Introduction

Expansion of urban and suburban land is at the expense of natural forests, wetlands, and agricultural land (Robinson et al. 2005). Continued urbanization is anticipated (McDonnell et al. 1997) along with a growing proportion of the global human population living in cities and suburbs (Sadik 1999; Pavao-Zuckerman 2008). Though urban ecology studies have focused on urbanization effects on animals (Ditchkoff et al. 2006) and plants (Lundholm and Marlin 2006; Neil and Wu 2006), there has been little emphasis on analysis of urban fungi (Newbound et al. 2010). Weiher (2007) calls attention to the bias of ecological restoration studies toward plants and animals, with a considerably lower proportion targeting fungi and soil microbes (less than 3% of papers reviewed).

The diversity of EMF communities in natural systems is impressive, and most studies have shown little overlap in species assemblages between replicate plots or sites (Horton and Bruns 2001; Taylor 2002). Mycorrhizal inoculum is pervasive in undisturbed soils (Allen et al. 2003). However, severely disturbed soils could be characterized by reduced mycorrhizal inoculum or altered fungal diversity (Newbound et al. 2010) and a characteristic urban EMF suite, much like many plant species ubiquitously encountered across disturbed sites (Del Tredici 2010). For example, high levels of N deposition have been found to decrease fungal abundance and change community composition (Baxter et al. 1999; Lilleskov and Bruns 2001; Lilleskov et al. 2002). Cousins et al. (2003) found a lack of arbuscular mycorrhizal fungal spores in soils of unvegetated urban lots, and Bainard et al. (2011) found reduced mycorrhizal colonization of urban trees compared to rural trees. Parsons et al. (1998) cite the lack of mycorrhizal associations on tree seedlings planted on a closed landfill where only three ectomycorrhizal morphotypes were found.

This study used a soil bioassay and morphological and molecular methods to examine the ectomycorrhizal fungal (EMF) communities associating with Quercus rubra (northern red oak) seedlings in soils of seven sites: five urban sites of varying disturbance severity and two less disturbed forest reference sites. We test the hypotheses that the less disturbed reference site soils exhibit higher EMF colonization of seedling root systems, higher EMF species richness, and different EMF species composition than the disturbed site soils.

Methods

Study sites

Seven sites differing in degree of anthropogenic disturbance and surrounding urbanization were used. The two least disturbed sites were mature (>170 years old) forests: Kilmer Woods (REF1) in Piscataway, Middlesex County, New Jersey (latitude 40°31′4″ N, longitude 74°26′23″ W) and Helyar Woods (REF2) in New Brunswick, Middlesex County, New Jersey (latitude 40°28′33″ N, longitude 74°25′17″ W). In this metropolitan area, these sites are among the closest extant systems to undisturbed forest sites. We call these two sites “reference sites” because they are representative of local mature woodlands and their biotic and abiotic structure could serve as a target for forest restoration. Sites characterized by intermediate disturbance were Duke Farms (URB1), a post-agricultural woodland in Hillsborough, Somerset County, New Jersey (latitude 40°33′9″ N, longitude 74°37′25″ W), and Greenbelt (URB2), a remnant woodland patch adjacent to a plant nursery operation in Staten Island, New York (latitude 40°35′46″ N, longitude 74°10′50″ W). A third disturbed site was a former arsenal/superfund site in Edison, Middlesex County, New Jersey (latitude 40°30′41″ N, longitude 74°21′22″ W; URB3). The most urban sites were a recreational park with a fabricated substrate of rubble and excavation material in Brooklyn, New York (latitude 40°35′2″ N, longitude 73°59′34″ W; URB4) and a sparsely vegetated lot adjacent to an oil refinery in Bayonne, Hudson County, New Jersey (latitude 40°39′34″ N, longitude 74°6′5″ W; URB5), in which the soil was obviously contaminated with oil. Quercus individuals were found at five study sites, but not at URB4 and URB5. Collectively, the disturbed sites are referred to as “urban.” A summary of site characteristics and soil properties is provided in Table 1.

Table 1 Selected study site soil and vegetation properties, distance from reference sites, and time since known major soil disturbance
Full size table

Though the majority of woodlands in this metropolitan region have been logged several times over, our urban study sites, if restrictive anthropogenic influences were absent, would likely sustain oak-dominated woodlands. Uplands in northern and central New Jersey are dominated by mixed oak forest (Collins and Anderson 1994), and before European settlement, even the uplands of the island of Manhattan were dominated by oak woodlands (Sanderson 2009). Post-agricultural succession in this area typically leads to oak-dominated forests, with the Duke Farms site (URB1) being such an example.

Experimental design

Soil collection

To bioassay the EMF communities, soil was collected from each of the seven sites in April 2008. Using a drain spade shovel rinsed with a 10% bleach solution after collection at each site, 15 soil cores approximately 15 cm in depth and diameter were removed from the ground. Cores were randomly collected from five points along three parallel transects, each transect being 20–25 m long and 1 m apart (total sampling area of approximately 50 m2). Soil cores were placed into sterilized standard plastic nursery pots 15 cm in height and diameter while keeping the vertical structure of the cores intact.

Although Taylor and Bruns (1999) showed that EMF bioassay experiments in which soil is removed from field sites and the experiment is performed in pots selects for the resistant fungal propagule community, Avis and Charvat (2005) showed that keeping the soil vertical profile intact during collection more accurately reflects field conditions and helps to retain field similarities in the resulting soil EMF community. Bioassays conducted in such a way provide mycorrhizal inoculum that could come from spores, sclerotia, soil hyphae, and colonized root fragments (Avis and Charvat 2005).

Bait plants

Q. rubra was chosen as a host species as its distribution includes virtually all of eastern North America (USDA NRCS 2007), and it is common at our reference sites. Since most EMF species form mycorrhizal associations with a broad range of host tree species, usually within a family (Horton and Bruns 1998), EMF found in association with Q. rubra are likely to associate with other members of Fagaceae.

Q. rubra acorns (Sheffield’s Seed Co., Inc., Locke, NY, USA; sourced in Michigan) were surface-sterilized with a 10% bleach solution, cold-stratified for 60 days in moistened peat moss, and germinated on steam-sterilized potting soil. Upon the first flush of leaves, one seedling was transplanted into each pot of collected soil, 15 pots of soil from each study site (105 pots), within 3 days of soil collection. Seedlings were also transplanted into three pots of steam-sterilized potting soil to check for EMF contamination in the greenhouse. Seedlings were maintained in the greenhouse and watered with tap water three times weekly for 26 weeks. Newton and Pigott (1991) showed EMF colonization rates of approximately 40–50% on oak seedlings planted in pots of collected field soil after 20 weeks. We allowed a few more weeks to approximate the length of the local growing season. Pots were arranged randomly and periodically rotated to minimize effects of bench position. Whole intact root systems were harvested from each soil pot and stored in individually sealed plastic bags at 4°C. Ten root systems from each site were randomly selected for EMF sampling.

EMF sampling and analysis

EMF sampling

Each root system was subsampled for EMF colonization within 3 weeks of harvest. Root systems were gently rinsed free of adhering soil by submersion and slight agitation in water. Fine roots were clipped from the root system into fragments 1 to 7 cm in length. Effort was made to clip roots from all around the root system at varying distances from the root collar. Root fragments were placed into three gridded petri dishes filled with water until fragments became crowded and overlapped.

Each petri dish contained 36 1.3 × 1.3 cm squares, arranged in six rows by six columns. In each dish, one square per row was randomly selected and root tips within that square were sampled for EMF colonization. Root tips (×6 to ×50 magnification) that were visibly colonized by EMF were classified as “colonized” and bare tips as “uncolonized.” Each EMF root tip was counted and classified into a morphotype as per the criteria of Agerer (1987–1996) based on external characteristics. Effort was made to overestimate morphotype diversity based on minor differences. For each root system sample, at least two (if available) representative EMF root tip samples of each morphotype were collected and stored in 2× CTAB in separate sterile 1.5 mL microcentrifuge tubes for subsequent molecular analysis.

EMF colonization was measured as a percentage by dividing the number of EMF-colonized root tips by the number of colonized plus uncolonized root tips. Measures of EMF species richness and community composition were based on subsequent molecular analysis.

DNA extraction and PCR amplification

DNA was extracted from EMF root tips representing each morphotype from each seedling root system using the CTAB extraction protocol (Gardes and Bruns 1993). Polymerase chain reaction (PCR) was performed in a GeneAmp 9700 thermal cycler (Applied Biosystems, Foster City, CA, USA) according to the methods of Gardes and Bruns (1993), with the following modifications: A PCR mastermix (GoTaq Colorless Mastermix, Promega Corporation, Madison, WI, USA) was used along with the primers ITS1F (Gardes and Bruns 1993) and ITS4 (White et al. 1990) to amplify the internal transcribed spacer (ITS) ribosomal DNA region. Samples that did not successfully amplify were noted, and replicates (EMF root tips of the same morphotype from the same root system) were extracted and run through PCR in the same way until amplification was successful or until no more replicate samples were available.

RFLP analysis and DNA sequencing

Successfully amplified samples were digested for restriction fragment length polymorphism (RFLP) using the restriction enzymes HinfI and DpnII (New England Biolabs, Beverly, MA, USA). Fragments were separated using agarose gel (3%) electrophoresis, visualized under UV light, and photographed on a ChemiDoc XRS molecular imager (Bio-Rad Laboratories, Hercules, CA, USA). Band patterns were visually analyzed, and samples with identical band patterns with both enzymes were grouped together into RFLP types. The ITS regions of representative samples of each RFLP type were sequenced at Cornell University’s DNA Sequencing Facility using an Automated 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA). Each ITS sequence was entered into a BLAST (NCBI) query to provide records of similar sequences in GenBank (Altschul et al. 1997). RFLP types whose sequences were at least 97% similar along at least 90% of the sequence length in a pairwise alignment comparison were combined under one taxon type (O’Brien et al. 2005). Taxonomic identities were assigned to each RFLP type based on affinity to BLAST records as follows: greater than or equal to 97% similarity = species level match, 95–96% similarity = genus level match, 90–94% similarity = family level match, and less than 90% similarity = order level match. Root tip morphology, in addition to sequence data, was used to identify Cenococcum sp. Some RFLP types failed to yield sequence data and were not given a taxonomic identity. DNA sequences from this study have been submitted to GenBank (accession numbers GU907781–GU907810).

Taxonomic (or RFLP type) identities were retroactively assigned to EMF morphotypes from each root system. From here, these identities will be referred to as separate EMF “species,” as they are all less than 97% similar to each other in ITS sequence comparisons. Morphotypes that failed to amplify were included in percent colonization counts but were excluded from community-level measures (species richness and community composition). RFLP typing was conducted on up to three tips for any particular morphotype within a root system. When RFLP analysis revealed more than one genetic type, the taxonomic identity was assigned equally among the tips in that morphotype.

Statistical analysis

Differences in EMF colonization and EMF species richness between site soils were determined by ANOVA (JMP®, version 8, SAS Institute Inc., Cary, NC, 1989–2009) followed by the post hoc Tukey–Kramer HSD test. Analysis of similarity (ANOSIM), devised by Clarke (1993), was performed using the program PRIMER v6 (Clarke and Gorley 2006) to compare EMF community composition among site soils. ANOSIM deals with non-parametric data and is especially geared toward species counts with multiple zero values. It tests the null hypothesis that there are no differences in community composition among sites. In this study, this null hypothesis is rejected at p < 0.002 (alpha value of 0.05 with Bonferroni correction for multiple comparisons). Community similarity was analyzed among site soils by aggregating EMF species into genera and creating a resemblance matrix using the Bray–Curtis similarity index on presence/absence data. Using generic identities is appropriate because sequences of many fungal species are not available in GenBank. When a list of species was retrieved with a BLAST query and all were of the same genus, we were confident of the generic identification. PRIMER v6 was also used to create a cluster diagram (CLUSTER) based on common EMF species presence/absence in each site soil.

Results

EMF sampling

In 11 cases, RFLP patterns could not be determined for a morphotype within a sample due to failed or weak amplification. This resulted in 35 EMF tips classified as “unknown.” Fifty-four different RFLP pattern types were identified. EMF tips from the same root system classified into different morphotypes often revealed identical RFLP patterns. No colonization was observed on seedlings maintained in sterilized potting soil.

After combining RFLP types with at least 97% similarity over at least 90% of their ITS sequence, 34 distinct EMF “species” remained. Comparisons of RFLP types found in this study to records in GenBank are shown in Table 2. Four RFLP types failed to yield sequence data, and so their original RFLP type numbers were retained as their designations (RFLP types 16, 23, 39, 43).

Table 2 DNA sequence analysis results of the ITS region from RFLP types based on GenBank BLAST searches
Full size table

EMF colonization

An ANOVA comparing the percentage of EMF-colonized root tips showed significant differences among site soils (F [6,63] = 10.445, p < 0.0001). Seedlings planted in URB4 and URB5 soils had significantly lower root colonization rates than those in the other five site soils (Tukey–Kramer HSD, α = 0.05; Fig. 1). EMF colonization in URB3 soil was distinctly bimodal, with root systems having either 30–60% colonization or 0–10% colonization.

Fig. 1
figure 1

Mean percent colonization of host plant root systems per site soil (n = 10). Error bars represent standard error; sites with different letters are significantly different (ANOVA, Tukey–Kramer HSD, α = 0.05)

Full size image

EMF species richness

Significant differences were found with an ANOVA comparing the number of EMF species present per seedling root system among site soils (F [6,63] = 13.02, p < 0.0001). An almost 17-fold gradient in richness is shown among sites (Fig. 2). On average, reference site soils had 3.0 EMF species per root system and urban site soils had 1.2 species per system, a significant difference (Student’s t test, p < 0.001).

Fig. 2
figure 2

Mean number of EMF species per host plant root system per site soil (n = 10). Error bars represent standard error; sites with different letters are significantly different (ANOVA, Tukey–Kramer HSD, α = 0.05)

Full size image

Site soils differed in total number of observed EMF species and jackknife estimates of species richness, with the greatest number of species found in REF1 soil (Table 3). Of the 34 total species found, 47% occurred exclusively in the reference site soils, 32% occurred exclusively in the urban site soils, and 21% occurred in both reference and urban site soils.

Table 3 Number of observed EMF species and first- and second-order jackknife estimates of total EMF species richness per site soil
Full size table

EMF community composition

The ANOSIM test indicated significant differences in EMF community composition on host roots between 11 site soil pairwise comparisons. Ten site pairs did not show significantly different EMF communities. In comparison to each other, the two reference sites showed non-significant differences in EMF community composition but were significantly different from all but one urban site (URB3). Several urban sites did not significantly differ from each other. Figure 3 graphically depicts EMF community similarity relationships between site soils.

Fig. 3
figure 3

Graphical depiction of ANOSIM results. Lines connecting sites represent non-significant differences (p > 0.002) in EMF community composition (presence/absence of genera) between those site soils

Full size image

A cluster analysis of EMF species based on their presence/absence in each of the seven site soils indicates species assemblages (Fig. 4). Since sites were defined a priori, this analysis serves only to show affinities between EMF species in terms of common site soils in which they occurred. The analysis shows separation first of a group of EMF found primarily in URB2 soil and then splits the remainder of EMF species between those found only in REF1 and REF2 and those found among several study site soils. The second cluster, containing EMF species Scleroderma citrinum, Pezizales sp., Tuber sp. 4, Thelephora terrestris, Tuberaceae sp., Scleroderma sp., and Cenococcum sp., represents species that were found broadly across the seven site soils. Relative abundance of each EMF species as a percentage of total colonized root tips across replicate seedlings in each site soil is also shown in Fig. 4. Of note is the lack of rare EMF species colonizing small proportions of the root systems in URB4 and URB5 soils.

Fig. 4
figure 4

Cluster analysis among EMF species using presence/absence data. Species are clustered based on common site soil presence. Relative abundance of each EMF species as percent of total root tips colonized across the ten replicate seedlings for each site soil is shown. Although the proportion of colonized root tips does not indicate the abundance of genetic fungal individuals, it can still suggest a level of dominance of fungal species within a root system. Four major species assemblages are indicated: reference site group, URB3 group, common group, and URB2 group

Full size image

Discussion

EMF community differences

Our results indicate that the level of site disturbance affects the type of impact on EMF communities. In URB1 and URB2, sites of intermediate disturbance, EMF root colonization rates were not significantly affected, but EMF community composition was different in comparison to the reference sites. High disturbance soils (URB4 and URB5), however, were associated with both altered community composition and significantly reduced root colonization rates. These results suggest that inoculum abundance or spread of fungi on a root system is not significantly limited unless the soil is extremely disturbed (polluted, high pH, low organic matter, compacted) or specific host plants are absent, potentially leading to fewer inoculum sources. Dickie and Reich (2005) found a similar relationship between EMF diversity and colonization rates in which reductions in fungal community diversity were observed before declines in seedling colonization rates as distance from forest trees increased. Additionally, other studies have found that disturbances such as clear-cutting affect EMF community composition but not root colonization rates (Byrd et al. 2000; Jones et al. 2003).

Jackknife estimates of overall EMF species richness show that with increased sampling, REF1 would likely retain its standing as the most species-rich soil and URB4 and URB5 as the most species-poor. These estimates also indicate that the full pool of species at each site was not observed given the sampling effort, particularly at the more species-rich sites, as has been observed repeatedly in EMF community studies (Horton and Bruns 2001; Taylor 2002). If a greater diversity of EMF mutualists associated with a seedling’s root system confers greater benefit to the seedling, then seedlings at URB4 and URB5 would be at a disadvantage. In these site soils, seedlings that exhibited any EMF colonization were colonized by only one species. Seedlings in REF1 and REF2 soils had up to six and five species per root system, respectively. EMF species differ in their physiological functioning traits (Baxter and Dighton 2005), capacities to acquire soil nutrients (Baxter and Dighton 2001), and tolerance to environmental factors (Baxter and Dighton 2005). Increasing EMF diversity on a host’s root system could allow for increased overall EMF functioning under changing conditions.

The ANOSIM results suggest some degree of convergence among EMF communities of the urban site soils, whereas communities of the reference site soils converge with each other and share significant similarities with only the most species-rich urban site (URB3). We suggested earlier the possibility of an “urban EMF suite”—an assemblage of EMF species found commonly across disturbed sites much like a suite of plant species predictably found in urban areas (Del Tredici 2010). As indicated in Fig. 4, a distinct set of fungal species was found on host roots across multiple sites (“common group”), perhaps indicating that these particular species are more tolerant of a wide range of environmental conditions. Amanitaceae sp., Russula sp. (as Russula sp. 2), and Tomentella sublilacina, the three EMF found exclusively on host roots in both REF1 and REF2 soils, are characteristic of late-successional or mature forest stands (Dighton and Mason 1985; Keizer and Arnolds 1994; Taylor and Bruns 1999; Lilleskov and Bruns 2005). Multiple Russula species were found in REF1, REF2, and, interestingly, URB3 soils. Cenococcum was found in five of our seven site soils, including the soils of three disturbed sites. Its distribution is known to be broad (Trappe 1964; LoBuglio 1999; Cripps 2003), and propagules of this genus have been found to persist for years even after clear-cutting (Shaw and Sidle 1982). Scleroderma, present in all but one site’s soil (as either Scleroderma sp. or S. citrinum), has been found to be typical of disturbed sites (Danielson 1984). It is not surprising that fungal species such as S. citrinum, Tuber sp., Thelephora terrestris, and Cenococcum sp. were found in REF1 and REF2 soils in addition to those of the disturbed sites, just as ruderal plants such as Daucus carota (Queen Anne’s lace) and Solidago rugosa (rough goldenrod) can be found in both urban and rural meadows. Also, although REF1 and REF2 are representative of relatively undisturbed forest systems in our study area, being located in a metropolitan region means that they are still impacted by some degree of disturbance, particularly nitrogen deposition (this could be a reason for the absence of Cortinarius spp. in these soils, as this genus has been shown to be among the first to drop out of high nitrogen soils (Wallenda and Kottke 1998; Lilleskov et al. 2001; Peter et al. 2001)). We would expect to see some disturbance-associated species at these sites.

Of note is the bimodal pattern of EMF colonization in URB3 soil. The distribution of colonization rates suggests that at a small scale (e.g., within meters), sources of EMF colonization at this site are either present at moderately high levels or are completely absent. Implications for establishing EMF-dependent seedlings could be significant; depending on where a seed lands and germinates, it could encounter either abundant EMF inocula or none at all. Boerner et al. (1996) found similar patchiness in disturbed sites, in which small patches of moderate inoculum availability were surrounded by areas completely lacking in inoculum. Dickie and Reich (2005) found spatial heterogeneity in EMF inoculum distribution, with decreasing ectomycorrhizal colonization of oak seedlings as distance from a forest edge increased. At URB3, we believe two historical factors may be responsible for zones of low inoculum availability: use of the site for an arsenal and soil dumping. Such heterogeneity in inoculum distribution could regulate spatial patterns of seedling establishment or growth. For example, Dickie et al. (2005) observed greater oak seedling growth at intermediate distances from mature trees, where canopy shading was not prohibitive but mycorrhizal inoculum remained abundant.

Reasons for different EMF communities

Many possible explanations could account for differences in EMF colonization, species richness, and community composition on host roots among site soils. Jumpponen and Egerton-Warburton (2005) discuss a series of “filters” fungal species must pass through until site-specific mycorrhizal communities are established: the host filter, in which compatibility between host plants and fungal species determines mycorrhizal communities; the environmental filter, which imposes abiotic constraints; and the biotic filter, in which facilitative or competitive interactions between fungal species determine community composition. Although an array of fungal species may be present in a site’s propagule bank, only a selection will be represented in the active mycorrhizal community, determined by successful passage through the above site-specific filters (Jumpponen and Egerton-Warburton 2005). Peay et al. (2010), for example, observed effects of soil type on EMF richness and community structure in a tropical forest and attributed these effects to either physical and chemical differences among soil types (abiotic filtering) or differences in host plant communities associated with the soil types (host filtering).

Differences among individual site histories likely contribute to observed differences in EMF communities. Past disturbance at our urban sites could have disrupted the mycelial network and may have caused extensive lags in EMF community recovery. We would expect an urban park fabricated with excavation material to have a different EMF community structure from that of a post-agricultural woodland. However, even though the particular sources of disturbance vary among the urban sites, we still see convergence in EMF communities among several of these site soils.

Other possible explanations of EMF community differences include adverse soil conditions, distance from reference sites, and landscape fragmentation. Poor soil conditions at the more highly disturbed sites could be prohibitive toward EMF survival or spread on host roots. Additionally, long distances from less disturbed woodlands and fragmentation of sites isolated within an urban matrix could present barriers to EMF spore dispersal. Although wind-dispersed spores could potentially travel across broad areas (Molina et al. 1992), we do not know whether urban environments physically interfere with such dispersal. Interestingly, Tuber spp. were highly represented in the urban sites and hypogeous fungi such as Tuber can be dispersed over long distances by mammals (Ashkannejhad and Horton 2006). It is also notable that of all the urban sites, URB3 is within closest proximity to the reference sites and is the only urban site whose EMF community composition is not significantly different from that of the reference sites.

Differences in successional stages and aboveground plant community composition could also explain EMF community differences among site soils. Highly disturbed and early successional systems are often dominated by non-mycorrhizal and facultatively mycorrhizal plants (Jumpponen and Egerton-Warburton 2005). Lack of reference site EMF species in the urban soils could be partly due to a lack of inoculum otherwise available through mycelial transfer from pre-existing mature root systems (Fleming 1983; Kranabetter and Frieson 2002) and fragments of such in our bioassay soils. However, successional state does not explain the depauperate EMF community in URB1 soil. Given the passing of 70+ years since agricultural activities and the subsequent development of a closed-canopy woodland, we would expect the EMF community of this site to be more similar to that of REF1 or REF2. URB2, also, has mature Quercus individuals and was lacking in EMF species typical of mature woodlands.

It is important to keep in mind that in this study we have observed EMF colonization only of Q. rubra seedlings and cannot make assumptions about fungi that might colonize other host plant species. Additionally, Q. rubra is known to associate with both ectomycorrhizal fungi and arbuscular mycorrhizal fungi (AMF; Dickie et al. 2001). It is possible that in the urban soils, any reduction in benefits provided to host seedlings due to a lack of EMF associates could be ameliorated through root colonization by AMF. In fact, Bainard et al. (2011) suggested that trees in urban environments may rely more on AMF than EMF.

Implications for community ecology

Lack of mycorrhizal mutualists at disturbed urban sites could be a constraint to plant community development (Dickie and Reich 2005), particularly if spore dispersal into isolated urban patches is limited. Nuñez et al. (2009) found that invasion of surrounding forests by non-native trees from Argentinian Pinaceae plantations was prevented by the low abundance of EMF propagules in soils far from the plantations. In our study, low potential for EMF colonization could be one of several urban stressors that prevents establishment of EMF-dependent plant species in severely disturbed sites.

Given the demonstrated importance of mycorrhizal fungi to plant survival and community structure (Read and Birch 1987; Horton and van der Heijden 2008), determining the status of urban mycorrhizal fungal communities will contribute to our understanding of urban plant community assembly. The mycorrhizal community should also be considered during ecological restoration of urban sites, as successful restoration of a target plant community could benefit from, if not require, the presence of an appropriate mycorrhizal fungal community (Eviner and Hawkes 2008; van der Heijden and Horton 2009; Newbound et al. 2010) that, as our study shows, is likely to be depauperate at urban sites.