Seasonal trends in the biomass and structure of bryophyte-associated fungal communities explored by 454 pyrosequencing


  • Marie L. Davey,

    1. Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, PO Box 5003, NO-1432 Ås, Norway
    2. Microbial Evolution Research Group (MERG), Department of Biology, University of Oslo, PO Box 1066 Blindern, NO-0316 Oslo, Norway
    Search for more papers by this author
  • Einar Heegaard,

    1. Norwegian Forest and Landscape Institute, Fanaflaten 4, NO-5244 Fana, Norway
    Search for more papers by this author
  • Rune Halvorsen,

    1. Department of Botany, Natural History Museum, University of Oslo, PO Box 1172 Blindern, NO-0318 Oslo, Norway
    Search for more papers by this author
  • Mikael Ohlson,

    1. Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, PO Box 5003, NO-1432 Ås, Norway
    Search for more papers by this author
  • Håvard Kauserud

    1. Microbial Evolution Research Group (MERG), Department of Biology, University of Oslo, PO Box 1066 Blindern, NO-0316 Oslo, Norway
    Search for more papers by this author

Author for correspondence:
Marie Louise Davey
Tel: +47 6496 5347


  • Bryophytes are a dominant vegetation component of the boreal forest, but little is known about their associated fungal communities, including seasonal variation within them.
  • Seasonal variation in the fungal biomass and composition of fungal communities associated with three widespread boreal bryophytes was investigated using HPLC assays of ergosterol and amplicon pyrosequencing of the internal transcribed spacer 2 (ITS2) region of rDNA.
  • The bryophyte phyllosphere community was dominated by Ascomycota. Fungal biomass did not decline appreciably in winter (= 0.272). Significant host-specific patterns in seasonal variation of biomass were detected (= 0.003). Although seasonal effects were not the primary factors structuring community composition, collection date significantly explained (= 0.001) variation not attributed to locality, host, and tissue. Community homogenization and a reduction in turnover occurred with the onset of frost events and subzero air and soil temperatures. Fluctuations in the relative abundance of particular fungal groups seem to reflect the nature of their association with mosses, although conclusions are drawn with caution because of potential methodological bias.
  • The moss-associated fungal community is dynamic, exhibiting seasonal turnover in composition and relative abundance of different fungal groups, and significant fungal biomass is present year-round, suggesting a winter-active fungal community.


The boreal forest is the world’s second-largest terrestrial biome, accounting for 8–11% of global landmass and encompassing significant areas of Alaska, Canada, Fennoscandia and Siberia (Whittaker, 1975; Melillo et al., 1993). The biome is of rising global concern as its soils and vegetation represent major storage and sequestration components of the global carbon cycle, and are expected to be affected by climate change (Melillo et al., 1993; Gower, 2003). It is becoming increasingly apparent that microbial communities not only play vital ecological roles in the boreal ecosystem and interact with other organisms: they are also of great importance in nutrient and carbon cycling (Hattenschwiler et al., 2005; van der Heijden et al., 2008; McGuire & Treseder, 2010; Allison & Treseder, 2011). Knowledge of boreal microbial communities is largely limited to soil microbes (Hattenschwiler et al., 2005; van der Heijden et al., 2008; Allison & Treseder, 2011) and those ectomycorrhizal fungi associated with vascular plants (Buscot et al., 2000; Read & Perez-Moreno, 2003; Lehto & Zwiazek, 2011). Bryophytes are a dominant vegetation component of the boreal forest with significant contributions to carbon and nitrogen cycling (Turetsky, 2003; Lindo & Gonzalez, 2010), and they provide unique microbial habitats as a result of their stems being a continuous gradient from dead to senescing to photosynthetic tissues. Despite this, relatively little is known about their associated microbial communities. Recent research suggests that cyanobacterial moss associates produce the majority of fixed nitrogen in the boreal forest (DeLuca et al., 2002) and demonstrates that this fixation rate is affected by precipitation (Jackson et al., 2011) and temperature (Gentili et al., 2005). By comparison, the fungal component of the moss-associated microbial community and their ecological function have remained poorly characterized, although Racovitza (1959) and Davey & Currah (2006) report a variety of fungal pathogens, parasites, and saprophytes to be associated with mosses. Because of the high fungal biomass associated with senescent bryophyte tissues (Davey et al., 2009a) and active bryophyte cell wall degradation exhibited by some fungi (Day & Currah, 2011a,b), it is thought that fungi are the primary decomposers of moss substrates. It has also been suggested that bryophytes may act as inoculum reservoirs for some tree pathogens (Davey et al., 2010), that moss pathogens create micro-disturbances that ultimately alter plant community composition on a local scale (Hoshino et al., 2001), and that ectomycorrhizal fungi associated with trees are able to scavenge nutrients from moss substrates (Carleton & Read, 1991). Such diverse ecological roles highlight the probable ecological significance of this poorly known group of fungi to the boreal forest ecosystem.

Seasonality of fungal reproduction has long been observed, particularly for macrofungi (Kauserud et al., 2010; Gates et al., 2011). However, fungal communities have also been found to exhibit seasonal variations in total biomass (Berg et al., 1998; Björk et al., 2008; Artigas et al., 2009; Randhawa et al., 2011), community composition and diversity (Schadt et al., 2003; de Román & de Miguel, 2005; Kennedy et al., 2006; Zinger et al., 2009; Dumbrell et al., 2011; Jumpponen, 2011), and functional activity (i.e. decomposition and mycorrhizal nutrient transfer) (Bentivenga & Hetrick, 1992; Buée et al., 2005; Uchida et al., 2005; Capps et al., 2011). Factors such as temperature, precipitation, snow cover, and nutrient availability have been identified as possible driving factors behind temporal variation in fungal communities (Steinaker & Wilson, 2008; Artigas et al., 2009; Zinger et al., 2009; Kauserud et al., 2011). While seasonal patterns have been well characterized in ectomycorrhizal (de Román & de Miguel, 2005; Steinaker & Wilson, 2008; Jumpponen et al., 2010) and arbuscular mycorrhizal fungi (Mandyam & Jumpponen, 2008; Dumbrell et al., 2011), aquatic hyphomycetes (Barlocher, 2000; Artigas et al., 2009; Anderson & Shearer, 2011), and soil fungal communities (Berg et al., 1998; Schadt et al., 2003; Björk et al., 2008; Zinger et al., 2009), relatively few studies have examined the seasonality of fungi associated with the phyllosphere of plants (Cabral, 1985; Osono & Mori, 2005; Osono, 2008; Albrectsen et al., 2010; Jumpponen & Jones, 2010). More specifically, seasonal variation in community composition and the fungal biomass associated with the phyllosphere of terrestrial bryophytes have not been previously investigated. However, decomposition rates of dead and senescent bryophyte material are known to be temperature dependent (Nakatsubo et al., 1997), which suggests that, at a minimum, seasonal variation in microbial community activity associated with senescent tissues occurs. Given the importance of bryophytes and their associated microbial communities in carbon and nitrogen cycling (Turetsky, 2003; Lindo & Gonzalez, 2010), understanding seasonal changes in moss-associated fungal communities is of particular relevance to understanding the functioning of the boreal forest and predicting its responses to global change.

Next-generation sequencing techniques have led to a revolution in microbial ecology by providing opportunities to generate unprecedented numbers of sequences and detect very rare or low-abundance organisms (Begerow et al., 2010; Ekblom & Galindo, 2011). These techniques have been successfully applied to examine seasonality of fungal communities in several instances (Jumpponen & Jones, 2010; Jumpponen et al., 2010; Dumbrell et al., 2011). In this study, 454 amplicon pyrosequencing and fungal biomass assays were used to investigate seasonal variation in fungal communities associated with the photosynthetic and senescent tissues of three common boreal mosses. The specific effects of host and tissue type on community composition are to be addressed in a forthcoming manuscript (M. L. Davey et al., unpublished) and will be mentioned only cursorily here.

Materials and Methods

Study sites and sampling

Five 1-m-diameter circular plots containing the mosses Hylocomium splendens (Hedw.) Schimp., Pleurozium schreberi (Brid.) Mitt., and Polytrichum commune Hedw. were established in each of two mature secondary Picea abies (L.) Karst. spruce forests near Ås in south-east Norway (Myrvoll: 59°39·687′N 10°50·703′E; Nordskogen: 59°40·463′N 10°45·706′E). Vegetation in the two forests was of the bilberry–Norway spruce forest type (Kielland-Lund, 1982) and the area exhibited a typical boreal climate with cold winters, relatively warm summers, and an average yearly temperature of 5.9°C (Supporting Information Fig. S1). In each plot, one colony of the three moss species was selected and marked, and representative shoots were collected for biomass and genetic analyses every 8 wk between late April 2009 and early January 2010. At the time of collection, all plant material was cleaned of coarse debris and rinsed in running water. Individual stems of each bryophyte species were separated into green photosynthetic and brown senescent sections, and the intervening parts discarded. For ergosterol analyses, shoot fragments were freeze-dried overnight, crushed to a powder using a Retsch ball mill (Retsch, Dusseldorf, Germany), and stored at −80°C until analysis. For sequencing, one shoot was chosen to be representative for each host colony at each sampling date. The resultant 150 shoots were individually rinsed in running water, washed with agitation in 0.01% Triton-X, and then rinsed three times in sterile distilled water before being divided into fragments as described above and freeze-dried. Dried shoot fragments were crushed using a Retsch ball mill, and stored in 2 × cetyl-trimethylammonium bromide (CTAB) extraction buffer at −80°C until further analysis.

Ergosterol extraction and quantification

Free and esterified ergosterol was extracted with saponification from all bryophyte tissues. Approximately 200 mg of powdered tissue was combined with 20 ml of 3 M KOH in ethanol and incubated at 80°C in a water bath for 1 h. Extraction mixtures were diluted with 5 ml of distilled water, and ergosterol was extracted from the supernatant by two successive applications of 7.5 ml of hexane. The organic fractions were then combined, evaporated, and resuspended in 1 ml of methanol. Extracts were analysed on an 1100 series HPLC (Agilent Technologies, Waldbronn, Germany). Ergosterol was separated on a Luna® 5-μm particle size 4.6 × 150 mm column (Phenomenex, Værløse, Denmark) with methanol as the mobile phase at a flow rate of 2 ml min−1. The total analysis time was 10 min, and the detection wavelength was 282.4 nm. Ethanol blanks were analysed after every tenth sample, and two samples were used to calculate the standard error by performing five successive injections and analyses of the sample. The identification of ergosterol was based on retention time, online UV spectra and co-chromatography with a commercial ergosterol standard (Sigma, St Louis, MO, USA).


Genomic DNA was extracted from each of the 300 shoot fragments using a modified CTAB-based extraction protocol (Murray & Thompson, 1980; Gardes & Bruns, 1993) and all extracts were cleaned and purified using the Wizard®SV Gel and PCR Clean-Up System (Promega, USA). A nested PCR approach was used to amplify and tag the internal transcribed spacer 2 (ITS2) region of rDNA. The ITS2 region has been identified as suitable for fungal taxon assignment (Mello et al., 2011), and has been successfully used in other comparative ecology studies where it gives results that are convergent with, if not comparable to, those for other markers (Mello et al., 2011; Arfi et al., 2012; R. Blaalid et al., unpublished). The nested approach was chosen as recommended by Berry et al. (2011) to generate highly replicable results (see Kauserud et al., 2012) and to avoid the ligation bias known to occur in barcode primer-free approaches (Gillevet et al., 2010). The entire ITS region of fungal DNA was first amplified using the fungal-specific primers ITS1-F and ITS4 (White et al., 1990; Gardes & Bruns, 1993), and then a second PCR using the primers ITS3 and ITS4 (White et al., 1990) targeted the ITS2 region and added an emulsion PCR adaptor and one of 19 different 10-bp tags to each sample to allow for individual sample recognition in downstream analyses after pooling. PCR reactions were conducted in 20-μl volumes and contained 0.4 units of Phusion polymerase (Finnzymes Oy, Vantaa, Finland), 0.5 μM of each primer, 1.7 μM of dNTP, and 1 × Phusion HF PCR-buffer (Finnzymes Oy). In the first PCR, 2 μl of a 10 × dilution of extracted genomic DNA was used as a template, while in the second 4 μl of a 50 × dilution of the first PCR product was used as a template. PCR conditions were as follows: initial heating to 98°C for 30 s, 30 cycles of denaturation at 98°C (10 s), annealing at 53°C (20 s) and extension at 72°C (10 s), followed by a 7-min extension at 72°C before storage at 4°C. Conditions in the second PCR were identical to those in the first, with the exception that the denaturing-annealing-extension cycle was repeated only 10 times, which was determined to be the minimum number of cycles required to reliably detect successful amplification using gel electrophoresis. Amplicons from the second PCR were cleaned and purified using the Wizard®SV Gel and PCR Clean-Up System (Promega) according to the manufacturer’s instructions. Of the 294 samples that were successfully amplified, seven were selected at random to be sequenced twice to determine the validity of read abundance data. Amplicons were then quantified and pooled into 16 equimolar libraries, each containing 19 uniquely tagged samples. The amplicon libraries were subsequently sequenced on a full plate split into 16 lanes using the Roche GS FLX Titanium Series 454 sequencing platform at the Norwegian High-Throughput Sequencing Centre (University of Oslo, Oslo, Norway).

Bioinformatics and statistical analyses

Raw sequences are publicly available at the MG-RAST metagenomics analysis server (Project Name: Seasonality of Bryophilous Fungi). Sequences were quality-filtered, denoised, and clustered using Qiime v. 1.3.0 (Caporaso et al., 2010). Reads with length < 200 bp, an average Phred quality score of < 25, or errors in the tags were discarded. Those sequences with homopolymers of length > 10 bp, more than one ambiguous base call (N ), and more than one error in the forward primer sequence were also filtered from the data set. In addition, a 50-bp sliding window was used to identify regions of low sequence quality (average quality score < 25) and truncate the sequence at the beginning of the low-quality window. Those truncated sequences still meeting the minimum length requirement (200 bp) were retained in the data set. The resulting 327 273 reads were subject to denoising using Denoiser v. 0.91 (Reeder & Knight, 2010) as implemented in Qiime v 1.3.0 (Caporaso et al., 2010). Denoised sequences were clustered into operational taxonomic units (OTUs) using a 97% similarity threshold and the uclust algorithm with optimal uclust settings as implemented in Qiime v. 1.3.0 (Caporaso et al., 2010; Edgar, 2010). Those clusters represented globally by a single sequence were discarded as likely sequencing errors, as suggested by Tedersoo et al. (2010). The most abundant sequence in each cluster was designated as the representative sequence and was subjected to BLAST searches against the NCBI-nr/nt database (downloaded November 2011) and a customized database combining the UNITE and INSDC databases (downloaded November 2011). Taxonomy was assigned at the order level, as Ovaskainen et al. (2010) suggest only a moderate fraction of 454 reads can be reliably identified to genus using BLAST. Among those sequences successfully assigned to the order level, the two databases agreed in 92% of cases; however, 62% of sequences could only be identified to the level of kingdom or phylum using the UNITE + INSDC database, and an additional 11% were only identified to the class level. Thus, all final taxonomic assignments were made at the level of order and based on critical reviews of the top 10 BLAST matches to the NCBI-nr/nt database. All OTUs with BLAST matches to nonfungal organisms were removed, as were those that could not reliably be identified as fungi (best BLAST match < 80% sequence similarity or < 50% coverage). When the top BLAST match was to an unclassified or uncultured fungus, the top 10 matches were screened for concordance and taxonomy assigned based on the most parsimonious lineage among those matches meeting the minimum thresholds of > 80% sequence similarity and > 50% coverage.

EstimateS v. 8.2.0 (Colwell, 2009) was used to calculate rarefaction curves for the entire data set. The seven samples that were sequenced twice were used to assess the validity of between-sample abundance comparisons. A global nonmetric multidimensional scaling (GNMDS) ordination (see the Methods section) coupled with vector fitting (Fig. S2) demonstrates that sample pairs display greater inter-sample variation than within-sample pair variation. A paired t-test of rarified abundances in each of the seven pairs of resequenced control samples indicates that there is no statistically significant difference between the abundance measures of individual OTUs between sequencing replicates of the same sample (= 0.582). Although read abundance cannot be used for direct quantification without calibration and does not allow for valid between-taxa comparisons, it has been suggested to provide reliable measures for within-taxon comparisons (Amend et al., 2010). Abundance-based comparisons were therefore made solely within selected taxonomic groups (ectomycorrhizal families, Exobasidiales, Chaetothyriales, Mycena, Pleosporales, and Helotiales) using an OTU table that was rarified in Qiime v 1.3.0 (Caporaso et al., 2010) to an even sampling depth of 500 sequences per sample in order to minimize effects of unequal sequencing depth between samples. Chi-square tests were used to determine if abundances at the different collection dates varied significantly from one another.

Linear mixed effects models and general linear mixed effects models were fitted using the lme4 and nlme packages of R (Bates et al., 2011; Pinheiro et al., 2011) to examine the response of fungal biomass (i.e. ergosterol) and OTU richness to host species, tissue type, and sampling date. For fungal biomass we assume a normal distribution, whereas for OTU richness we assume a Poisson distribution. In both cases, models were nested in accordance with the experimental design. We recognized random contributions by the components forest and plot (for biomass data) or shoot (for OTU richness). For the biomass analysis, the inclusion of a level-specific residual variance at level combinations of host and forest was required to compensate for differences in variance. The analysis of OTU richness required the inclusion of a specific term accounting for number of reads (Zuur et al., 2009) and an observation-specific random term accounting for potential overdispersion. Log-likelihood ratio tests were used to determine the overall significance of the collection date as a fixed effect in the model and to determine whether interaction terms should be included in the final model.

The data set was transformed using the Hellinger equation to account for so-called ‘blind sampling’ and large numbers of absences in the data sets, as suggested by Ramette (2007), and then split into host/tissue-type subsets, and a subset containing only the seven samples that were sequenced twice. The vegan v.2.0.3 package implemented in R v.2.15.0 (Oksanen et al., 2011; R Development Core Team 2011) was used to conduct GNMDS (Kruskal, 1964a,b; Minchin, 1987) with the following options (following recommendations by Økland, 1996a and Liu et al., 2008): distance measure = Bray–Curtis distance, dimensions = 2 and 3, initial configurations = 100, maximum iterations = 200, and convergence ratio for stress = 0.9999. Because the underlying gradients in our study are unknown, detrended correspondence analysis (DCA) (Hill, 1979; Hill & Gauch, 1980) was also run with default options on all data subsets and all ordinations were inspected for known artefacts such as the arch effect (in GNMDS), tongue effect and other patterns (Økland, 1990; Økland & Eilertsen, 1993). Similar results obtained using the two methods and absence of visual artefacts is interpreted as a strong indication that a reliable gradient structure has been found (Økland, 1996b) and this was confirmed by calculating correlations between DCA and GNMDS axes using Kendall’s rank correlation coefficient τ for GNMDS ordinations with both two and three dimensions (Table S1). The larger number of dimensions was only accepted when correlations between corresponding DCA and GNMDS axes were better at three dimensions than at two. The envfit function in vegan was used to fit the variables of sampling date, precipitation, minimum air temperature, and soil temperature at 2-cm depth to the ordination diagrams. The effect of collection date on the complete data set was investigated using a partial canonical correspondence analysis (CCA) (ter Braak, 1986) in which the effects of location, host species, and tissue type were partialled out of the constrained ordination. The significance of collection date as an explanatory variable for the remaining variation in the data set was assessed by conducting 999 random permutations of the community data.


Data characteristics

Of the 491 280 reads generated, 327 273 were retained after the filtering and denoising steps, and subsequently clustered into 9367 OTUs. After removal of singletons (> 60% of OTUs) and nontarget organisms, the final data set included 2676 OTUs encompassing 78% of the reads retained after filtering and denoising (Table S2). The average number of OTUs detected per sample was 91 (range: 20–184) with a mean of 1031 reads generated per sample (range: 502–2423). On average, each OTU occurred in 8.4 samples (range: 1–224) and contained 96 reads (range: 1–11 716), with the majority of OTUs having a frequency of < 10 (83%) and containing < 10 reads (62%). Neither the species accumulation curves (excluding global singleton OTUs) for the entire data set nor those for each sampling date approached an asymptote (Fig. S3). OTUs representing 67 orders of fungi were detected. Fifty per cent of OTUs (56% of reads) belonged to the Ascomycota, while 37% (38% of reads) belonged to Basidiomycota, and a small proportion (4%; 1.6% of reads) were assigned to the Chytridiomycota, Glomeromycota, and Mucormycotina. Dominant orders among the Ascomycota were the Helotiales, Chaetothyriales, Pleosporales and Capnodiales. Dominant basidiomycete groups included the Agaricales, Tremellales, and Sporidiobolales (Table 1).

Table 1.  Summary of the distribution of operational taxonomic units (OTUs) among fungal lineages including the 10 most abundant orders detected
Taxnomic affinityPercent of OTUsPercent of reads
 Orbiliomycetes0.6< 0.1
 Taphrinomycetes0.4< 0.1
 Unassigned Ascomycota8.610.4
 Unassigned Basidiomycota4.07.4

Seasonal variations in community biomass

Ergosterol was detected in variable amounts in all species and tissues throughout the year, although large variations in ergosterol content often occurred within a single host-tissue class. Ergosterol content per gram DW moss tissue was higher in senescent tissues than in photosynthetic tissues. Ergosterol content also varied among the three hosts with P. schreberi >  H. splendens Pol. commune, regardless of the host tissue being examined (Fig. 1). The linear mixed effects model identified: an effect of both tissue type and host species on the amount of ergosterol present; no effect of tissue type on ergosterol variation between collection dates; and a significant difference in ergosterol variation between sampling dates among the three hosts: ergosterol levels remained relatively constant throughout the year in both Pol. commune and H. splendens, while they peaked in August in P. schreberi (Fig. 1, Table 2).

Figure 1.

Linear mixed model-estimated ergosterol levels for photosynthetic (green) and senescent (brown) tissues of Hylocomium splendens, Pleurozium schreberi, and Polytrichum commune from April to January. The fixed effects of tissue type and host, and the interactions between tissue and host, and host and month were found to be significant (see Table 2).

Table 2.  Summary statistics for the best linear mixed model fitted to the amount of ergosterol detected in moss tissues with the fixed effects of tissue type, host, month, and the interactions between tissue and host, and host and month
 numDFdenDF F-value P-value
  1. numDF, numerator degrees of freedom; denDF, denominator degrees of freedom.

  2. Forest type and sampling plot were included as random factors. An additional contribution of difference in variance was included at the combined levels of host and forest. The standard deviations of random components were as follows: plot : forest = 80.3113; forest = 0.02576.

  3. Statistically significant values are indicated in bold text.

Tissue : host22435.7311 0.0037
Host : month82433.0691 0.0026

Seasonal variation in fungal diversity

Per-sample OTU richness showed large variation within a single sampling event, but also varied between collection dates (Fig. 2). Collection date was a statistically significant component (< 0.001) of the generalized linear mixed model, which was conditioned on a term for the number of reads per sample and identified an effect of both tissue type and month on the number of OTUs detected; and no differences in variation between collection dates for the different hosts and tissue types (Fig. 2, Table 3).

Figure 2.

Generalized linear mixed model-estimated per sample operational taxonomic unit (OTU) richness for photosynthetic (green) and senescent (brown) tissues of Hylocomium splendens, Pleurozium schreberi, and Polytrichum commune from April to January. See Table 3 for the significance of fixed and random effects.

Table 3.  Fixed effects table for the generalized linear mixed model fitted to the number of operational taxonomic units (OTUs) detected in moss tissues with the fixed effects of tissue type, host, and month, and a term representing the number of reads per sample
 EstimateSE z valuePr(> |z|)
  1. Forest locality, shoot number (shootID) and observation number (obs) were included as nested random factors (the standard deviation of the random components are as follows: obs:(shootID : forest) = 0.2745; shootID : forest = 0.0650; forest < 0.0001).

  2. Statistically significant values are indicated in bold text.


Specific fungal groups detected in the mosses also exhibited seasonal variations in abundance that were often tissue-type specific (Fig. 3) but were generally not host-specific (data not shown). Ectomycorrhizal (ECM) fungi exhibited a peak in relative abundance (measured as the proportion of ECM reads per sampling event) during August in senescent moss tissues, while this peak occurred in October in photosynthetic tissues. The Exobasidiales were detected in very low numbers in senescent tissues, but were more abundant in green tissues, exhibiting a clear seasonality, with abundance peaking in August. The Pleosporales were present in both tissue types and exhibit a summer abundance peak in photosynthetic tissues. Mycena, a basidiomycete saprophyte, was more abundant in senescent tissues than photosynthetic tissues, and peaked during the summer months, a trend that also occurred in the ascomycete order Chaetothyriales. The most abundant group present in both tissues was the Helotiales, which exhibited a slight decline in relative abundance during August.

Figure 3.

Seasonal trends in the relative abundance of selected taxonomic groups in photosynthetic (green lines) and senescent (brown lines) moss tissues. Groups in which fluctuations in abundances are statistically significantly larger than expected at random are indicated by a ‘*’ at the end of the line. ECM, ectomycorrhizal.

Seasonal variation in community structure

GNMDS ordination of Hellinger-transformed abundance data partitions from all species–tissue combinations showed similar trends: vector fitting indicated that all the data are structured to some extent by sampling date, precipitation, minimum air temperature, and soil temperature at 2-cm depth (Fig. 4). In the CCA ordination in which the effects of locality, host, and tissue type were removed (3% of total variation in species composition), 2% of the remaining total variability in species composition was significantly explained by seasonal effects (= 0.001) (Fig. 5). The samples distributed along the axis CCA1 largely according to minimum air temperature and soil temperature at 2 cm, and low CCA axis 1 scores represented minimum air temperatures above the freezing point and soil temperatures > 5°C. Samples were distributed along CCA axis 2 somewhat according to precipitation received. High scores along this axis were associated almost exclusively with samples from August, where precipitation was nearly three times higher than in any other month.

Figure 4.

Ordination diagrams for global nonmetric multidimensional scaling (GNMDS) ordinations of the fungal operational taxonomic unit (OTU) composition for each combination of host and tissue type. Arrows representing the direction of maximum increase for minimum air temperature (AirTemp), soil temperature at 2-cm depth (SoilTemp), and monthly precipitation (Precipitation) are black when statistically significant (< 0.05), and grey when nonsignificant. Centroids for each collection date are marked with a black ‘X’ when significantly different from one another (< 0.05) and a grey ‘+’ when the difference is nonsignificant. Stress values for each ordination are as follows: Hylcomium green: 23.99957; Hylocomium brown: 26.29371; Pleurozium green: 20.54731; Pleurozium brown: 24.62266; Polytrichum green: 16.47877; Polytrichum brown: 25.96876.

Figure 5.

Canonical correspondence analysis (CCA) ordination of the fungal operational taxonomic unit (OTU) composition with the effects of forest locality, host, and tissue partialed out. Arrows point in the direction of maximum increase of minimum air temperature (AirTemp), soil temperature at 2-cm depth (SoilTemp), and monthly precipitation (Precipitation), while ‘X’ marks the centroid for each collection date. All variables and factors had significant effects (< 0.05) on the ordination configuration. Conditioned and constrained variables represented 3.0% and 2.2% of the total variation, respectively, while 94.8% remained unconstrained.


Overall community OTU richness and taxonomic diversity

A large and diverse assemblage of moss-associated fungi was detected, and species accumulation curves failed to approach an asymptote, suggesting that sequencing depth was insufficient to capture the full community diversity. More than 2600 nonsingleton OTUs were recovered, which is an order of magnitude more than in other published phyllosphere studies (Jumpponen & Jones, 2009; Arfi et al., 2012), studies of agricultural soils (Xu et al., 2012), and studies of mycorrhizal root systems (Jumpponen et al., 2010; Tedersoo et al., 2010; Blaalid et al., 2012). The OTU richness in mosses is of comparable magnitude to the diversity reported in natural soils (Buée et al., 2009; Lentendu et al., 2011). Possible explanations for this high diversity are the perennial life history strategy of the mosses which allows for continual fungal growth and colonization, and the close physical proximity of the senescent portions of the moss to the soil fungal spore bank. However, the validity of between-study comparisons is questionable, as sequencing depth is generally insufficient to capture complete community diversity. Furthermore, methodologies are inconsistent among existing fungal diversity pyrosequencing studies, and OTU detection and richness estimations are dependent on experimental methodology, sequencing depth, and library size, as well as the bioinformatic processing of the raw data (ex/denoising) (Quince et al., 2009; Kunin et al., 2010; Schloss, 2010; Gihring et al., 2012).

The bryophyte-associated fungal community detected was dominated by ascomycetes. This may partly be a product of preferential ascomycete amplification by the ITS3/ITS4 primer pair (Bellemain et al., 2010). However, other studies of vascular plant phyllosphere communities using different or multiple primer combinations also recover ascomycete-dominated communities (Jumpponen & Jones, 2009; Arfi et al., 2012). Furthermore, the most prevalent orders detected (Helotiales, Chaetothyriales, Agaricales, and Tremellales) were consistent with those described from the same host species by Kauserud et al. (2008), who used a cloning and Sanger sequencing-based approach with a primer pair not specifically biased to ascomycete amplification. It therefore seems likely that our recovery of an ascomycete-dominated community is not an artefact. Despite insufficient sampling and sequencing depth, the pyrosequencing-based approach detected an order of magnitude more OTUs than the clone-based approach, further highlighting the usefulness of this technology in complex systems.

Seasonal variation in fungal community biomass

As observed by Davey et al. (2009a), significant amounts of fungal biomass are associated with each moss host, and biomass varies between hosts and tissue types, with brown tissues containing more fungal biomass than green, and Pol. commune hosting less fungal biomass than P. schreberi and H. splendens. Notably, in all hosts and tissue types, winter measurements of fungal biomass (October/January), during which soil and air temperatures were subzero and the hosts were beneath the snowpack, did not fall significantly below levels observed when soil and air temperatures were above zero (April/June sampling dates). Given that ergosterol is considered a proxy for living fungal biomass (Ekblad et al., 1998), our findings are consistent with a growing body of research that indicates that fungal cells not only survive subzero temperatures, but actively grow beneath the winter snowpack (Schadt et al., 2003; Schmidt et al., 2008; Matsumoto, 2009; Haei et al., 2011). Although hyphal mats have been reported growing on soil and plant litter beneath snow cover (Schmidt et al., 2008; Matsumoto, 2009), such mats were not observed on winter-collected bryophtyes, and there was no detectable abundance increase in the psychrotolerant Mucormycotina which often form such mats (Schmidt et al., 2008). The presence of a significant, living, winter fungal community suggests that boreal bryophyte-associated fungi are similar to those associated with vascular plant litter and soil in alpine ecosystems, which are important in carbon and nitrogen cycling, and decomposition during winter months (Bardgett et al., 2005). This points to moss-associated fungi possibly playing a significant role in winter carbon flux in the boreal forest.

Fluctuations in fungal biomass between sampling dates were inconsistent between hosts: biomass associated with Pol. commune and H. splendens was relatively constant between sampling dates, while P. schreberi exhibited a significant biomass peak in August. This biomass increase was not associated with an abundance or diversity increase in a particular OTU or group of OTUs. However, bryophytes are known to release bursts of aqueous nutrient leachates with repeated drying–wetting cycles that are useable by ectomycorrhizal fungi (Carleton & Read, 1991; Wilson & Coxson, 1999) and presumably other fungi. It is possible that P. schreberi releases more nutrients than H. splendens and Pol. commune during these cycles, and consequently is able to sustain a larger phyllosphere community during the August collection date, when precipitation is maximal. Tissue type did not affect seasonal variation of ergosterol content, suggesting that those factors controlling seasonal biomass fluctuations act equally on photosynthetic and senescent tissues. Although different fungal groups produce varying amounts of ergosterol and the compound is absent in some groups (i.e. Chytridiomycota and Glomeromycota) (Ruzicka et al., 2000; Weete et al., 2007), we assume that biomass comparisons are valid because few fungi lacking ergosterol were detected (< 5% of OTUs) and communities had similar order-level composition across all sampling dates.

Seasonal variation in community structure

Bryophyte-associated fungal communities are primarily structured by host identity and tissue type (M. L. Davey et al., unpublished); however, effects of collection date were also detected. CCA analysis found 2% of variation remaining after partialling out of the effects of host, forest, and tissue, which is statistically significant (= 0.001). This variation is explained by the collection date (note that the total variation as given by the total inertia of CCAs is inflated by lack-of-fit-of-data-to-model ‘variation’ of unknown quantities (Økland, 1999) so that the fraction of ‘explainable’ variation in species composition is therefore likely to be 1.5–2.5 times > 2%). This demonstrates that, like soils and vascular plant roots and shoots (Osono & Mori, 2005; Courty et al., 2008; Zinger et al., 2009; Jumpponen & Jones, 2010; Jumpponen et al., 2010), bryophytes host a seasonally dynamic fungal community. Precipitation, minimum air temperature and soil temperature at 2-cm depth significantly structure ordination analyses, with individual collection dates tending to cluster, indicating that compositional turnover takes place in the fungal community. This is consistent with vascular plant-associated communities and soils (Schadt et al., 2003; Walker et al., 2008; Jumpponen & Jones, 2010; Jumpponen et al., 2010; Dumbrell et al., 2011), where significant community composition turnover can occur in a matter of months. However, soil (Schadt et al., 2003) and mycorrhizal (Koide et al., 2007; Dumbrell et al., 2011) communities exhibit more dramatic seasonal and temporal turnover than bryophyte-associated communities, with nearly complete turnover occurring in some cases (i.e. Schadt et al., 2003; Dumbrell et al., 2011). In CCA analysis it becomes clear that the winter samples (October/January) represent a community that is distinct from other sampling dates, and that compositional turnover is reduced during months experiencing subzero temperatures. Zinger et al. (2009) observed similar winter homogenization of alpine soil communities that was concomitant with temperature and soil pH. It is unclear whether the observed homogenization phenomenon can be attributed to the existence of a fungal assemblage that is strongly favoured by its tolerance to freezing events (Henis et al., 1987), or whether other seasonally variable environmental factors (i.e. pH) are the driving factor reducing community turnover. The observed changes in community composition are concurrent with fluctuations in soil and air temperatures, and precipitation, suggesting that these communities are likely to be sensitive to the environmental perturbations expected from global climate change. While bryophyte-associated fungi are clearly responsive to seasonal environmental changes, season is not the dominant factor structuring the community composition and other factors (i.e. host and tissue type; M. L. Davey et al., unpublished) play a more important role. Nevertheless, further research is warranted to determine to what extent the ecological function of this community shifts in conjunction with the observed seasonal changes in community composition, and to determine which environmental factors drive these shifts.

GLMM modelling of OTU richness indicated significant differences between hosts and tissue types, as well as significant changes in OTU richness between collection dates. Seasonal fluctuations in alpha diversity and species richness have also been reported in mycorrhizal fungi (de Román & de Miguel, 2005; Jumpponen et al., 2010; Dumbrell et al., 2011) soil fungi (Toberman et al., 2008) and phyllosphere fungi (Osono & Mori, 2005; Osono, 2008; Jumpponen & Jones, 2010). Although our data suggest that similar variation occurs in moss-associated fungal communities, the magnitude of the observed differences was low (10–15% of total OTUs), and given the insufficient sequencing depth of this study and its impact on OTU detection, we cannot conclusively determine that OTU richness varies seasonally.

Seasonal variation within taxonomic groups

The relative abundance of particular fungal groups varied between sampling dates in some cases, and remained constant in others. For some groups, these patterns provide additional insight into the interactions between the fungi and their moss hosts. For example, ECM fungi peaks in abundance in moss tissue during late summer or autumn, as does their biomass and abundance in boreal forest soils (Wallander et al., 2001). This corroborates reports of ECM fungi actively colonizing moss tissues, as demonstrated by Carleton & Read (1991), and that moss species in addition to P. schreberi (Carleton & Read, 1991) can act as substrates for ECM fungi. Similarly, the Exobasidiales, a group of Ericaceous plant pathogens, are associated primarily with photosynthetic tissues and peak in abundance during August. This coincides with a late summer dispersal and reproduction peak of Exobasidium, which infects the abundant Vaccinium species in the plots (Pehkonen et al., 2002; Pehkonen & Tolvanen, 2008). It seems likely the Exobasidiales are a casual phyllosphere member of the fungal community that are ostensibly deposited on the mosses by nearby infected Ericaceous plants. Although mosses have been reported as potential alternate hosts of vascular plant pathogens (Davey et al., 2010; M. L. Davey, unpublished), it appears unlikely that mosses represent an overwintering host or inoculum reservoir for Exobasidium pathogens. Pleosporalean fungi exhibit the same gradual abundance increase in green tissues throughout the growing season followed by a decrease concomitant with the onset of frost events. However, given that some members of the Pleosporales are known to be moss pathogens (Racovitza, 1959; Davey et al., 2009b; Stenroos et al., 2010; M. L. Davey, unpublished) it seems unlikely that these fungi are casual epiphyllous members of the phyllosphere like the Exobasidiales, and their seasonal abundance variation rather represents a typical Northern Hemisphere fungal disease cycle with an epidemiological peak and dispersal occurring in summer.

A large proportion of the taxonomic groups detected are primarily saprophytic and present in varying abundance throughout the year. Although temporal shifts in the relative importance of bacterial vs fungal decomposers have been documented in soils (Berg et al., 1998; Bardgett et al., 2005; Björk et al., 2008), similar shifts within the fungal community are not well characterized. The dominant order in the bryophyte-associated community, Helotiales, experiences a slight summer decline in abundance. Conversely, other saprophytic groups found in senescent tissues, including the basidiomycete Mycena and the Chaetothyriales, experience spring and summer abundance climaxes. This may reflect shifts in the relative importance of different groups in decomposition throughout the season, although it is unclear whether this could be attributable to competitive ability, or responses to abiotic climatic factors such as temperature and moisture availability. Seasonal changes in saprophyte abundance may be particularly pertinent to understand boreal ecosystem functioning. Despite the importance of bryophytes and their associated microbial communities in carbon and nitrogen cycling (Turetsky, 2003; Lindo & Gonzalez, 2010) and the temperature dependence of bryophyte decomposition rates (Nakatsubo et al., 1997), the bryophyte component of the ecosystem is largely overlooked in models predicting ecosystem responses to global change, if only because of a lack of knowledge (Hobbie et al., 2000; Cornelissen et al., 2007; Lindo & Gonzalez, 2010). Although the seasonal abundance shifts detected herein appear both biologically feasible and ecologically relevant, all conclusions drawn are tentative, as insufficient sequencing depth and the inherent uncertainties of taxonomic identification based on existing sequence databases (Nilsson et al., 2006; Ovaskainen et al., 2010) add significant uncertainty to the reliability of these comparisons.


The bryophyte phyllosphere hosts a large fungal community comparable in magnitude to soil fungal communities, and dominated by the Helotiales (Ascomycota) and the Agaricales (Basidiomycota). Fungal biomass does not decline significantly in winter, which may reflect a fungal community that is active at temperatures near and below 0°C. Variation of fungal biomass between collection dates differed between host species. Although seasonal effects were not the primary factors structuring communities, composition and OTU richness changed between sampling dates concomitantly with changes in precipitation and soil and air temperatures. Some community homogenization and reduction in turnover were detected in conjunction with frost events and subzero temperatures. Fluctuations in the relative abundance of particular fungal groups appear to reflect the nature of their association with the mosses. Saprophytic fungi experienced abundance peaks at different collection dates, highlighting the complexity of carbon cycling in the bryophyte component of the boreal forest ecosystem. It is increasingly evident that the fungal communities associated with mosses are both dynamic and complex systems.


This research was supported by a Miljø-2015 grant from the Research Council of Norway to M.O., and by a post-doctoral fellowship from the Natural Sciences and Engineering Research Council of Canada to M.L.D. The authors thank Tor Carlsen and MERG colleagues for discussion of 454 set-up and data analysis. The University of Oslo is acknowledged for providing laboratory facilities for molecular analyses.