mM MgCl2, 0.2 mM dNTPs, and 2.0 μl of 1 : 10 diluted genomic DNA. Initial PCR amplification used previously published primers and thermocycler protocols (Table 1). New primers were designed using Primer3 (Rozen & Skaletsky, 2000) for RPB1, RPB2, and GAPDH (Table 1). Thermocycler (PTC-220 thermocycler, Programable Thermal Controller) parameters for amplification of the ITS region were 95°C for 5°min, 35 cycles of 94°C for 30°s, 54°C for 40°s, and 72°C for 60°s, followed by a 10°min final extension at 72°C. PCR conditions for ATP6 followed Kretzer & Bruns (1999). Using the new primers (Table 1), PCR amplification was performed using a touchdown program: 95°C for 5°min, 10 cycles of 94°C for 30°s, 68°C for 30 s, −1.0°C per cycle, 72°C for 1.5 min, then 29 cycles of 94°C for 30 s, 58°C for 30°s, 72°C for 1.5 min, and a final extension of 72°C for 10 min for GAPDH. The touchdown annealing temperatures were decreased to 65 and 55°C for the new sets of primers for both RPB1 and RPB2. PCR amplicons were electrophoresed in a 1.5% agarose gel, and visualized by staining with ethidium bromide.

Table 1.   Primers used in this study
LocusPrimer nameNucleotide sequence (5′–3′)Reference
  1. ITS, nuclear ribosomal internal transcribed spacer region (ITS1 and ITS2 plus the 5.8S ribosomal subunit gene); TEF1-α, translation elongation factor 1-alpha gene; ATP6, the sixth subunit of the adenosine triphosphatase gene; RPB1, the largest subunit of the RNA polymerase II binding protein; RPB2, the second largest subunit of the RNA polymerase II binding protein; GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

ITSITS1FCTTGGTCATTTAGAGGAAGTAAGardes & Bruns (1993)
ITSITS4TCCTCCGCTTATTGATATGCWhite et al. (1990)
TEF1-αEF1-983FGCYCCYGGHCAYCGTGAYTTYATRehner & Buckley (2005)
TEF1-αEF1-1953RCCRGCRACRGTRTGTCTCAThttp://aftol1.biology.duke.edu/pub/primers/viewPrimers
ATP6ATP6-3fTCTCCTTTAGAACAATTTGAKretzer & Bruns (1999)
ATP6ATP6-6rAACTAATARAGGAACTAAAGCTAM. Binder (unpublished; http://www.clarku.edu/faculty/dhibbettProtocols_Folder/Primers/Primers.htm)
RPB1gRBP1-A forwardGA(G/T)TGTCC(T/G)GG(A/T)CATTTTGGStiller & Hall (1997)
RPB1aRPB1-B reverseTCCGC(A/G)CC(C/T)TCTTC(C/T)TTGGMatheny et al. (2002)
RPB1RPB1tp1fCCATCCTGGTACGCGCATTTCThis study
RPB1RPB1tp1rTCC GCG CCT TCT TCT TTG GThis study
RPB2bRPB2-6fTGGGGYATGGTNTGYCCYGCMatheny (2005)
RPB2bRPB2-7.1RCCCATRGCYTGYTTMCCCATDGCMatheny (2005)
RPB2RPB2tp1fAGGTCAAGCGTGCGGTCTCThis study
RPB2RPB2tp1rTCGAGGAGACTATTGGGAGATGThis study
GAPDHGPD1FCGGCCGTATCGTCCTCCGTAATGCden Bakker et al. (2004)
GAPDHGPD2RGAGTA(AT)CC(GC)CATTCGTTATCGTACCden Bakker et al. (2004)
GAPDHGPD1tp1fATGCCTACAGCCCTTTCATCThis study
GAPDHGPD1tp1rGCACAACCAAGTCCACGAThis study

Polymerase chain reaction amplification products were cleaned using ExoSAP-IT (US Biochemical Corp., Cleveland, OH, USA). Cycle-sequencing of purified PCR products was performed using the same primers as for PCR amplification and Applied Biosystems (ABI, Carlsbad, California, USA) BigDye® version 3.1 Terminator Kit. Samples were run on an ABI 3130 automated capillary DNA sequencer at the Core Facility for Nucleic Acid Analysis in the Institute of Arctic Biology, University of Alaska Fairbanks. ITS sequences from 19 North American isolates used in this study were generated by the MIT Broad Institute as part of a large-scale fungal biodiversity project of D. Lee Taylor. Sequences were assembled into contigs and edited using the program codoncode aligner version 1.52 (CodonCode Inc., Dedham, MA, USA). Coding regions and introns were identified using the web server AUGUSTUS version 2.4 (Stanke et al., 2008). Sequences were deposited in GenBank as accessions JN019383JN019533 and JN019584JN019753 (Table 1). Only one haplotype was identified for ATP6 for all T. populinum isolates and thus this locus was not included in further analyses.

Haplotype reconstruction

Haplotypes of heterozygous sequences were determined using three methods. Analysis of chromatograms revealed heterozygous multiple base insertion/deletions (indel) or more than one heterozygous indel in three ITS sequences. PCR products for these ITS sequences were cloned using the TOPO-TA cloning kit (Invitrogen) following manufacturer’s instructions and sequenced. For other loci with single heterozygous indel sites, chromatograms were of high quality with clear single peaks until the point of the indel and then double peaks were present for the remainder of the sequence. These sequences were subtracted and a gap added to the appropriate sequence, resulting in two haplotypes following Peters et al. (2007). Finally, for heterozygous sequences with single nucleotide substitutions, haplotypes were reconstructed from analyses in a Bayesian framework using the program phase version 2.1.1 (Stephens et al., 2001; Stephens & Donnelly, 2003). To check for consistency of results, haplotype frequency estimates and goodness-of-fit measures from five independent runs were compared using the parameters: 10 000 iterations, 100 thinning intervals, and a burn-in of 10 000 generations. Haplotype pairs with a posterior probability of ≥ 0.90 were accepted and used for haplotype analyses.

DNA polymorphism and nucleotide divergence

Estimates of nucleotide polymorphism were conducted using DnaSP version 5.10 (Librado & Rozas, 2009). Alignment gaps were removed for all analyses. The number of haplotypes, h, segregating sites, S, Watterson’s θ (an estimate of the population mutation rate, θ = 4Neμ, where Ne is the effective population size and μ is the mutation rate per generation (Watterson, 1975)), and average nucleotide diversity, π (Nei, 1987), were calculated to determine the degrees of DNA polymorphism in Scandinavia vs North America and across hosts.

Estimates of nucleotide divergence were conducted using DnaSP (Librado & Rozas, 2009). Nucleotide divergence between populations based on hosts and continents was estimated as the number of fixed nucleotide differences between populations, the number of polymorphic mutations that occur in one population but are monomorphic in the second, the number of shared mutations, the average number of nucleotide differences between populations, K, nucleotide divergence (average number of nucleotide substitutions per site between populations), Dxy, and the number of net nucleotide substitutions per site between populations, Da.

Population differentiation

Population structure was identified using the Bayesian clustering method implemented in Structure v2.3.3 (Pritchard et al., 2000; Hubisz et al., 2009). Haplotypes were allowed to have mixed ancestry in each of the K clusters (admixture model; Pritchard et al., 2000; Hubisz et al., 2009). To assess within continent genetic structure, a prior was set for each individual’s sampling location (locprior model) that detects weak population structure (Hubisz et al., 2009). Markov chain Monte Carlo (MCMC) simulations were run for = 1 to = 10 genetic partitions, with 15 replicate runs for each K. For each run a burn-in of 20 000 MCMC iterations was followed by 80 000 MCMC iterations. The number of genetic partitions was determined by estimating the ad-hoc statistic ΔK, which is the largest second-order rate of change in the likelihood distribution between successive runs of K (Evanno et al., 2005), using Structure harvester v0.6 (available at http://taylor0.biology.ucla.edu/struct_harvest/).

Phylogenetic analysis

FASTA files of nonphased sequences were compiled into multiple sequence alignments using the program Opal of the Opalescent version 1.0 (Wheeler & Kececioglu, 2007) package within Mesquite version 2.73 (Maddison & Maddison, 2010) and manually edited. For ITS sequences that had two alleles as a result of indels, a consensus sequence was determined for each isolate by deleting indels that represented intra-isolate allelic polymorphism. GenBank sequences of Tricholoma portentosum and Tricholoma myomyces were added as outgroups for TEF1-α (EF421084, DQ367429), ITS (AF357015, DQ825428), RPB1 (EF421047, DQ842013) and RPB2 (EF421014, DQ367436). Tricholoma scalpturatum (GU060260) and Tricholoma argyraceum (GU060239) were outgroups for GAPDH. Alignments were deposited on TreeBASE (Accession 12241). Bayesian analyses were conducted to estimate gene genealogies using MrBayes v. 3.1.2 (Huelsenbeck & Ronquist, 2001; Ronquist & Huelsenbeck, 2003). Nucleotide substitution models for each locus were estimated by employing the Akaike Information Criterion (AIC; Akaike, 1974) using MrModeltest v. 2.3 (Nylander, 2004) and PAUP* (Swofford, 2003). The GTR model of DNA substitution + gamma (Γ) was employed for TEF1-α, RPB2, ITS, GAPDH and GTR + invariant (I) for RPB1. Gamma distribution was approximated using four rate categories. Gaps were treated as missing data. Two runs of 10 000 000 generations were performed, sampling every 1000 generations and saving 10 000 trees. The first 2500 genealogies were discarded as burn-in. To determine if a stationary phase was reached at this burn-in, scatterplots were generated. Convergence for each run was also assessed by plotting the log-likelihood values against generation time using Tracer 1.5 (available at: http://tree.bio.ed.ac.uk/software/tracer). Posterior probabilities were determined from the remaining 7500 trees and a 50% majority rule consensus tree was computed and viewed in FigTree ver.1.3.1 (http://tree.bio.ed.ac.uk/software/figtree/).

Divergence time estimates for Tricholoma

A Bayesian MCMC method for multispecies coalescent analyses was implemented in *Beast ver. 1.6.1 (Drummond et al., 2006; Drummond & Rambaut, 2007; Heled & Drummond, 2010) for T. populinum North America associated with P. balsamifera and T. populinum Scandinavia associated with P. tremula. One advantage of using *Beast is that the standard error decreases with increased sample size for the species analyzed (Heled & Drummond, 2010). We had 17 T. populinum isolates from the North American P. balsamifera host and 12 T. populinum isolates from the Scandinavian P. tremula host. The three isolates that formed a monophyletic clade in all single gene phylogenies (O-F170922, O-F220341, AT2005140) were included as the outgroup. Closely related Tricholoma species included in the single gene genealogies, for example, T. aurantium, T. spp. DAVFP26750, were not included in *Beast analyses because the minimum requirement for estimation of population size is at least two isolates per species (Heled & Drummond, 2010). An assumption of *Beast is that sequences are free of recombination. The largest nonrecombining block of DNA sequence for each locus was identified using the online server IMgc (Woerner et al., 2007).

Nexus files for each alignment of phased, recombination-free haplotypes of five loci (TEF1-α, GAPDH, RPB1, RPB2 and ITS) were imported separately into BEAUti v1.6.1. Alignment gaps were deleted. Loci were unlinked for nucleotide substitution model, molecular clock model and tree model. The Hasegawa, Kishino and Yano (HKY) model of nucleotide substitution was used for each locus. The molecular clock hypothesis was tested for each locus and could not be rejected using the relative rate test of Tajima (1993) employed in Mega v. 4.0 (Tamura et al., 2007). Therefore, a strict molecular clock was set for each locus. A substitution rate was set for the ITS of 1.4 × 10−9 substitutions per site yr–1 (Kasuga et al., 2002) and the relative rates for the other loci were estimated in Beast. This method was chosen instead of fixing all loci to the same rate because that method has been shown to substantially decrease the estimate of species divergence (Heled & Drummond, 2010). Analyses were also conducted using the lower and upper ranges for the estimated substitution rates of 0.1 × 10−9 and 2.7 × 10−9 for ITS (Kasuga et al., 2002). The tree model parameters were a random starting tree for each locus, with the species tree prior set to a Yule process. BEAUti v1.6.1 was used to create Beast XML files. Short preliminary runs were performed to check operators for optimal performance that were adjusted in the final runs. To assess consistency across runs, two independent runs were performed, each with 20 000 000 generations, sampling every 2000 generations. Log files for each run were viewed in Tracer ver. 1.5 to verify that effective sample size (ESS) values were > 200 and that parameters had reached a stationary stage after a 10% burn-in. The two species tree files (10 000 trees each) were combined using LogCombiner ver. 1.6.1 after discarding a burn-in of 2500 trees from each run. The LogCombined tree file of 15 000 trees was imported into TreeAnnotator ver.1.6.1. A summary tree was produced and viewed in FigTree.

A second method of estimating divergence time employed a calibration based on divergence of Ustilaginomycotina and Agaricomycotina using Beast v. 1.6.1 (Drummond et al., 2006; Drummond & Rambaut, 2007). Because of the lack of fossil evidence for Tricholoma, a prior for the time to the most recent common ancestor (tMRCA) with a normal distribution was set between Ustilago (Ustilaginomycotina) and Agaricomycotina at 430 (± 50) million yr ago (Ma) based on Berbee & Taylor (2001). Nexus files of sequence alignments of RPB2 (707 bp) and the 5′ end of the nuclear large subunit ribosomal RNA gene (nLSU; 513 bp) for a 12 taxon data set were imported into BEAUTi ver. 1.6.1 to produce XML files for input in BEAST. The taxa chosen followed Geml et al. (2004) and Matheny et al. (2009) and are listed here including GenBank accession numbers for nLSU and RPB2, respectively: Endocronartium harnessii (AY700193, DQ234551), Ustilago tritici (DQ094784, DQ846896), Auricularia sp. (AY634277, DQ366278), Hygrophoropsis aurantiaca (AY684156, AY786059), Marasmius alliaceus (AY635776, AY786060), Laccaria ochropurpurea (AY700200, DQ472731), Alnicola escharoides (AY380405, AY337411), T. portentosum (U76464, EF421014), Tricholoma myomyces (U76459, DQ367436), Tricholoma spp. AT2005140 (JN019647, JN019704), T. populinum O-F70009 (Norway) (JN019648, JN019710), T. populinumLCG2003 (Alaska) (JN019649, JN019720). Nucleotide substitution model and molecular clock model were unlinked for the two loci. The GTR + Γ nucleotide substitution model with four rate categories was employed for RPB2 and the Tn93 + Γ model for nLSU as estimated by jModeltest v. 0.1.1 (Guindon & Gascuel, 2003; Posada, 2008). An uncorrelated lognormal relaxed molecular clock (Drummond et al., 2006) was employed with the tree prior set to a Yule process using a randomly generated starting tree. Prior constraints on monophyletic clades were based on previously published studies (Matheny et al., 2006, 2009). Tricholoma was constrained to be a monophyletic group, with no further constraints within the genus except that T. populinum Norway and Alaska were constrained to be monophyletic. The analysis was run for 10 000 000 generations, sampling every 1000 trees for each of two replicates to assess consistency among runs. Results were assessed as described earlier.

Divergence time estimates for Populus

*Beast analyses of Populus host species included 48 phased haplotypes (P. balsamifera, 22; P. trichocarpa, 20; P. tremula, six) of five nuclear loci (195487, 230673, 560714, 751129, and 816152; Levsen et al., 2012). All loci adhered to molecular clock assumptions and were set as unlinked for nucleotide substitution model, molecular clock model, and tree model. Nucleotide substitution models for each locus were determined using jModeltest v. 0.1.1, to be as follows: 195487, HKY; 230673, GTR + I; 560714, HKY; 751129, GTR + I; 816152, GTR. The strict molecular clock model was employed for each locus with mean rate of 2.5 × 10−9 substitutions per site yr–1 (Tuskan et al., 2006). Each locus had a random starting tree generated under a Yule process. We performed two independent *Beast runs of 100 000 000 generations each with a 10 000 generation sampling interval. Gene sequences for P. trichocarpa, P. balsamifera, and P. tremula are available on GenBank from previous studies (Keller et al., 2010; Levsen et al., 2012).

Results

Identification of isolates

Six herbarium collections of T. populinum were misidentified. The results of Blast searches of the ITS sequences for these collections is presented in Table S2.

Haplotype reconstruction

Haplotypes determined by Phase for T. populinum sequences from TEF1-α, ITS, RPB1, and RPB2 all had posterior probabilities (PPs) > 0.90. Eight GAPDH sequences from North American collections (LCG2191, LCG2213, LCG2246, LCG2088, LCG2074, LCG2050, LCG2056, LCG2065) had PPs < 0.90 and were not included in analyses requiring phased haplotypes. PCR errors were not observed in the three ITS sequences that were cloned.

Haplotype and nucleotide diversity and divergence

Aligned Tpopulinum sequence lengths, with gaps, for loci GAPDH, RPB1, RPB2 and TEF1-α were 829, 715, 628, and 800 bp, respectively. The ITS alignment consisted of 588 bp when indels that represented intra-isolate allelic polymorphism were included as phased haplotypes. Colorado isolate LCG2307 had one allele with a single 10 bp deletion in the ITS2 region but was otherwise identical to the second allele. Intra-isolate indel ITS variation was not present in the Scandinavian isolates.

Tricholoma populinum haplotypic diversity was similar across populations and hosts (average Scandinavia Hd = 0.487; average North American Hd = 0.479 Table 2). Haplotypes were not shared between Scandinavia and North America; in contrast, haplotypes were shared among North American populations for all loci. Average nucleotide diversity (π) varied from a high of 0.00254 in the GAPDH locus in T. populinum North America population to a low of 0.00014 in ITS of the T. populinum host P.trichocarpa population (Table 2).

Table 2.   Nucleotide polymorphism for populations of Tricholoma populinum identified by host and geographic location
HostLocation and locusnhSHdθW/siteπ
  1. aFor analyses of the North American host P. balsamifera, P. trichocarpa and Populus spp. category, isolate TRTC150613 and isolates from Colorado from the unverified Populus host were included.

  2. n, number of phased haploid sequences; h, number of haplotypes; S, number of segregating sites; Hd, haplotypic diversity; θW, Watterson’s theta; π, average pairwise nucleotide diversity.

Populus tremulaNorway/Sweden
GAPDH24320.4240.000650.00055
RPB124430.6490.001120.00109
RPB224760.8510.002560.0037
TEF118210.2090.000360.00026
ITS24320.3010.000920.00053
Mean   0.4870.0010.001226
Populus balsamifera, Populus trichocarpa, Populus spp.aNorth America
GAPDH5413110.8550.002910.00254
RPB170680.3910.002320.00193
RPB270670.4890.002310.00179
TEF1-α64640.520.001060.00105
ITS70430.1380.001080.00025
Mean   0.4790.0020.002
Populus balsamiferaNorth America
GAPDH20870.8470.002380.00235
RPB134480.3650.002740.00182
RPB234460.2730.002340.00133
TEF1-α34430.5440.000920.0011
ITS34320.1690.000840.0003
Mean   0.4400.0020.001
Populus trichocarpaNorth America
GAPDH22680.8310.002650.00246
RPB124360.3040.002250.00134
RPB224440.6090.001710.00198
TEF1-α24430.3080.0010.00059
ITS24210.0830.000460.00014
Mean   0.4270.0020.001

In all six T. populinum nuclear loci, only one nucleotide position in RPB1 was polymorphic in both Scandinavian and North American regions (Table 3). The remaining polymorphic sites were either fixed for differences between Scandinavia and North America, or polymorphic in one region and monomorphic in the other (Table 3). By contrast, comparisons of North American T. populinum populations collected near different host species shared polymorphic sites (range of one to six shared mutations) and no fixed differences (Table 3). Higher pairwise nucleotide divergence was found between Scandinavian T. populinum populations and each T. populinum population within Populus hosts in North America (Scandinavian to North American average = 8.1888, average Dxy = 0.01152, average Da = 0.01016; Table 3) than among North American T. populinum populations with two Populus hosts for K (average = 1.0170), Dxy (average Dxy = 0.001 38), and Da (average Da = 0.000 04) (Table 3).

Table 3.   Nucleotide divergence between Tricholoma populinum populations from Scandinavia and North America regions and by host
Region or host and locusNo. of polymorphic sitesNo. of fixed differencesP1 (continent or host)P2 (continent or host)Shared mutationsKDxyDa
  1. aFor analyses of the North American region, isolate TRTC150613 and isolates from Colorado from the unverified Populus host were included.

  2. P1, mutations polymorphic in population 1, but monomorphic in population 2; P2, mutations polymorphic in population 2, but monomorphic in population 1; K, average number of nucleotide differences between populations; Dxy, nucleotide divergence (average number of nucleotide substitutions per site between populations); Da, number of net nucleotide substitutions per site between populations.

Scandinavia vs North Americaa  ScandinaviaNorth America    
GAPDH21821109.954000.012010.01046
RPB11112714.294000.006010.00449
RPB21306704.946000.007880.00513
ITS1281309.139000.015950.01563
TEF1-α171214012.611000.015760.01511
Mean     8.188800.011520.01016
Populus tremula vs Populus balsamifera  P. tremulaP. balsamifera    
GAPDH1782709.850.011880.01043
RPB11112714.245000.005940.00448
RPB21206604.846000.007720.00520
ITS1282209.196000.015770.01536
TEF1-α161213012.670000.015840.01516
Mean     8.161400.011430.01013
P. tremula vs Populus trichocarpa  P. tremulaP. trichocarpa    
GAPDH1882809.886000.01193930.01042
RPB11022514.396000.006150.00493
RPB21226405.042000.008030.00519
ITS1182109.125000.015650.01531
TEF1-α161213012.361000.015450.01503
Mean     8.162000.011440.01018
P. balsamifera vs P. trichocarpa  P. balsamiferaP. trichocarpa    
GAPDH901262.000000.002410.00001
RPB1802061.108000.00155−0.00003
RPB2703131.1420.001820.00016
ITS201010.127000.000220.00000
TEF1-α401120.708000.000890.00004
Mean     1.017000.001380.00004

Population structure

The population structure of T. populinum within and among Scandinavian and North American regions and between Populus host species was assessed with the program Structure (Pritchard et al., 2000). Two clusters were identified (= 2) based on the ad hoc statistic ΔK of Evanno et al. (2005)K = 326) that align perfectly with Scandinavian and North American regions and Populus host species (P. tremula vs P. trichocarpa/P. balsamifera), with no admixture (Fig. S1). Structure results did not reveal genetic clusters within regions or hosts within North America.

Phylogenetic analyses

Bayesian inference of the T. populinum phylogenetic tree using our five nuclear loci provided strong support (PP = 1.0, four phylogenies) for monophyly of Scandinavian and North American populations of T. populinum (Fig. 2). Isolates O-F170922, O-F220341 and AT2005140 formed a strongly supported clade (PP = 1.0, four phylogenies) outside of T. populinum for each gene phylogeny. Isolate DAVFP26750 was the sister taxon to T. populinum in all gene phylogenies.

Figure 2.

Bayesian consensus single gene phylogenies of Tricholoma populinum collections from North America and Scandinavia. (a) Phylogram based on RPB1 locus (54 taxa, 742 characters). The DNA substitution model was GTR + invariant (I). (b) Phylogram based on RPB2 locus (54 taxa, 628 characters). The DNA substitution model was GTR + gamma (Γ). (c) Phylogram based on internal transcribed spacer (ITS) sequences (55 taxa, 604 characters). The DNA substitution model was GTR + Γ. (d) Phylogram based on GAPDH locus (54 taxa, 885 characters). The DNA substitution model was GTR + Γ. (e) Phylogram based on TEF1-α locus (51 taxa, 801 characters). The DNA substitution model was GTR + Γ. Posterior probabilities > 0.95 are shown. Tricholoma myomyces and Tricholoma portentosum were used as outgroups. Closely related herbarium collections that were identified as T. populinum were included in order to assess their phylogenetic relationship across several loci (DAVFP26750, OF170922, OF220341, AT2005140). Tricholoma aurantium LCG2307 is a closely related species.

Divergence time estimates for Tricholoma

The largest recombination-free sequence blocks were used in *Beast analyses: TEF1-α was truncated to 1–675 bp, RPB1 to 1–589 bp, RPB2 1–561 bp with deletion of phased haplotypes O-F70600a, O-F70600b, O-F51644b, O-F51642b, LCG2088b, LCG2074b, and for GAPDH the entire sequence was used after phased haplotypes AI02a and PC05b were removed. Recombination was not detected in ITS sequences. *Beast analyses resulted in a PP of 1.0 at both nodes and recovered T. populinum North America and Scandinavia as sister lineages (Fig. S2). The mean estimate of the divergence time between Scandinavia (P. tremula host) and North America (P. balsamifera host) was 1.7 Ma (95% highest posterior density (HPD): 0.76, 2.95 Ma). Estimates based initially on the upper and lower estimates of mutation rate for ITS from Kasuga et al. (2002) resulted in 0.87 Ma (95% HPD: 0.35, 1.46) and 23.55 Ma (95% HPD: 9.77, 41.98), respectively.

An estimate of divergence using a calibration point of divergence of the Ustilaginomycotina and Agaricomycotina resulted in an estimated divergence between North American and Scandinavian T. populinum at approx. 1.0 Ma (95% HPD: 0.17, 2.40) (Fig. 3). All nodes show evidence of strong support based on a PP of 0.98–1.0 (Fig. 3).

Figure 3.

Chronogram of select Basidiomycota based on BEAST analyses using a relaxed molecular clock. Numbers above branches are the estimated mean divergence dates in millions of years (Ma) based on a 430 ± 50 Ma calibration for the divergence of the Ustilaginomycotina and the Agaricomycotina, designated by the black arrow. Numbers in brackets at each node are the 95% highest posterior density (HPD), which encompasses 95% of the mean estimates for divergence time at a particular node from 15 000 trees. Posterior probabilities (PPs) were all = 1.0, except one had PP = 0.99 (**Agaricaceae/Tricholomataceae) and one had PP = 0.98 (*Tricholoma myomyces/Tricholoma spp. and T. populinum).

Divergence time estimates for Populus

Populus tremula and P. balsamifera/P. trichocarpa clades were recovered in *Beast with a PP of 1.0 (Fig. S3). The mean estimate of divergence time between P. tremula and P. balsamifera/P. trichocarpa was 5.04 Ma (95% HPD: 3.24–6.84 Ma).

Discussion

Intercontinental lineage divergence

Our results strongly support a hypothesis of lineage divergence and a lack of contemporary gene flow between North American and Scandinavian populations of T. populinum. Migration corridors were present either via the Bering Land Bridge (BLB) from eastern Siberia to Alaska or via the North Atlantic Land Bridge (NALB) between Europe and North America through Greenland. The NALB was intact until c. 40 Ma, but functional as late as 25–15 Ma (Milne, 2006). However, recently, Denk et al. (2010) concluded that the NALB may have been an active route into the late Miocene (11.6–5.3 Ma). The BLB was severed approx. 5.4–5.5 Ma (Gladenkov et al., 2002) and may have been functional continuously between 65 and 5.4 Ma, (Milne, 2006). Ingvarsson (2005), using an average (five gene) silent site divergence between P. tremula and P. trichocarpa of 6.1% and assuming a silent site substitution rate of 5.0–8.0 × 10−9, estimated the divergence of P. tremula and P. trichocarpa to have been between 3.8 and 6.2 Ma. Our coalescent-based estimates also place the divergence of P. tremula and P. balsamifera/P. trichocarpa during the Late Miocene to Pliocene (5.3–2.6 Ma) epochs, roughly coincident with the opening of the Bering Strait (Milne, 2006). However, estimates of divergence in Tricholoma are much more recent and place the divergence of these lineages during the Pleistocene (2.6–0.01 Ma) epoch. Our divergence estimates strongly suggest that Tricholoma and its Populus hosts did not undergo intercontinental migration in tandem.

Possible explanations for the strongly supported, but more recent, intercontinental lineage divergence within T. populinum include the following: allopatric isolation of T. populinum populations in Scandinavia and North America that was coincident with speciation of Populus hosts (P. tremula and P. balsamifera/P. trichocarpa); effects of population contraction and expansion during the glacial and interglacial cycles of the Pleistocene; and founder effect followed by genetic drift after a single, rare long-distance dispersal event. Since there were no shared haplotypes between continents, analyses of directional migration were not possible.

Determining the possible role coevolution with different Populus species (P. tremula vs P. balsamifera/P. trichocarpa) had on intercontinental T. populinum divergence will require additional sampling of host populations within Europe and North America or reciprocal inoculation trials. Based on the taxonomic distribution of the host species of T. populinum, it is unlikely that the North American and European lineages of T populinum diverged solely as a result of cospeciation with Populus. It is curious that there are few collections of T. populinum in North America with poplar species other than P. trichocarpa and P. balsamifera, because P. tremula and P. trichocarpa/P. balsamifera reside in two different and well-diverged sections of the genus. This enigma suggests that environmental factors associated with boreal forest-like environments may limit the distribution of T. populinum.

Overall, the extent to which T. populinum occurs in Europe, Asia and North America is not well documented. T. populinum is known to occur with P. alba and P. nigra and these populations need to be sampled to assess population structure and host specificity across Europe. Populations of T. populinum from other Populus species, including quaking aspen, P. tremuloides, which is the North America sister species to P. tremula, are needed to provide a more complete North American phylogeograhic history. Within North America the only potential population sampling of another Populus host came from samples from Colorado where P. tremuloides is dominant, although pockets of P. balsamifera are present. These isolates clearly group within the North American clade in each of the gene trees; however, without a clear identification of host species, no further conclusions are possible at this time.

Estimates of average genetic diversity were low but similar in both Scandinavia and North America (Hd, π, θ; Table 2), reflecting recent demographic events. Low diversity is indicative of a recent founding event or postglacial recolonization bottlenecks (Hewitt, 2004), but also may reflect low mutation rate. Scandinavian populations of P. tremula survived in glacial refugia on the European continent (Birks et al., 2008; Fussi et al., 2010). In North America, recent studies of P. balsamifera found evidence for glacial refugia in southern central Canada (Keller et al., 2010; Levsen et al., 2012) and possibly in a glacial-free region of Beringia (Hultén, 1937; Breen et al., 2012) during the last glacial maximum c. 28 000–15 000 yr before present (Brubaker et al., 2005). T. populinum may have survived with these hosts in glacial refugia and would then have under gone population contraction and expansion during glacial and interglacial periods.

While the number of biogeographic studies of ectomycorrhizal fungi is increasing (Douhan et al., 2011), few Holarctic ectomycorrhizal fungi have well-documented phylogeographies. Studies are beginning to show a recurrent pattern of intercontinental divergence. The ectomycorrhizal fungi Leccinum scabrum and Leccinum holopus both have phylogeographic structure similar to T. populinum, and both Leccinum spp. have genetic discontinuities between the North American and European continents, and little intracontinental phylogeographic structure (den Bakker et al., 2007). These Leccinum spp. are also host specialists, associated only with Betula spp., and occur only in the Northern Hemisphere (den Bakker et al., 2007). In a recent study using microsatellites and nuclear loci, Vincenot et al. (2012) found the ectomycorrhizal fungus Laccaria amethystine formed phylogenetic lineages in Europe and Japan with no shared haplotypes between regions. Genetic structure was not detected and only weak isolation by distance (IBD) was noted within Japan (FST = 0.04; 960 km between the two populations) and Europe (FST = 0.041; 2900 km between the most distance populations) (Vincenot et al., 2012). Our results are very similar to those of Vincenot et al. (2012) at both inter- and intracontinental scales. We also found low differentiation of T. populinum populations in North America that were separated by as much as 2500 km (the maximum distance between our Pacific Northwest and interior Alaskan populations). A major difference between these study systems, however, is that L. amethystine associates with a variety of hosts, while T. populinum only occurs with Populus species. An interesting suggestion by Vincenot et al. (2012) is that L. amethystine may constitute a ring species. Future studies of Holarctic ectomycorrhizal fungi should also consider this hypothesis.

There are, however, also examples of intracontinental population structure within ectomycorrhizal species. A series of studies of A. muscaria, traditionally thought of as a host-generalist mycorrhizal fungus with a broad geographic range, have revealed strong allopatric divergence between Eurasia/Alaska and North American lineages from similar habitats as a result of a lack of gene flow (Oda et al., 2004; Geml et al., 2006, 2008). However, in stark contrast to T. populinum, A. muscaria was demonstrated to be a phylogenetic species complex with strong divergence within North America attributed to ecoregional endemism (Geml et al., 2006, 2008, 2010). Furthermore, phylogenetic species were found in sympatry in several regions (Geml et al., 2006, 2008).

Tricholoma divergence time estimates

In this study, we estimated divergence time in T. populinum using two methods: one based on an estimated average nucleotide substitution rate for the ITS and the second based on a calibration point within the Basidiomycota but external to Tricholoma. Since the likelihood of discovering Tricholoma in the fossil record is very low, internal calibration of estimates of divergence time may never be possible. Hence we used calibration points between major groups within the Basidiomycota. It is clear that the lower substitution rate, 0.1 × 10−9, is not appropriate, as these estimates of divergence for Tricholoma spp. from T. populinum (c. 150 Ma) are approximately the same as the estimated diversification of the Agaricales (Geml et al., 2004; Matheny et al., 2009). By contrast, the congruence of the estimates based on upper (0.87 Ma (95% HPD: 0.35, 1.46)) and average (1.7 Ma (95% HPD: 0.76,2.95 Ma)) ITS nucleotide substitution rates and from calibration points (1.0 Ma (95% HPD: 0.17, 2.40)) within the Basidiomycota is striking. As sequences from more fungal genomes become available, robust estimates of nucleotide substitution rates for increasing numbers of representative Basidiomycota will be developed as well as development of numerous additional loci, both of which will increase the accuracy of divergence time estimates. It should be possible to refine the hypotheses presented here on the divergence time estimates among T. populinum lineages in future studies.

This was the first phylogenetic and phylogeographic study of T. populinum. The only other study examined genet size and longevity at a very fine spatial scale associated with P. nigra in France (Gryta et al., 2006). Thus we know very little about the reproductive biology of T. populinum. The two T. populinum lineages identified here may be considered phylogenetic species, or cryptic species, according to genealogical concordance phylogenetic species recognition (GCPSR; Taylor et al., 2000). The number of different lineages in the T. populinum complex can only be ascertained after additional sampling of hosts and geographic locations as discussed earlier. The degree to which T. populinum lineages remain interfertile and undergo hybridization as a result of secondary contact between North American T. populinum lineage(s) and T. populinum lineage(s) associated with European or Asian poplars has not been examined. In this study, North American T. populinum populations from field collections were associated with host populations devoid of introduced Populus spp. Populus spp. frequently hybridize and it may be possible to examine hybridization of T. populinum from known Populus hybrid zones in future work.

Our results provide strong evidence for a lack of ongoing gene flow between North American and Scandinavian populations of T. populinum. The emerging trend of intercontinental genetic breaks in ectomycorrhizal fungi from mid-latitudes implies that fungal biogeography will be a rich area of inquiry, with relevance to past and present organismal responses to climate change.

Acknowledgements

Funding for this research was provided by NSF EPSCoR grant EPS-0346770, additional support to L.C.G. from NSF grants DEB-1050315 and DEB-1050292, and NSF DBI-0701911 and DBI-1137001 to M.S.O. We thank the following herbaria for loans of Tricholoma populinum: Natural History Museum University of Oslo, Denver Botanical Gardens, Field Museum of Natural History of Chicago, Canadian Forest Service Pacific Forestry Center Forest Pathology Herbarium (DAVFP), Oregon State University Mycological Collection, and the Royal Ontario Museum. We also thank the following people for contributing collections of T. populinum to this project: Gary Larsen, Drew Parker, Ed Swanberg, Andy F. S. Taylor and Jim Trappe. We thank three anonymous reviewers and Marc-André Selosse for thoughtful, critical comments and suggestions that improved the quality of the manuscript. We also thank Roger Ruess and Amy Breen for assistance collecting specimens and Shawn Houston at UAF Life Sciences Informatics for assistance with technical computer analysis. The Computing Portal managed by UAF Life Sciences Informatics (http://biotech.inbre.alaska.edu) was used to run preliminary analyses using Phase, MrModeltest, PAUP* and MrBayes. UAF Life Science Informatics as a core research resource is supported by grant number RR016466 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH). Part of this work was carried out by using the resources of the Computational Biology Service Unit from Cornell University which is partially funded by Microsoft Corporation.

Ancillary