Although experiments show a positive association between vascular plant and arbuscular mycorrhizal fungal (AMF) species richness, evidence from natural ecosystems is scarce. Furthermore, there is little knowledge about how AMF richness varies with belowground plant richness and biomass.
We examined relationships among AMF richness, above- and belowground plant richness, and plant root and shoot biomass in a native North American grassland. Root-colonizing AMF richness and belowground plant richness were detected from the same bulk root samples by 454-sequencing of the AMF SSU rRNA and plant trnL genes.
In total we detected 63 AMF taxa. Plant richness was 1.5 times greater belowground than aboveground. AMF richness was significantly positively correlated with plant species richness, and more strongly with below- than aboveground plant richness. Belowground plant richness was positively correlated with belowground plant biomass and total plant biomass, whereas aboveground plant richness was positively correlated only with belowground plant biomass. By contrast, AMF richness was negatively correlated with belowground and total plant biomass.
Our results indicate that AMF richness and plant belowground richness are more strongly related with each other and with plant community biomass than with the plant aboveground richness measures that have been almost exclusively considered to date.
Interactions between the above- and belowground components of plant communities are crucial for understanding biodiversity patterns and ecosystem function (Wardle et al., 2004). Most terrestrial plants form associations with arbuscular mycorrhizal fungi (AMF). Because AMF play a key role in plant nutrient uptake, they can influence both vascular plant community richness and primary productivity (Maherali & Klironomos, 2007; Smith & Read, 2008; van der Heijden et al., 2008). AMF sporulation and root colonization rates, which are taken as proxies for fungal biomass, can vary depending on host identity (Bever et al., 1996; Hart et al., 2013), and AMF, in turn, elicit a wide range of growth responses in host plant species (van der Heijden et al., 1998; Klironomos, 2003; Pringle & Bever, 2008; Uibopuu et al., 2012).
Patterns of soil microbial diversity are often explained by the ‘plant diversity hypothesis’, which proposes that greater plant diversity increases microclimatic variability and habitat complexity, for instance in soil structure and root architecture (Hooper et al., 2000; Waldrop et al., 2006). Although this concept is widely accepted for directly interacting organisms such as plants and AMF, macrocosm and field experiments have revealed both positive (van der Heijden et al., 1998; Vogelsang et al., 2006) and negative (Antoninka et al., 2011) relationships between AMF and plant richness. However, experimental investigations are often limited by the relatively small numbers of either fungal or plant taxa included, or are influenced by a sampling effect because those fungal species that are especially good at promoting plant growth or richness are more frequently studied (van der Heijden et al., 1999; Wardle, 1999). Surprisingly, data about whether and how AMF richness and plant richness are related in natural ecosystems is very scarce. Lekberg et al. (2013) found no relationship between AMF operational taxonomic unit (OTU) richness and overall plant richness in their study site. However, Landis et al. (2004) found a positive relationship between AMF spore richness in soil and aboveground plant richness (Landis et al., 2004). As spore identification can underestimate AMF taxa that do not establish in cultures or rarely sporulate (Sanders, 2004), it would be informative to test this relationship using molecular approaches that target the AMF structures actually present in plant roots. As far as we are aware, the only published attempts to correlate molecularly detected AMF richness and (conventionally detected) plant richness found no relationship at local (Öpik et al., 2008) or global scales (Öpik et al., 2010). Moreover, previous studies have related AMF richness to aboveground, rather than belowground plant richness. This potentially overlooks the significantly greater plant richness below- than aboveground, caused by dormant species, clonal species with extensive rhizome networks, and ephemerals (Hiiesalu et al., 2012). Most importantly, one might expect AMF richness to be more strongly correlated with below- than aboveground plant richness, because these fungi associate directly with roots. As AMF can be selective for the host plant species (Öpik et al., 2009; Davison et al., 2011), the higher the richness of plant roots, the greater the range of niches or habitats they would potentially have.
Related to the question of whether the richness levels of AMF and plant taxa are correlated in natural environments is the question of how they both vary along environmental gradients. Plant species richness is known to respond to net primary productivity (NPP), plant biomass and soil fertility. Aboveground plant richness frequently peaks at medium, but declines at high values of aboveground plant biomass (Grime, 1973; Grace, 1999; Waide et al., 1999; but see Adler et al., 2011), at least in temperate regions (Pärtel et al., 2007). It is notable that although the majority of plant growth (i.e. 50–90% of NPP) in cool and arid ecosystems, such as grassland, tundra, steppe and desert, occurs belowground (Stanton, 1988; Schenk & Jackson, 2002), it is not known if the plant diversity–productivity patterns described aboveground apply to the belowground part of plant communities. So far, only one study has related belowground plant richness to soil fertility, which is one of the main determinants of habitat productivity (Hiiesalu et al., 2012). That study found that the number of plant species detected belowground increased along the soil fertility gradient, in contrast to the negative response shown by aboveground plant richness. The decline in aboveground plant richness with increasing soil fertility can be explained by asymmetric competition, whereby tall plants gain a disproportionate share of the light resource (for their relative size) eventually excluding smaller plants from the community (Weiner, 1990). However, this reduction in plant species richness might not occur belowground because competition for soil resources is considered symmetric with regards to the size of the root systems (Cahill & Casper, 2000). In other words, soil resources can be acquired from all directions, as opposed to light, which comes only from above. Thus, competitive exclusion of smaller root systems by larger root systems might be less important in limiting species coexistence belowground. In addition, if plants excluded aboveground remain dormant belowground for a period of time (Pärtel et al., 2012), a pattern of increasing belowground plant richness with increasing soil fertility could be expected. While relatively more focus is placed on studying the relationships between plant diversity and productivity, plant biomass production is also likely to be associated with the diversity and species composition of AMF communities.
Very few studies have considered the richness of both plants and their associated fungi when examining relationships between plant community diversity and biomass (but see Klironomos et al., 2000; Koch et al., 2012). AMF can have a considerable impact on plant nutrient uptake, diversity and productivity (van der Heijden et al., 2008). Whether the relationship between AMF richness and plant biomass production is negative, positive or nonlinear might depend on the identity and functional attributes of the organisms involved, as well as the environmental conditions, especially the availability of limiting nutrients, under which they interact (Klironomos, 2003; van der Heijden et al., 2008). Theoretical studies have predicted a positive relationship between AMF diversity and plant biomass, and this has largely been borne out by controlled empirical experiments (van der Heijden et al., 1998; Lovelock & Ewel, 2005; Vogelsang et al., 2006; Maherali & Klironomos, 2007; Wagg et al., 2011a; Koch et al., 2012). However, some studies have recorded a negative effect of AMF presence on individual plant species' growth (Wilson & Hartnett, 1998; Klironomos, 2003), indicating that the relationship may not be straightforward. To the best of our knowledge, the associations between AMF richness, plant richness (including the richness of roots actually interacting with AMF as opposed to aboveground parts) and plant biomass have not been studied in a natural community.
In this study, we measured AMF species richness, above- and belowground plant species richness, and plant biomass in a native grassland at the northern edge of the North American Great Plains. AMF and belowground plant taxa were identified by 454-sequencing of DNA extracted from mixed root samples. Aboveground plant species were identified visually. First, we aimed to quantify species richness of coexisting AMF and plant taxa from mixed root samples. Second, we examined correlations between AMF and plant richness and plant biomass. We hypothesized that: (1) AMF richness is positively correlated with plant richness, and that this relationship is stronger in the case of belowground plant richness; (2) aboveground plant richness is negatively and belowground plant richness is positively correlated with several components of plant biomass (above-, belowground and total biomass); and (3) AMF richness is positively correlated with plant biomass.
Materials and Methods
Study site and sampling
We measured root-associated AMF richness and above- and belowground plant richness at White Butte near Regina, Saskatchewan, Canada (50°28′N, 104°22′W). Vegetation in the area is native mixed-grass prairie, dominated by Hesperostipa comata (Trin. & Rupr.) Barkworth, Carex duriuscula C.A. Mey and Bouteloua gracilis Vasey, with patches of the shrub Symphoricarpos occidentalis R.Br. (Pärtel & Wilson, 2002). The site harbours c. 75 vascular plant species (I. Hiiesalu, pers. obs.). The climate is continental with mean daily temperatures of −19°C in January and 18°C in July. Average annual precipitation is 384 mm, which mostly falls in May and June (Environment Canada, 1993).
Sampling locations at the c. 2-ha site were arranged contiguously along ten randomly placed 1-m-long transects (separated from one another by > 25 m), with ten locations per transect, resulting in a total of 100 sampling locations. Sampling was conducted at the end of June 2008, which corresponds to the peak of the plant growth period. Volumes of 10 × 10 × 10 cm were sampled above and below the soil surface at each sampling location, following Hiiesalu et al. (2012). This volume captures the scales at which herbaceous plants interact and is representative of the scales used to study plant aboveground richness in grasslands (Grace, 1999; Wilson et al., 2012). Also, the vertical component, that is, the 10 cm above or below the soil surface, represents a layer that captures the majority of grassland biomass both below- (Steinaker & Wilson, 2005) and aboveground (Kull & Zobel, 1991). To aid aboveground species identification and biomass collection, we delineated the vertical edges of the aboveground sample volume with 10 cm metal wires fixed at the lower corners of the sample volume. Aboveground plant species richness was determined by identifying all vascular plant species in each sample volume. This included species that were rooted in the samples, as well as the occasional species that occurred in the sample volume but was rooted outside (mean 0.5 species per sample). The biomass of shoots in the sample volumes was collected, litter removed, oven-dried at 70°C for 24 h and weighed. AMF richness and belowground plant species richness were measured from mixed root samples removed from soil volumes of 10 × 10 × 10 cm located directly below the corresponding aboveground samples. The litter layer was removed, roots were sieved from soil and dead roots were removed on the basis of colour and physical appearance (Gregory, 2006). Root biomass was dried in a freeze-drier and weighed. Subsequently, each root sample was crushed using liquid nitrogen, and thoroughly mixed. A root subsample of 100 mg dry weight was used for molecular analysis (100 samples in total).
Root DNA extraction and 454-sequencing of AMF DNA
Root subsamples were pulverized with 2.3-mm chrome-steel beads (BioSpec Products Inc., Bartlesville, OK, USA) in a Mixer Mill 301 (Retsch GmbH, Haan, Germany). DNA was extracted using the PowerSoil-htp™ 96 Well Soil DNA Isolation kit (Mo Bio Laboratories Inc., Carlsbad, CA, USA) with the modifications of adding 100 μl of sterile H2O to weighed roots and eluting in a final volume of 50 μl. Glomeromycota nuclear SSU rRNA gene sequences were amplified from root DNA extracts using the primers NS31 and AML2 (Simon et al., 1992; Lee et al., 2008). In order to identify reads originating from different samples, we used a set of 8-bp barcodes designed following Parameswaran et al. (2007). A two-step PCR procedure was conducted: in the first PCR reaction PCR primers were linked to barcodes and partial 454-sequencing adaptors A and B; in the second reaction the full 454-adaptors A and B served as PCR primers, completing the full 454-adaptor+barcode+PCR primer construct. Thus, the composite forward primer in the first PCR reaction was: 5′ GTCTCCGACTCAG (NNNNNNNN)TTGGAGGGCAAGTCTGGTGCC 3′; and the reverse primer: 5′ TTGGCAGTCTCAG (NNNNNNNN)GAACCCAAACACTTTGGTTTCC 3′, where the A and B adaptors are underlined, the barcode is indicated by N-s in parentheses, and the specific primers NS31 and AML2 are shown in italics. The 10 × diluted product of the first PCR reaction was used in the second PCR with 454-adaptors A (5′-CCATCTCATCCCTGCGTGTCTCCGACTCAG-3′) and B (5′-CCTATCCCCTGTGTGCCTTGGCAGTCTCAG-3′) serving as PCR primers. The PCR reactions were performed in a total volume of 10 μl containing 5 μl of HotStarTaq Master Mix (Qiagen), 0.2 μM each of the primers and 1 μl of template DNA. The reactions were run on a 2720 Thermal Cycler (Applied Biosystems, New York, NY, USA) under the following conditions: 95°C for 15 min; five cycles of 42°C for 30 s, 72°C for 90 s, 92°C for 45 s; 35 (first PCR) or 20 (second PCR) cycles of 65°C for 30 s, 72°C for 90 s, 92°C for 45 s; followed by 65°C for 30 s and 72°C for 10 min. PCR products were separated by electrophoresis through a 1.5% agarose gel in 0.5 × TBE, and the PCR products were purified from the gel using the QIAquick Gel Extraction kit (Qiagen) and further purified with Agencourt® AMPure® XP PCR purification system (Agencourt Bioscience Co., Beverly, MA, USA). DNA was quantified using a NanoDrop 1000 Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). A total of 250 ng of the resulting DNA mix was sequenced on a Genome Sequencer FLX System, using Titanium Series reagents (Roche Applied Science, Penzberg, Germany) at GATC Biotech (Konstanz, Germany).
454-sequencing of plant DNA
Plant chloroplast trnL (UAA) gene sequences were amplified from the same root DNA extracts as above using primers c and d (Taberlet et al., 1991), linked to 454-sequencing adaptors A and B, respectively. PCR was conducted in two steps, as described above. The composite forward primer in the first PCR reaction was: 5′-GTCTCCGACTCAG(NNNNNNNN)CGAAATCGGTAGACGCTACG-3′; and the reverse primer: 5′- TTGGCAGTCTCAG(NNNNNNNN)GGGGATAGAGGGACTTGAAC-3′, where A and B adaptors are underlined, the barcode is indicated by N-s in parentheses and specific primers c and d are shown in italics.
The PCR reactions were performed in a total volume of 10 μl containing 5 μl of Smart-Taq Hot Red 2 × PCR Mix (Naxo, Tartu, Estonia), 0.2 μM each of the primers and 1 μl of template DNA. The reactions were run on a Veriti® 384-Well Thermal Cycler (Applied Biosystems) with the same cycling conditions as described above. PCR products were purified and quantified as already described. A total of 250 ng of this DNA mix was sequenced on a Genome Sequencer FLX System as described above.
Bioinformatical analysis of AMF sequences
AMF 454-sequencing reads were included in subsequent analyses only if they carried the correct barcode and forward primer sequence, and were ≥ 199 bp long (including the barcode and primer sequence). Because 454-sequencing reads were shorter than the full amplicon we did not consider the reverse primer (AML2) here. Potential chimeras were identified using UCHIME (Edgar et al., 2011) in reference database mode using the default settings, and excluded from the analyses (2423 chimeras). After stripping the barcode and primer sequences, we used a closed reference operational taxonomic unit (OTU) picking approach (sensu Bik et al., 2012) to match obtained reads against taxa in the MaarjAM database of published Glomeromycota SSU rRNA gene sequences (http://maarjam.botany.ut.ee, Öpik et al., 2010; accessed 14 June 2013). The MaarjAM database contains representative sequences covering the NS31/AML2 amplicon from published environmental Glomeromycota sequence groups and morphologically described taxa. As of May 2013 it contained a total of 7618 records, of which 6064 records carried identities of 341 SSU sequence-based taxa, or so-called virtual taxa (VT cf. Öpik et al., 2009, 2014). 454-reads were assigned to VT by conducting a BLAST search (soft masking of DUST filter) against the reference database with the following criteria required for a match: sequence similarity ≥ 97%; an alignment length not differing from the length of the shorter of the query (454-read) and subject (reference database sequence) sequences by > 10 nucleotides; and a BLAST e-value < 1e−50. VT occurring only once in the dataset were omitted following our previously applied procedures (Öpik et al., 2009; Moora et al., 2011). In addition, only samples that yielded ≥ 10 reads were included in further analyses. A set of representative sequences were submitted to EMBL nucleotide collection (accession numbers HG780136–HG780302; the set contains up to three of the longest reads per VT).
Bioinformatical analysis of plant sequences
Plant 454-reads were quality-checked as already described, except that reads were included in subsequent analyses only if they were ≥ 170 bp long, including the barcode and primer sequences. We detected 299 chimeras as already described and excluded them from further analyses. After stripping the barcode and primer sequences, we identified plant 454-reads using the same BLAST-based OTU picking approach as used for analysis of AMF reads, but this time used a custom made trnL (UAA) intron sequence reference database. The database was compiled from three sources: (1) plants sampled at the study site and Sanger-sequenced (published in Hiiesalu et al., 2012); and sequences from species occurring in our study system or closely related taxa that were (2) available in the International Nucleotide Sequence Database (INSD); or (3) generated by the EcoChange Project (EU FP6 Integrated Project EcoChange). Plants collected from our study system and its surroundings were identified and stored as vouchers at the George F. Ledingham Herbarium (USAS), University of Regina, Canada. Only samples which yielded at least 10 sequences and MOTUs that occurred more than once were included in further analyses.
Because the trnL (UAA) intron sequence between c and d primers does not distinguish certain closely related plant species, we defined molecular operational taxonomic units (MOTUs) within our reference database by grouping species that exhibited sequence similarity of ≥ 97% using the BLASTclust algorithm (ftp://ftp.ncbi.nih.gov/blast/).
Twenty-three species recorded aboveground were assigned to the corresponding MOTUs in order to use the same taxonomic resolution above- and belowground. Consequently, a few groups of closely related species were grouped into single MOTUs (Solidago missouriensis Nutt., Heterotheca villosa (Pursh) Shinners and Erigeron glabellus Nutt. were merged, Elymus trachycaulus (Link) Gould ex Shinners, E. smithii (Rydb.) Gould. and Agropyron cristatum (L.) Gaertn. were merged, Koeleria macrantha (Ledeb.) Schult. and Trisetum spicatum (L.) K. Richt. were merged, and Artemisia campestris L., A. frigida Willd. and A. ludoviciana Nutt. were merged). This approach slightly underestimates the actual plant species richness, but at the same time it standardizes plant taxonomic resolution above- and belowground, thus making plant richness estimates comparable to each other (Hiiesalu et al., 2012). Plant MOTUs are hereafter referred to as species or species groups. A set of representative sequences was submitted to EMBL nucleotide collection (accession numbers HG780303–HG780356; the set contains up to three of the longest reads per MOTU).
The strengths and limitations of molecular plant root identification by 454 sequencing of the trnL amplicon of c and d primers as applied in this study are discussed in our earlier papers (Hiiesalu et al., 2012; Pärtel et al., 2012).
We considered two aspects of plant species richness: aboveground richness, sampled visually, and belowground richness (aboveground richness of species rooted in the sample plus additional belowground richness comprising species only detected on the basis of DNA sequences). Two plant species occurred in aboveground samples but were not detected belowground: Psoralea argophylla Pursh (present in five samples) and Erysimum asperum (Nutt.) DC. (three samples). Nonrecovery may have been because of amplicon competition during mixed-species PCR, whereby the DNA of some taxa is amplified more efficiently than that of others (Schlichter & Bertioli, 1996). In order to have directly comparable MOTU richness estimates above- and belowground, we omitted these species from the analysis, following Hiiesalu et al. (2012). Both richness and biomass values were log10-transformed before analysis.
In order to examine the relationships among AMF and plant species richness above- and belowground, and their relationships with total, above- and belowground plant biomass measures, we used partial correlation. This technique measures the relationship between two parameters while controlling the potential effect of other measured variables. This allows spurious correlations (i.e. correlations explained by the effect of other variables) to be avoided and hidden correlations (i.e. correlations masked by the effect of other variables) to be revealed. In addition we used this technique to account for the potential effect of spatial autocorrelation. The partial correlation between variables x and y is a correlation between the residuals of x and y from linear regressions against other potentially influencing variables. In the regressions (R package nlme, function lme, using Maximum Likelihood ‘ML’; R Development Core Team, 2013), transect was specified as a random factor to account for spatial autocorrelation, and other variables besides x and y that were relevant for the comparison were included in the model. For instance, when we correlated AMF and belowground plant richness, we added total plant biomass to the model to account for its potential effect on x and y. If two variables were not independent (e.g. belowground plant richness and aboveground plant richness, or total plant biomass and belowground biomass, as one includes the other), only the more complete measure (i.e. belowground plant richness and total plant biomass) was included. We used Pearson correlation in our analyses. Instead of presenting raw values of x and y, we plotted their residuals in order to account for the effects of other variables. We used a Type II regression line (R package lmodel2, functions lmodel2; R Development Core Team, 2013) to display the linear relationship between variables if the correlation was significant (P <0.05). In these graphs the selection of x and y axes is arbitrary (there are no dependent and independent variables), and it does not influence the relationship. To illustrate the relationship between above- and belowground plant richness, we plotted untransformed richness data without testing for correlation because in our calculations aboveground richness is theoretically a subset of belowground richness.
In order to test whether our results were dependent on between-sample variation in the number of sequence reads, we repeated all analyses with rarefied data (Supporting Information Methods S1). Standardising sequencing depth per sample by rarefying to the median sample size in the dataset did not significantly change any of the correlations (Table S1), although the correlation coefficients tended to be slightly lower due to the lower statistical power of the analysis (Figs S1–S3).
We recovered 20 321 quality-filtered AMF SSU rRNA gene reads from root samples (minimum: 23, mean: 303, maximum: 1037 sequences per sample). The reads were assigned to 63 AMF virtual taxa (VT) in five families: Archaeosporaceae (1 VT), Claroideoglomeraceae (3), Diversisporaceae (3), Glomeraceae (57) and Paraglomeraceae (1) (Table 1). The five most dominant AMF VT were taxa formerly recorded from various continents including North America and various habitat types including grasslands: Glomus VT 177, 113, 212, 155 and 83 (in the order of sequence numbers; distribution range data from MaarjAM database, status 5 September 2013; Öpik et al., 2010). We detected on average 12.0 AMF VT (minimum 1, maximum 29) per mixed-root sample.
Table 1. Detected arbuscular mycorrhizal (AM) fungal SSU rRNA gene virtual taxa (VT) based on the nomenclature of the MaarjAM database (status 14 June 2013)
INSD acc. no. of the VT type sequence
Total no. of sequences
VT are referred to as taxa or species in the text. Total counts of 454-sequences matching taxa in the MaarjAM database are reported for each VT.
We recovered 11 740 quality-filtered plant chloroplast trnL intron reads from root samples (minimum: 10, mean: 165, maximum: 763 sequences per sample). These reads were assigned to 19 molecular operational taxonomic units (MOTUs) in eight families (Table 2): Asteraceae (3 MOTUs), Boraginaceae (1), Cyperaceae (3), Elaeagnaceae (1), Fabaceae (1), Onagraceae (1), Poaceae (8), Rubiaceae (1). The most frequently encountered MOTUs were Koeleria-Trisetum group, Thermopsis rhombifolia, Erigeron-Heterotheca-Solidago group, Galium boreale and Calamovilfa longifolia (Table 2). Three species found belowground were not found in aboveground samples: Carex duriuscula, Poa pratensis L. and Lithospermum incisum Lehm. These taxa are all present in the local species pool of the study site (S. D. Wilson, pers. obs.). Belowground plant richness was on average 5.0 MOTUs per sample (maximum: 9), whereas aboveground plant richness was on average 3.4 MOTUs per sample (maximum: 5). Aboveground plant richness generally corresponded to belowground richness, and belowground richness was on average 1.5 times higher than that aboveground (Fig. 1).
Table 2. Plant molecular operational taxonomic units (MOTUs – referred to as species or species groups in the text) detected using the trnL (UAA) intron
INSD accession no.
Total no. of sequences
International Nucleotide Sequence Database (INSD) accession numbers of reference sequences generated in this study are highlighted in bold. Total counts of 454-sequences matching the reference sequences are reported for each MOTU.
AMF VT richness was positively correlated with both above- and belowground plant MOTU richness (r =0.28, P =0.03 and r =0.48; P <0.001, respectively; Fig. 2a,b).
Aboveground plant MOTU richness was positively correlated with belowground plant biomass (r =0.27, P =0.03; Fig. 3a), but not with aboveground or total plant biomass. Belowground plant MOTU richness was positively correlated with both belowground and total plant biomass (r =0.32, P =0.01 and r =0.31, P =0.014 respectively; Fig. 3b,c). By contrast, AMF VT richness was negatively correlated with belowground and total plant biomass (r =−0.29, P =0.023 and r =−0.31, P =0.014, respectively; Fig. 4a,b). None of the richness measures were significantly correlated with aboveground plant biomass.
AMF and plant taxon richness
By utilizing 454 sequencing of mixed root samples we were able to identify root colonizing AMF and belowground plant taxa coexisting in a natural grassland system. Overall, we recorded high fungal and plant taxon richness from mixed root samples. Our results also provide evidence that traditional measures of aboveground plant diversity underestimate by c. 50% the number of coexisting grassland plant species at the scale of 0.01 m2 (Hiiesalu et al., 2012). We detected three plant species belowground which were not detected in aboveground samples most probably due to clonality (e.g. Carex and Poa can form extensive belowground rhizome networks with only a few aboveground shoots) or dormancy (Lithospermum was not detected from the entire site in the year of sampling). However, it is conceivable that a greater sequencing depth for both AMF and plants might have yielded even higher belowground richness estimates. Our total richness values for both plants and fungi are comparable to those of earlier studies (Davison et al., 2011; Dumbrell et al., 2011; Saks et al., 2014). Thus, any richness underestimation, if present, seems to have been moderate.
We detected 63 AMF taxa at the study site, which is among the highest AMF richness values recorded in a natural plant community of comparable size. When comparing this number with earlier data, it is important to recognize methodological differences that may affect AMF richness estimation. First, very few studies have used mixed root samples for AMF identification (Heinemeyer & Fitter, 2004; Dumbrell et al., 2011). More commonly, roots of individual plants are sampled (cf. Öpik et al., 2010). Because coexisting plant species can differ in their associated AMF assemblages (Vandenkoornhuyse et al., 2002; Davison et al., 2011), mixed root samples potentially allow the detection of those AMF that are specific to infrequent plant species and would be overlooked if only the most common plants at a site are considered. Second, 454-sequencing methodology typically yields orders of magnitude higher numbers of sequences per sample (sequencing depth) than cloning-Sanger sequencing, resulting in higher detected richness values (Öpik et al., 2009, 2013). Dumbrell et al. (2011) used a similar approach – 454-sequencing of AMF SSU rDNA amplified from mixed root samples from limestone grassland in the UK – and detected 70 AMF taxa. However, Dumbrell et al. (2011) sampled 11 times through the year, detecting some taxa in the cool or warm season only, whereas we sampled once, in summer. Third, it is important to consider whether AMF are sampled in plant roots or soil, because some AMF taxa may be present in soil in the form of extra-radical mycelia and active or dormant spores, but not present in colonized roots (Hempel et al., 2007). Indeed, an even higher number of AMF VT (83) was detected in an Estonian boreo-nemoral forest where both soil and root samples of six plant species were analyzed (Saks et al., 2014). Fourth, AMF richness estimates can vary depending on the marker region used and the MOTU delimitation principles applied (Kõljalg et al., 2013; Lindahl et al., 2013; Kohout et al., 2014; Öpik et al., 2014). Overall, our results indicate that next-generation sequencing of mixed root samples together with surveys of traditional aboveground vegetation enables a more complete estimate of the biodiversity of coexisting plant and AMF taxa.
Relationship between AMF and plant taxon richness
AMF richness was positively associated with measures of plant richness, with the relationship involving belowground plant richness stronger than the one involving aboveground plant richness. A few experiments (van der Heijden et al., 1998; Vogelsang et al., 2006) and one observational study from a natural community, in which AMF were identified on the basis of spore morphology (Landis et al., 2004), have found positive relationships between AMF and host plant richness. However, earlier field studies identifying AMF on the basis of DNA sequences have found no relationship between AMF MOTU and vascular plant richness (Öpik et al., 2008; Lekberg et al., 2013). The studies described above all measured only aboveground plant species richness when examining the association between AMF and plant richness. Data on belowground plant richness is lacking, mostly due to methodological constraints related to the identification of plant roots (Mommer et al., 2011; Pärtel et al., 2012). When sampling only shoots, up to half of the plant species coexisting at small spatial scales may be overlooked (this study and Hiiesalu et al., 2012). Such plant species, which exhibit spatial or temporal variability in their aboveground appearance, may nonetheless associate with AMF, and as indicated by our results, AMF richness can be more strongly linked with below- than aboveground plant richness. Our results suggest that greater richness in one of the symbiotic partners may promote richness in the other. Alternatively, the richness of both partners could be positively influenced by some other variable, for example, an environmental factor.
At the studied scale, possible mechanisms causing fungal and plant richness to co-vary in space include host specificity, AMF functional trait complementarity, and mycelial networks shared among plants (reviewed by Hart et al., 2003). Because our study was observational, we cannot determine whether changes in AMF richness caused changes in host plant richness or vice versa (Sanders, 2004). Our results are in agreement with the ‘plant diversity hypothesis’ which suggests that the higher the richness of plants, the more niches are available for microorganisms (e.g. thanks to the diverse soil structure, root architecture etc.) and the more likely are microbes to find a suitable host (Hooper et al., 2000; Waldrop et al., 2006). Recent evidence shows that preferential associations can occur between certain ecological groups of AMF and plants (Öpik et al., 2009; Öpik & Moora, 2012; Lekberg et al., 2013), such that the probability of finding a preferred host may be important in generating patterns of covariation. Moreover, both AMF and plants differ in functional traits, meaning that some combinations of fungi and hosts may be more mutually beneficial than others (Klironomos, 2003; Kiers et al., 2011; Hart et al., 2013). In addition, higher AMF richness might allow the plant community to more completely acquire available resources through AMF functional trait complementarity (Koide, 2000; Maherali & Klironomos, 2007; Chagnon et al., 2013), resulting in reduced competition between plant species (Wagg et al., 2011b) and, consequently, in increased coexistence. Lastly, common mycelial networks shared by neighbouring plants might promote species coexistence by equalizing the distribution of soil resources among dominant and subdominant plant species in the community (Hart et al., 2003). These are just some possible mechanisms that could contribute to a positive association between AMF richness and plant richness. Further experimental research is needed to test these mechanisms more explicitly.
Relationship between plant richness and plant biomass
We found that belowground plant richness increases with increasing belowground and total plant biomass, which contrasts with the general plant diversity–productivity theory based on observations of plants aboveground. Belowground plant richness has never before been related to plant community biomass measures, mostly due to the technical difficulty of assigning roots to species (Mommer et al., 2008, 2010; Frank et al., 2010; Kesanakurti et al., 2011; Hiiesalu et al., 2012). Moreover, aboveground plant richness has typically been related to aboveground biomass, but only rarely to total (Tilman et al., 2001) or belowground plant biomass (Liira & Zobel, 2000), despite the fact that greater biomass occurs below- than aboveground in temperate and arid ecosystems (Schenk & Jackson, 2002). Although there are significant fluctuations in aboveground plant biomass in this prairie system (Wilson, 2007), the lifespan of roots may be as long as 5 yr (Milchunas, 2009). In order to fully capture the seasonal dynamics of plant biomass fluctuation, the observed relationship between plant species richness and biomass needs to be tested by repeated sampling in time.
The positive association between belowground plant richness and biomass supports the idea that plant competition belowground is symmetric with respect to the relative size of the root systems involved (Weiner, 1990; Lamb & Cahill, 2008) and therefore probably not crucial in shaping belowground plant communities (Price et al., 2012). The decline in aboveground plant richness with increasing plant biomass is usually attributed to competitive exclusion resulting from asymmetric competition, whereby taller species have a disproportionate advantage over smaller species, and the latter are excluded. Roots can grow preferentially into fertile patches, and plant species may be more equal in their ability to acquire soil resources that are accessible from all directions, in comparison to access to light, which only comes from above (Cahill & Casper, 2000). This potentially results in greater overlap of individual root systems than of individual shoot canopies. Even when smaller plant species are excluded aboveground by competition for light, some belowground meristems are able to remain dormant and hidden for years or even decades (Shefferson, 2009; Reintal et al., 2010). Our results complement those of a previous study where soil fertility – rather than plant biomass which is used as a proxy of fertility in this study – was measured and shown to be positively related to plant species richness belowground (Hiiesalu et al., 2012).
An alternative mechanism that might underlie the positive relationship between belowground plant richness and belowground and total plant biomass is a complementarity effect whereby communities with high plant richness are able to access resources more completely and exhibit higher net primary productivity than less diverse communities. Such a positive relationship has been reported previously for aboveground plant richness and biomass in experimental systems (Tilman et al., 1996, 2006). However, belowground plant species richness can be considered a more complete measure of plant community richness, as it includes species that were absent aboveground at the time of sampling (Hiiesalu et al., 2012).
Relationship between AMF richness and plant biomass
Contrary to our expectations, AMF richness in roots was negatively correlated with both belowground and total plant biomass. This result contradicts some earlier experiments reporting positive effects of AMF richness on plant biomass (van der Heijden et al., 1998; Vogelsang et al., 2006; Wagg et al., 2011a; Koch et al., 2012). Some plant–AMF combinations are known to elicit negative plant growth responses (Wilson & Hartnett, 1998; Klironomos, 2003), but it is hard to predict how these would affect community-level fungal richness–plant biomass relationships. The direction of these relationships might also depend on the seasonal dynamics of AMF community composition and of plant biomass production. Experiments using radioactive phosphorus (P) isotopes have shown that AMF colonization can reduce direct P uptake by plant roots, replacing it with uptake via the fungus and eliciting a reduction in plant biomass (Smith et al., 2011). It has been proposed that if P uptake by AMF does not compensate for root uptake, plants could suffer from P deficiency, explaining the decline in plant biomass production (Grace et al., 2009). It might also be the case that if an AMF community in plant roots is diverse and includes less cooperative fungal taxa that mainly act as carbon sinks, then AMF may not produce a positive growth response in plant individuals (Johnson et al., 1996; Hart et al., 2003, 2013; Kiers et al., 2011). Therefore, future experiments should include more than just a few fungal taxa in order to explore the effects of AMF on plant growth and coexistence.
Alternatively, the observed negative relationship between AMF richness and plant biomass could have been mediated by the effects of soil nutrient concentrations (e.g. phosphorus) on plant biomass and AMF richness. It has been reported previously that AMF spore richness (Egerton-Warburton & Allen, 2000), evenness (Eom et al., 1999) and plant carbon allocation to AMF structures (Johnson et al., 2003) decrease with increasing soil fertility. Furthermore, AMF biomass has been shown to decrease with increasing plant biomass (Hedlund et al., 2003). Although we did not measure soil nutrient concentrations and fungal biomass it seems plausible that a negative association between AMF richness and plant biomass may have arisen due to higher soil fertility in sites where plants allocated less carbohydrate to fungal structures. The resulting increase in plant belowground biomass in such sites and reduced dependence on the symbiont could cause roots to compete with fungi for soil resources (Schnepf et al., 2008; Smith et al., 2011), thus reducing AMF biomass and richness.
We recorded the relationships between plant richness, the richness of their associated fungi and plant productivity using 454-sequencing of both AMF and plants in mixed root samples from a natural grassland system. We detected a positive relationship between AMF and plant richness, suggesting that richness in one group could promote richness in the other. Belowground plant richness was positively related to belowground and total plant biomass, while aboveground plant richness did not vary in relation to aboveground plant biomass. This suggests that with increasing plant biomass, more plant species are able to persist belowground. By contrast, AMF richness was negatively related to belowground and total plant biomass, which could be explained either by a negative effect of AMF on plant growth or by the relative unimportance of AMF to plants in more productive conditions. Our results from a natural ecosystem suggest that belowground components encompassing plants and their associated fungi can be more important than aboveground factors in plant community functioning.
This research was funded by grants from the Estonian Science Foundation (9050, 9157, 8323, 8613); targeted financing (SF0180098s08, SF0180095s08); grants IUT 20-27 and IUT 20-28; the European Regional Development Fund (Centre of Excellence FIBIR), Natural Sciences and Engineering Research Council of Canada and Environmental Protection and Technology R&D programme project ERMAS. This article was written during implementation of the project ‘Integration of the experimental and population biology using new methods of interdisciplinary issues – the way to excellence with young scientists’, Reg. no.: CZ.1.07/2.3.00/30.0048, which is funded by the European Social Fund (ESF) and the state budget of the Czech Republic through the Operational Programme Education for Competitiveness (OPEC). We thank Yoann Le Penize-Pinguet for comments on the manuscript, Dr Mary Vetter for help with plant identification, Jonathan Misskey for assistance in the field, and three anonymous reviewers for constructive comments on the manuscript. Sample preparation for 454-sequencing was performed by BiotaP Ltd (Tallinn, Estonia).