- Top of page
- Materials and Methods
Arbuscular mycorrhizas (AM) are likely a primitive state for all vascular plants, and the glomalean fungi that produce them (AMF, phylum Glomeromycota) are among the most abundant fungi on the planet (Smith & Read, 1997; Brundrett, 2002). Glomalean fungi have been shown to affect plant species richness in greenhouse microcosms and natural communities. Plant species richness often increases with the number of AMF species added in greenhouse experiments (Grime et al., 1987; van der Heijden et al., 1998, 2002, 2004) but can decline with AMF species richness in natural grasslands, based on comparisons after vs before fungicide treatment (Newsham et al., 1995; Hartnett & Wilson, 1999). These studies have led to models predicting that plant diversity should increase with AMF diversity (van der Heijden, 2002), or that AMF effects on plant diversity should depend on the mycorrhizal status of competitively dominant and subordinate plant species (Hartnett & Wilson, 2002; Allen et al., 2003; Urcelay & Diaz, 2003). According to these models, plant diversity should increase with AMF diversity, unless all plants are independent of mycorrhizae, or unless dominant plant species are strongly dependent on AMF while subordinate ones are weakly dependent. In the field, different plant species, life-stages, or carbon cycling strategies are often associated with different AMF species (Clapp et al., 1995; Eom et al., 1999; Cornelissen et al., 2001; Husband et al., 2002; Vandenkoornhuyse et al., 2002), and experiments show that plants affect AMF community composition (Johnson et al., 1992; Johnson et al., 2004, 2004). Together, these studies suggest that plant and AMF communities influence each other profoundly; in addition, there are strong empirical and theoretical reasons to expect that external ecological conditions (e.g. light, moisture, phosphorus (P) and nitrogen (N) supplies) would also affect plant and AMF communities, both directly and indirectly (Givnish, 1986; Tilman, 1988; Johnson (1992); Smith & Read, 1997; Clark & Zeto, 2000; Corkidi et al., 2002; Facelli & Facelli, 2002; Miller et al., 2002; Bauer et al., 2003). To date, however, no study has simultaneously examined associations (and potential interactions) among plants, AMF, and ecological conditions, especially in diverse communities.
Here, we examine the relationships among plant and AMF communities and environmental conditions in species-rich, environmentally patchy oak savannas of the US Midwest. To simplify the task, we focus on species richness and community composition of vascular plants and AMF. Our goal is to determine how plant and AMF species richness and composition are related to each other and/or environmental gradients within and among sites.
Midwestern oak savannas are among the most floristically diverse plant communities in the central USA. One survey found 417 plant species in 722 m2 across 12 remnant Wisconsin savannas (Leach & Givnish, 1999). This represents about 22% of the state's vascular flora, and c. 90% of these species are herbs or shrubs. Studies of Wisconsin oak savannas have shown that the understory composition correlates with gradients in soil texture/fertility and light availability. The soil gradient involves collinear increases in per cent clay and silt, Kjeldahl N content, exchangeable calcium (Ca), magnesium (Mg) and per cent organic matter, and explains roughly twice as much variation in plant species composition as light availability, which varies from full sun in openings to deep shade near tree boles (Leach, 1996; Leach & Givnish, 1999; Meisel et al., 2002). The AMF response to such gradients and/or vascular plant composition is, however, almost unknown, with just one study relating AMF composition to the single factor of light availability across a natural prairie–savanna–forest gradient in Illinois (Benjamin et al., 1989).
To elucidate the relationships among plant communities, AMF communities, and the physical environment, we surveyed three remnant oak savannas in southern Wisconsin that vary internally in light regime and collectively span the range of a previously investigated soil-texture gradient (Leach & Givnish, 1999). We used the resulting data to determine whether plant and AMF species richness and composition are correlated with each other and/or light availability and soil texture and fertility. We sampled the biotic communities at two different spatial scales to determine whether associations depend on sampling scale. Our hypotheses are that: (1) AMF and plant species richness should be positively correlated with each other (van der Heijden et al., 1998, 2002); (2) AMF and plant species richness should decrease with increasing soil fertility, as a result of increased selection for AM on nutrient-poor soils; and (3) plant and AMF community composition should be correlated with each other, and with light availability and (especially) soil fertility and texture. Given that AMF and plants might respond directly to light and soil gradients, or indirectly to effects of the latter on partners or enemies of a given fungus or plant, there is the potential for complex patterns of association among AMF, plants, and environment. Since oak savannas are among the most endangered communities in the upper Midwest (Curtis, 1959; Nuzzo, 1986; Leach & Givnish, 1999), the results of this study may have fundamental implications for the role of mycorrhizal diversity and composition in their conservation and restoration.
Morphological vs molecular approaches
We used spores to quantify AMF community composition and species richness. This is a natural approach, given that AMF species are described based on spore morphology. Nevertheless, spore records are subject to bias, as with the pollen records used to investigate past communities of higher plants. Sporulation by different AMF species probably depends on biotic and abiotic conditions, and presence of species in the ‘spore community’ may thus not always reflect their abundance as mycelia or physiological importance as vascular plant symbionts. However, sporulation itself is an important aspect of AMF life cycles, so the ecological distributions of spores produced by different species are important in their own right, in addition to their use as proxies for species distributions. Spores have a limited lifetime (a few months or years) in the soil, at least as entities identifiable to species level, so spore samples reflect a constantly updated record of recent AMF reproduction.
Molecular approaches have the potential to revolutionize studies of AMF communities by bypassing the need to study spores, but currently present several problems. First, rDNA sequences are the basis for most current identification systems at present, but AMF species are notorious for having polymorphic rDNA sequences (Sanders, 2002; Redecker et al., 2003). This has led some to claim that spores contain ribosomal DNA from several different morphospecies (Clapp et al., 2001). The extent to which such variation reflects simple polymorphism among multiple copies of rDNA in a single genome or multigenomic mycelia is still controversial (Sanders, 2002; Sanders et al., 2003; Pawlowska & Taylor, 2004). Individual spores can contain a thousand or more nuclei, and it is normal to recover multiple sequences by polymerase chain reaction (PCR) amplifications from a single spore (Lanfranco et al., 1999; Antoniolli et al., 2000; Pawlowska & Taylor, 2004). Second, the diversity of sequences detected depends on the details of the molecular screening approach taken, and many (if not all) approaches currently in use are inherently unable to detect all rDNA sequences present (Redecker, 2000; Vandenkoornhuyse et al., 2002; Redecker et al., 2003; Schussler et al., 2003). Competitive interactions during PCR can overwhelm uncommon sequences, so accessions must be intensively cloned and sequenced before rDNA sequence diversity can be reliably established. Given the large number of species potentially present in a sample, and the large number of genomes present in individual mycelia, it might be necessary to sequence 100 or more clones from a given sample to ensure adequate representation of the AMF species present. For ecological studies of AMF community variation along gradients such as that reported here, involving scores or hundreds of samples, the cost of cloning and direct sequencing could be prohibitive (> US$ 250 000). DNA-based microarrays may soon reduce costs dramatically but are not yet developed. Finally, there are currently no individual rDNA primers that permit identification of all major glomalean lineages, let alone species, and no agreed-upon set of primers to use for such purposes. Given that the identities of rDNA sequences are traced to spore morphology, and the substantial issues regarding methodology and cost of molecular approaches, and inherent interest in conducting an initial study of relationships among plant communities, AMF (or at least AMF spore) communities, and environmental conditions in natural communities, we have based the present study on spore morphology. We incorporate molecular approaches in other facets of our systematic and ecological research and look forward to the development of powerful and cost-effective molecular techniques for projects to complement the results of the one presented here.
- Top of page
- Materials and Methods
We found 137 vascular plant species in 63 quadrats distributed across the three savannas studied. Of these, 62 species were found at Marshall, 62 in TCL and 43 in TCU. At Marshall, the dominant understory species was Carex lacustris, at TCU, Carex pensylvanica, and at TCL, Rubus hispidus (Table 2). Of these dominants, both Carex species are considered nonmycorrhizal and Rubus is considered facultatively AM (Trappe, 1987; Miller et al., 1999). Plant species showed substantial heterogeneity in their distribution across savannas; only four species are common to all three savannas (Table 2). Marshall shared six and 13 species with TCU and TCL, respectively, and TCU and TCL shared 17 species.
Table 2. Forty most common plant species of 137 found in this study
|Species||Frequency and cover ranking|
|Carex pensylvanica||–||0.71 (12)||1.00 (1)|
|Quercus (subg. Erythrobalanus)||–||0.43 (17)||0.71 (4)|
|Prunus serotina||–||0.43 (10)||0.62 (2)|
|Solidago canadensis||1.00 (6)||–||–|
|Carex lacustris||0.81 (1)||–||–|
|Rubus idaeus var. strigosus||0.33 (4)||0.14 (13)||0.33 (8)|
|Helianthus strumosus||0.67 (5)||–||–|
|Asclepias syriaca||–||0.05 (30)||0.62 (3)|
|Cornus racemosus||0.33 (10)||0.33 (25)||–|
|Potentilla simplex||–||0.67 (15)||–|
|Athyrium filix-femina||–||0.62 (2)||–|
|Phalaris arundinacea1||0.14 (23)||0.48 (5)||–|
|Fragaria virginiana||0.48 (20)||0.10 (49)||0.05 (30)|
|Silphium terebinthaceum||0.57 (3)||–||–|
|Rubus hispidus||–||0.57 (1)||–|
|Acer rubrum||0.05 (58)||0.43 (20)||0.10 (37)|
|Elymus riparius||–||0.29 (19)||0.24 (11)|
|Zizia aurea||0.52 (14)||–||–|
|Urtica procera||–||0.19 (24)||0.33 (15)|
|Aster sagittifolius||0.48 (17)||0.05 (54)||–|
|Oxalis sp. (stricta)||–||0.48 (29)||–|
|Carpinus caroliniana||–||0.43 (4)||–|
|Vaccinium angustifolium||–||0.19 (38)||0.24 (6)|
|Helianthus grosseserratus||0.43 (12)||–||–|
|Rumex acetosella1||–||–||0.43 (7)|
|Geranium maculatum||0.19 (28)||0.24 (26)||–|
|Impatiens capensis||0.29 (18)||0.10 (8)||–|
|Ilex verticillata||–||0.38 (9)||–|
|Thalictrum dasycarpum||0.38 (15)||–||–|
|Polygonum sagittatum||–||0.38 (23)||–|
|Glechoma hederacea||0.14 (41)||0.24 (32)||–|
|Pinus strobus||–||0.05 (55)||0.33 (18)|
|Viola sp. (sororia)||–||0.33 (33)||0.05 (42)|
|Spartina pectinata||0.33 (2)||–||–|
|Apios americana||0.33 (7)||–||–|
|Carex sp. (tenera)||–||0.33 (11)||–|
|Monarda fistulosa||0.05 (57)||0.14 (31)||0.14 (14)|
|Aster lateriflorus||–||0.24 (18)||0.10 (36)|
|Lysimachia lanceolata||–||–||0.33 (16)|
|Amphicarpaea bracteata||0.05 (59)||0.29 (35)||–|
We documented a total of 18 named AMF species, with 14 at Marshall, 17 at TCL, and six at TCU (Table 3). The AMF species are, on average, more widely distributed across savannas than the plant species, with only five AMF taxa restricted to single savannas, and five found in all three savannas. Glomus constrictum was most frequently found in Marshall, Acaulospora scrobiculata in TCL, and Glomus intraradices and Scutellospora dipapillosa were equally most frequent in TCU (Table 3).
Table 3. Species of glomalean fungi identified from spores in this study
|Paraglomus sp. (brasilianum?)||–||0.05||–|
The mean number of plant species was 11.4 ± 3.7 (SD) per quadrat, and 7.0 ± 2.9 per point (Table 4). TCU had significantly fewer plant species per sample than did the other two savannas at both scales (P < 0.001). Numbers of AMF species per sample showed a similar, significant pattern of differentiation, with TCU again having the fewest species per sample (P < 0.001). Across savannas, an average of 3.7 ± 2.9 species were found per sample, with the mean value ranging from 1.4 ± 1.4 at TCU to 6.0 ± 2.0 at Marshall (Table 5). As expected, the savannas differed significantly in Kjeldahl N, available P, and soil texture (Table 4). Silt, clay, and N content increased in moving from TCU to TCL to Marshall; available P increased from TCU to Marshall to TCL. The savannas did not differ significantly in estimates of understory PFD at any of the times studied, although Marshall was, in general, somewhat shadier than the other savannas.
Table 4. Environmental data summary
| ||#PlantSpQuad||#PlantSpPt||#SporeSp||N (g g−1)||P (g g−1)||% Sand||% Silt||% Clay||PFDEqui (mol m−2 d−1)||PFDSumM (mol m−2 d−1)||PFDSum (mol m−2 d−1)|
|Mean (SD)||18.5 (12.2)||11.35 (3.73)||7.03 (2.92)|| 3.73 (2.70)||4083.7 (3426.9)||28.9 (19.0)||61.4 (27.4)||17.4 (14.4)||21.2 (15.2)|| 8.7 (4.7)||14.5 (8.1)|
|Mean (SD) per savanna|
|Mar||11.4 (2.9) ab|| 8.6 (2.3) a||6.0 (2.0) a||8105.0 (2313.7) a|| 16.6 (4.5) a||30.7 (10.8) a||29.4 (11.5) a||39.9 (5.8) a|| 5.96 (4.69) a||10.02 (7.79) a||11.46 (10.33) a|
|TCU|| 8.4 (2.2) a|| 4.1 (1.7) b||1.4 (1.4) b||1035.1 (466.3) b|| 23.3 (8.1) a||91.3 (2.1) b|| 1.5 (1.5) b|| 7.2 (1.5) b||10.33 (3.25) a||16.73 (6.71) a||22.26 (10.22) a|
|TCL||14.4 (3.5) b|| 8.4 (2.2) a||3.7 (2.2) c||2861.3 (1267.1) b|| 48.5 (21.6) b||63.8 (15.2) c||20.8 (8.4) a||15.4 (7.2) c||9.84 (4.95) a||16.93 (8.01) a||22.28 (13.15) a|
Table 5. Linear correlations among variables in the environmental data set
|#PlantSpPt|| 0.59***|| || || || || || || || || |
|#SporeSp|| 0.18|| 0.57***|| || || || || || || || |
|N|| 0.12|| 0.54***|| 0.63***|| || || || || || || |
|P|| 0.43**|| 0.09||−0.23||−0.38*|| || || || || || |
|Sand||−0.27*||−0.59***||−0.70***||−0.83***|| 0.31*|| || || || || |
|Silt|| 0.37|| 0.55***|| 0.60***|| 0.66***||−0.12||−0.92***|| || || || |
|Clay|| 0.14|| 0.54***|| 0.70***|| 0.88**||−0.44**||−0.93***|| 0.72***|| || || |
|PFDEqui|| 0.07||−0.16||−0.18||−0.49***|| 0.12|| 0.29*||−0.14||−0.39*|| || |
|PFDSumM|| 0.17||−0.17||−0.18||−0.45***|| 0.10|| 0.28*||−0.16||−0.36*||0.85***|| |
|PFDSum|| 0.18||−0.09||−0.17||−0.42***|| 0.12|| 0.28*||−0.16||−0.35*||0.79***||0.92***|
Plant and AMF species richness were strongly and positively correlated when measured at the point scale (r2 = 0.31, P < 10−5) (Fig. 1), but not when plants were surveyed at the quadrat scale (r2 = 0.03, P > 0.24) (not shown). The number of plant species per point increased significantly with Kjeldahl N, clay and silt contents (r = 0.54–0.59, P < 10−4) and decreased significantly with per cent sand (r = –0.59, P < 10−4) (Table 5). Numbers of AMF species also increased significantly with Kjeldahl N, per cent silt and per cent clay (r = 0.60–0.70, P < 10−4), and decreased with per cent sand (r = –0.70, P < 10−4). The number of plant species per quadrat only showed a significant correlation under the Bonferroni correction with available P (r = 0.43, P < 0.001) (Table 5). As expected, there were highly significant correlations (P < 10−4) among Kjeldahl N and all soil texture variables, with correlation coefficients ranging from 0.66 for N–silt, to −0.92 and −0.93 for sand–silt and silt–clay, respectively. Available P showed a significant correlation with clay content (r = –0.44, P < 0.001). No measure of species richness correlated significantly with any of the light variables, but the latter showed a strong positive correlation with each other (r = 0.79–0.92, P < 10−4) and a negative correlation to per cent clay (r = –0.42 to −0.49, P < 0.001).
Figure 1. Number of plant species vs number of arbuscular mycorrhizal fungal (AMF) species, all points (n = 63). For the linear regression shown, r2 = 0.31, P < 0.00001. TCU, Upper Tarr Creek; TCL, Lower Tarr Creek.
Download figure to PowerPoint
Generally, there was less than a 0.02 difference between SEM partial correlation coefficients and the linear correlation coefficients in Table 5 (data not shown). The four exceptions, in which the differences between partial and linear coefficients ranged from 0.03 to 0.07, disappeared when we calculated linear correlations on the square-root transformed data used in the SEM. Thus, there appear to be no meaningful indirect correlations, and standard linear correlations (Table 5) seem to represent relationships among plant species richness, AMF species richness and the environmental variables measured as accurately as does a simple structural equation model.
Mantel tests showed a significant positive correlation between spore composition and quadrat plant composition across savannas (r = 0.24, P < 0.001). Spore composition and point plant composition showed an almost identical pattern of correlation (r = 0.24, P < 0.001).
In addition, all three NMS ordinations showed significant correlations (r2 from 0.31 to 0.62) among the samples’ axis scores on all three ordinations (Table 6). The NMS returned a two-dimensional ordination for the quadrat data (total r2 = 0.51, final stress = 16.98, instability = 10−4) and three-dimensional ordinations for the point data (total r2 = 0.69, final stress = 19.51, instability = 10−5) and spore data (total r2 = 0.79, final stress = 17.49, instability = 0.003) (Fig. 2 and Table 6).
Table 6. Nonmetric multidimensional scaling (NMS) ordination statistics
|Data Set||Axis||r2||#Plant SpQuad||#Plant SpPt ||#Spore Sp ||N||P||Sand||Silt||Clay||Plot axis||Point axis||Spore axis|
Figure 2. Joint plots of nonmetric multidimensional scaling (NMS) ordinations for Quadrat, Point and Spore datasets. The two strongest axes are shown, and figures are rotated to show the overlays in the same orientation. S, final stress; I, final instability. Overlays show environmental variables as listed in Table 1. In addition, axes from each ordination are overlaid on each other. Quadrat1 and Quadrat2 are scores from the two Quadrat axes, Point1–Point3 are from the Point ordination, and Spore1–Spore3 are from the Spore ordination. Correlation coefficients are given in the Table 6.
Download figure to PowerPoint
Joint plots of the quadrat, point, and spore compositional data showed the same correlations with environmental variables: strong positive correlations with N, clay and silt and numbers of plant and AMF species per point, and a strongly negative correlation with per cent sand (Fig. 2). That is, AMF spore communities and plant communities at the quadrat and point scales responded in quite similar ways to ecological conditions. The major axes aligned with the N and percent clay gradient explained 25% of the variance in the spore data, 29% of that in the point data, and 34% of that in the quadrat data (Table 6). Scores on the first axes of the quadrat and point ordinations showed much higher correlations with Kjeldahl N (quadrat r2 = 0.82, point r2 = 0.55) and soil texture variables (quadrat r2 = 0.53–0.90, point r2 = 0.59–0.74) than did the corresponding scores for the spore data (r2 = 0.33 for N, and r2 = 0.29–0.39 for other soil variables) (Table 6). Intriguingly, plant and AMF diversity also showed similar responses to this major environmental gradient: numbers of plant species per point had r2 values of 0.32, 0.40 and 0.42 for the first axes of the quadrat, point, and spore ordinations, respectively, while numbers of AMF species had a r2 values of 0.50, 0.41, and 0.23 for same axes of these ordinations (Table 6). Available P (quadrats and point ordinations r2 = 0.20) and numbers of plant species per quadrat (spores and point ordinations r2 = 0.20) correlated more strongly with other axes (Table 6).
- Top of page
- Materials and Methods
We found support for our first hypothesis: the numbers of plant and AMF species per sample were indeed positively correlated across savannas and microsites within savannas (Fig. 1). Given that the dominant understory species in each of the savannas studied are nonmycorrhizal (C. pensylvanica and C. lacustris) or facultative AM (R. hispidus), these results are consistent with the models proposed by Hartnett & Wilson (2002), van der Heijden (2002), and Urcelay & Diaz (2003): increased mycorrhizal diversity is likely to benefit competitively subordinate plants and thus increase overall plant diversity. Significant correlations between AMF and plant species richness were found only at the point level and not between AMF species richness and plant species richness at the quadrat level. This result underscores the importance of sampling scale in determining patterns in species richness. Reports of scale dependence of patterns in AMF species richness have surfaced in recent years (Schultz, 1996; R. M. Miller, unpubl. data), but few, if any, detailed accounts have been published. Correlations between plant and AMF diversity tend to be lower within individual savannas than across savannas, but this would be expected simply based on restricting an identical correlation structure across savannas to a narrower range of ordinates within savannas. On average, each sample contained the largest number of plant and AMF species at Marshall, and the fewest plant and AMF species at Upper Tarr Creek (TCU).
Our second hypothesis was contradicted, at least at first glance: the numbers of AMF and plant species increased significantly with soil fertility (specifically, Kjeldahl N content, per cent clay and per cent silt). Our prediction had been to expect the greatest number of AMF species in the sandiest, least fertile substrates at TCU (Table 4). However, Marshall, with the highest N concentrations, had the greatest numbers of AMF and plant species (Table 4; Leach & Givnish, 1999). Paradoxically, our data might support a rephrased version of our second hypothesis, in which increased AMF incidence is favored by a greater relative limitation of plant growth by phosphorus vis-à-vis nitrogen or other soil nutrients. Most importantly, AMF may provide no added value in harvesting P from sandy soils with large pores that can be penetrated by root hairs. Alternatively, low P : N ratios might favor the dominance of strongly mycorrhizal plants, if P uptake is one of the major benefits of mycorrhizal association. Soil P : N ratio is lowest (0.002) at Marshall and highest (0.225) at TCU, and shows a strong correlation (r = 0.63) with AMF species number across all samples. Paralleling our expections, Phoenix et al. (2004) recently found that N deposition induced greater root-surface phosphatase activity in graminoids and forbs in calcareous grasslands. Soil P : N ratio might not provide the best estimate of the actual relative supplies of these elements to plants and AMF. However, several studies suggest that foliar P : N ratios are reliable indicators of the degree to which plant growth is limited by P vs N (Koerselman & Meuleman, 1996; Bedford et al., 1999; Lockaby & Conner, 1999; Tessier & Raynal, 2003).
Our data demonstrate some distributional differences among genera of Glomales along the major nitrogen–soil texture gradient. Scutellospora was most common in the TCU, the sandiest, low-N savanna, whereas most Glomus species were found in Marshall, the high clay, high-N savanna, and Acaulospora showed little apparent preference. This distribution pattern of Scutellospora, Glomus and Acaulospora by soil texture has been observed elsewhere (Egerton-Warburton & Allen, 2000; Treseder & Allen, 2002).
We found support for our final hypothesis: plant and AMF community compositions were significantly correlated with each other and showed strikingly similar responses to environmental gradients (Table 6; Fig. 2). Mantel tests showed significant positive correlations between AMF composition and plant composition at the point and quadrat levels. Furthermore, NMS axis scores for all three ordinations (point, quadrat and spore data sets) were correlated with each other, suggesting that there are fundamental similarities in the ways in which plant and AMF community composition vary along environmental gradients generally, and not merely those we measured directly. These ordinations represent reductions of 18- to 137-dimensional data (one dimension per species) to two or three dimensions, so there is no a priori mathematical reason to expect significant correlations among ordination scores as an artifact.
Both AMF and plant community composition were significantly correlated with the N–soil texture gradient and with species richness per point at all measurement scales. No relationship was found between community composition and the light gradient, however, which seems consistent with the weak effect of the latter on plant communities in the more extensive sample analysed by Leach & Givnish (1999). The remarkable collinearity of shifts in AMF composition and species richness, plant composition and species richness, soil silt and clay content, and N content, observed in this study makes it difficult to infer any possible interaction of the effects of two or more of these variables on the others (e.g. plant species richness and soil texture on AMF species richness) using studies of direct or partial correlations. However, our results are perhaps the first to extend the finding of a positive association between AMF species richness and plant species richness – originally documented in greenhouse microcosm experiments (Grime et al., 1987; van der Heijden et al., 1998, 2002, 2004) – to natural communities.
The rise in AMF species richness with soil N content has a number of potential explanations beyond those based on soil texture or soil P : N ratio. Sporulation might be stimulated by higher soil N content. Leach & Givnish (1999) found that Kjeldahl N content was positively correlated with the levels of Ca, Mg and soil organic matter across the savannas surveyed, so one or more of these nutrients may be driving AMF species richness. Finally, because a substantial fraction of variation in soil N involves variation among the three study sites, it is possible that historical contingencies associated with different savannas (e.g. different burning frequencies or differential invasion by various organisms) might be largely responsible for the differences seen. Although these possibilities are worth pursuing, we cannot test them with current data.
We did not find support for any ecological effect of light availability similar to those reported in previous studies on savanna plant communities (Leach & Givnish, 1999; Meisel et al., 2002). We believe that this reflects the low sample size in our study and the inherently weaker association of plant communities with light vs soil seen in previous studies, rather than being an artifact of Type II errors introduced by the Bonferroni correction (use of P < 0.001). The previous studies cited had at least twice as many samples per savanna as ours did, and none of the correlations between AMF or plant communities and light availability in our study would have been significant even if we had used P < 0.05 as the cut-off.
As indicated in the Introduction, a key question in result interpretation is whether the sampling protocol for AMF was adequate, given that we studied only spores and thus missed nonsporulating taxa, and could not determine which fungi were active on plants roots at the time (Clapp et al., 1995; Husband et al., 2002). Trap culturing and some molecular techniques could sample nonsporulating fungi, but we chose to use spore morphology after evaluating these methods. We experimented with trap cultures, but cultures from all three savannas generally returned only about 50% of the AMF species we found in the samples that were visually analysed, and the cultured species were all found among spores identified microscopically (F. C. Landis and C. J. Andrew, unpubl. data). Spore identification thus appears to provide a more accurate measure of AMF species composition and richness than trap culturing for this study.
Molecular surveys using techniques such as terminal restriction fragment length polymorphisms (T-RFLPs) might, in principle, provide more accurate estimates of AMF community composition and diversity than spore data. However, the methodological problems associated with current molecular approaches are by no means trivial to resolve (see Introduction). More importantly, a molecular version of the present study would have been far more costly and logistically complex than any AMF study published to date (e.g. Husband et al., 2002; Vandenkoornhuyse et al., 2002; Vandenkoornhuyse et al., 2003; D Johnson et al., 2004), and extrapolating AMF species diversity from sequence diversity would not be straightforward (Clapp et al., 2001; Kuhn et al., 2001). Finally, most objections to the use of spore data appear to come from previous uses of spore numbers to estimate AMF abundance and activity (Clapp et al., 1995). In this study, we used spore data only to infer the presence/absence of AMF species in different savannas. We look forward to testing the preliminary results reported here by using molecular approaches once they become sufficiently powerful and cost-effective. However, we find it extremely hard to believe that the striking correlations among AMF community composition and species richness, plant community composition and richness, and environmental factors that are demonstrated in this paper by our use of spore morphology could be artefacts.
Our findings indicate that the composition and diversity of understory plant and AMF assemblages in Midwestern oak savannas display similar responses to environmental gradients, including primarily soil N content, per cent silt and per cent clay. AMF and plant composition and species richness are positively correlated with each other across the savannas studied, although the correlations of species richness across plants and fungi are scale-dependent. Surprisingly, AMF species richness was highest on the richest soils, contrary to expectations based on proposed advantages to plants of mycorrhizal nutrient capture, but consistent with the observed trend for plant species richness to increase toward richer soils in savannas (Leach & Givnish, 1999). In Midwestern oak savannas, plant and AMF communities appear to be linked to each other across the mycorrhizal interface, and to the soil they both inhabit. It is remarkable that both groups show such strong and similar responses to soil texture and N content.