description of field sites and sampling
Ten fields, five on sand and five on black clay, were selected within the Tsholotsho communal area (27°49.73′E, 19°51.07′S) of Zimbabwe. The climate is semi-arid, with a mean annual rainfall of 540 mm concentrated in the growing season from November to April. Average yearly maximum and minimum air temperatures are 26 and 11 °C, respectively (ICRISAT, unpublished data). All fields had been converted during the past 10–15 years to agriculture from native bush vegetation (based on interviews with farmers). Since then, common subsistence crops such as maize, sorghum (Sorghum bicolor L.), groundnut (Arachis hypogaea L.) and cowpea [Vigna unguiculata (L.) Walp.] have been cultivated on all fields. Maize was grown the previous year without any fertilizer on each experimental plot. All 10 fields were located within 25 km of each other (Fig. 1). Soil samples were taken randomly at a depth of 0–15 cm from five locations per field on 5 December 2002 and pooled by field. Soil characteristics were analysed by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in Zimbabwe and are presented in Table 1.
Figure 1. Map of Tsholotsho communal area in Zimbabwe indicating locations of the 10 study fields. C, clay soil; S, sand soil.
Download figure to PowerPoint
Table 1. Mean (SE) of selected soil characteristics of the two soils included in the field experiment. P-values were calculated based on a two-sample t-test, n = 5 (significant values in bold type)
| ||Sand soil||Clay soil||P|
| Clay||4.8 (0.9)||48.4 (3.3)||< 0.001|
| Silt||8.2 (1.3)||14.7 (1.4)||0.011|
| Sand||87.0 (1.9)||36.9 (4.1)||< 0.001|
|Moisture (%)*||4.4 (0.5)||15.1 (1.2)||< 0.001|
|Organic C (g C kg−1 soil)||3.9 (0.5)||9.0 (0.8)||< 0.001|
|pH (CaCl2)||5.2 (0.6)||7.1 (0.2)||0.053|
|Olsen P (mg P kg−1 soil)||6.2 (3.0)||1.3 (0.3)||0.093†|
|Total P (g P kg−1 soil)||0.37 (0.1)||0.63 (0.1)||0.067†|
|N available (mg kg−1 soil)||3.2 (1.2)||5.0 (1.5)||0.380|
|Total N (g N kg−1 soil)||0.26 (0.04)||0.61 (0.1)||0.016|
Maize seeds (SC 401, a common short-duration variety sold in Zimbabwe) were planted on a 0.5-ha ploughed section of each field between 7 and 19 December 2002. Twenty randomly selected maize plants per field were harvested 7 weeks after planting, allowing time for any slow colonizing AM fungal species to be detected in the roots (Hart & Reader 2002). Root subsamples from each plant were washed, blotted dry, further dried with a desiccant (Drierite, W.A. Hammond Drierite Company Ltd, Xenia, OH, USA), and shipped to The Pennsylvania State University, Pennsylvania, USA, where they were stored at −20 °C. Shoots and remaining roots were dried at 65 °C to constant mass, and root colonization by AM fungi was assessed by the line intersect method (Koide & Mooney 1987) after staining with trypan blue.
t-rflp analysis and species identification
We characterized the AM fungal community in maize roots in each field using terminal restriction fragment length polymorphisms (T-RFLP) of the internal transcribed spacer (ITS) region within the ribosomal gene complex. DNA was extracted from a single, 3-cm-long, root segment per plant using the Microbial DNA Isolation Kit (Mo Bio Laboratories, Solana Beach, CA, USA) according to Koide et al. (2005), and amplified using a nested polymerase chain reaction (PCR) protocol (Renker et al. 2003). Our PCR mixture was according to Koide et al. (2005), with the primer set for the second PCR (PCR2) labelled with fluorescent tags. DNA was purified following PCR2 using a commercially available kit (UltraClean™ PCR Cleanup Kit, Mo Bio).
The PCR2 product was digested with the restriction enzyme HinfI (New England Biolabs, Beverly, MA, USA) as in Koide et al. (2005). GeneScan analysis was performed on a nine-fold dilution of HinfI digestion fragments and on an 18-fold dilution of the PCR2 product according to Koide et al. (2005). To compensate for differences in DNA concentration across samples, the GeneScan output was standardized so that the tallest peak in each sample had a height of 1000; no peak with a height of < 50 after standardization was recorded. T-RFLP analyses were also performed on spores that had been isolated and identified from all 10 fields using trap culture techniques (Morton et al. 2004). The PCR protocol for spores was the same as for roots, except that DNA was extracted by crushing the spore in the PCR tube immediately prior to the first PCR.
Each T-RFLP profile comprised three data points: the length of the PCR2 product and the lengths of the two terminal restriction fragments (T-RFs) from the PCR product. A profile from a root sample was considered to match a spore profile if corresponding lengths of the PCR2 product and T-RFs were within 2 bp. This criterion was based on the precision of the ABI sequence analyser (our personal observations). Each known T-RFLP profile is hereafter referred to as a ribotype.
Profiles that did not match those of spores were compared with ribotypes in two other databases: (i) T-RFLP profiles from spores collected in Costa Rica, and (ii) virtual profiles determined for 435 ITS sequences from spores of 37 species published in GenBank (Aldrich-Wolfe, in press). Frequently encountered T-RFLP profiles that did not match ribotypes in our three databases (Zimbabwe spores, Costa Rican spores and spore virtual database from GenBank) were cloned and sequenced at the Nucleic Acid Facility at The Pennsylvania State University. PCR2 products were purified from agarose gels, ligated into the plasmid pGEM®-T Easy (Promega, Madison, WI, USA) and electroporated into Escherichia coli DH10. Transformants were amplified directly from colonies using the PCR protocol described previously. One transformant per sample (verified as producing the correct profile) was sequenced and matched to sequences previously reported using the BLAST function in GenBank (Altschul et al. 1997). Our sequences have been submitted to GenBank under accession numbers DQ294944–DQ294955. The PCR2 products from roots were also sequenced directly without the cloning step in a limited number of cases. The names of our ribotypes, their respective T-RFLP profiles and their matches with spores or GenBank sequences are listed in Supplementary Appendix S1.
Several ribotypes matched Glomus intraradices sequences from GenBank. Gl. intraradices strains have been documented to contain considerable genetic diversity in the ITS region (Jansa et al. 2002), which could be one potential explanation for this result. In other cases, ribotypes from morphologically different spores were within 2 bp of each other and therefore these were grouped within a single ribotype according to the criteria we set. It is possible that HinfI was not sufficient to detect sequence variability among species. Consequently, our various ribotypes may span taxonomic levels and comprise isolates, single species or species groups. We did not consider use of our ribotype designations to be a major limitation in our study for two main reasons: (i) the species concept in this seemingly asexual organism may be of questionable relevance given the considerable genetic and ecological variability that has been observed within a species (Munkvold et al. 2004; Koch et al. 2006); and (ii) our main objective was to compare differences in communities between two soil types. Any shortcomings of our ribotype designations would apply to both soil types.
DNA was successfully amplified from all root samples, with a majority of the PCR2 product lengths within a range (500–620 bp) typical for AM fungi (Aldrich-Wolfe, in press). Of all T-RFLP profiles, 62% matched AM fungal ribotypes, 13% matched non-AM fungal ribotypes and 25% could not be placed in either group with confidence. The great majority of the unmatched profiles (92%) were detected only once, and the remaining 8% (comprising three profiles) occurred between two and four times over several sites. The PCR2 product lengths of unmatched profiles did not differ between the two soil types (F1,143 = 1.86, P = 0.18). Because PCR product length alone appears to be a good indicator of AM fungal family (Appendix S1), our inability to identify all profiles to ribotypes probably did not affect our comparisons between soil types. All unmatched profiles and non-AM fungi were removed prior to subsequent analyses. We considered only members of the Gigasporaceae and Glomeraceae because AM fungi of other families were rarely detected.
glasshouse niche space experiment
Seeds of Sorghum bicolor L. were planted in 164-mL Cone-tainer tubes (Stuewe and Sons Inc., Corvallis, OR, USA) on 14 February 2004. Sorghum was chosen as a host plant for the glasshouse experiment as in our experience it is better suited for growth in Cone-tainers than maize. Each tube contained either sand (3 : 7 : 90 of clay, silt and sand; pH = 6.0), clay (41 : 11 : 48 of clay, silt and sand; pH = 7.8), or a sand/clay mixture (4 : 1, v/v) from one randomly selected clay and sand field from Tsholotsho. The sand/clay mixture resulted in a textural combination of the two soil types (18 : 9 : 73 of clay, silt and sand), with a pH more similar to that of the clay soil (7.6). All three soils were autoclaved for 2 h on two consecutive days before use in the experiment. Based on the family-level differences in the T-RFLP analyses, we inoculated all three soils with either 2.5 mL of a Glomeraceae inoculum (Gl. etunicatum and Gl. mosseae, originally isolated from trap cultures of the sand and clay soils in Tsholotsho) or 17.5 mL of a Gigasporaceae inoculum (Gi. gigantea and Gi. rosea, isolated from trap cultures of the sand soil in Tsholotsho), or a mixture of both inocula (2.5 mL Glomeraceae and 17.5 mL Gigasporaceae inocula). The difference in inoculum volume resulted from differences in spore densities and ensured that approximately 400 spores from each inoculum type were inoculated into each Cone-tainer. Sand was added to each inoculum for Cone-tainers receiving a single inoculum type to bring the total volume added to 20 mL for each Cone-tainer. The AM fungal species were chosen based on their field distributions and ability to sporulate under glasshouse conditions.
Each treatment was replicated 20 times. Half of the replicates were harvested for shoot dry weight and mycorrhizal colonization at 54 days, and half were harvested for spore production at 89 days. However, as a result of poor seed germination in some tubes, the number of replicates harvested varied from 8 to 10. Plants were thinned to a single individual after emergence and grown at 21 °C (± 3 °C) air temperature with 100 µmol m−2 s−1 additional photosynthetically active radiation from supplemental lighting for 16 h per day. Plants were fertilized with 20 mL of a quarter-strength Hoagland solution (Machlis & Torrey 1956) containing 1.5 mg P L−1 at weekly intervals for 7 weeks and twice weekly thereafter. Cone-tainers were arranged randomly with respect to treatment on the glasshouse bench. At harvest, shoots were cut at the soil surface and dried at 65 °C to constant mass. Root colonization by AM fungi was assessed as before and spores were counted under an 80× dissecting microscope following wet sieving (Brundrett et al. 1996).
To compare the occurrences of Glomeraceae and Gigasporaceae in the two soil types in the field, we counted the number of root samples in the five sand fields and the five clay fields for which we detected members of the two families and conducted a chi-squared test of independence with a Yates correction (Zar 1999). A second chi-squared test of independence was conducted to determine if ribotype distribution within a family differed among the five fields in either soil type (Zar 1999). We did not analyse Glomeraceae ribotype distribution in sand or Gigasporaceae ribotype distribution in clay, because their relative frequencies were too low.
To determine the relationship between environmental variables and ribotypes, we first conducted a principal component analysis (PCA), which sequentially identifies hypothetical environmental gradients that account for the greatest remaining variation in taxa dispersion among samples. Only ribotypes with an incidence ≥ 3 was used in these analyses because a majority of T-RFLP profiles with an incidence of < 3 occurred only once and could not be matched with known ribotypes. We then used redundancy analysis (RDA), which constrains the taxa and samples to axes determined by supplied environmental variables (in this case, soil variable; ter Braak & Prentice 1988). PCA and RDA were selected because the average ribotype response to the hypothetical gradient was linear (2.15 standard deviations) rather than unimodal, and because they allow for straightforward variance partitioning, unlike non-metric multidimensional scaling (NMDS). Results generated from NMDS (conducted using Sørensen distances in PC-ORD v. 5.01; McCune & Mefford 1999) were consistent with those generated from the PCA and RDA and therefore are not presented.
The eight soil variables measured (Table 1) were passively projected post-hoc onto the PCA diagram, whereas the RDA was constrained to environmental variables identified as significant by Monte Carlo permutations under the full model (P < 0.05; using a forward stepwise selection procedure with 999 iterations). Ordination analyses and hypothesis testing were conducted using canoco 4.5 (ter Braak & Šmilauer 2002) with log-transformed ribotype data and a focus on interspecies correlations. Biplots were created using CanoDraw 4.12 to display ordination results (ter Braak & Šmilauer 2002). The ribotype scores were post-transformed so that correlations of the ribotypes and soil variables with the ordination axes could be inferred by perpendicular projection.
We conducted a one-tailed Mantel's test to determine whether there was a relationship between distance among sites (based on Euclidean distance) and the similarity of their ribotypes (based on a Sørensen index) and soil characteristics (based on the relative Euclidean index because soil characteristics were measured in different units). The a priori hypothesis was that both ribotype and soil dissimilarity would increase with distance.
We used the methods of Borcard et al. (1992) to determine how much of the ribotype variation was accounted for by (i) spatial variables alone, (ii) soil variables alone and (iii) how much was shared by the two. Spatial predictors were the nine terms of a cubic trend-surface equation based on the x- and y-coordinates of each site (Borcard et al. 1992). This variance partitioning procedure extracts these three variance components and residual variation by first identifying minimal subsets of environmental and spatial predictors (using step-wise forward selection), and then using RDA, partial RDA and subtraction to calculate the components. These analyses were conducted on Hellinger-transformed data (see Legendre & Gallagher 2001), and probability values were calculated using Monte Carlo tests (999 iterations).
Finally, mycorrhizal colonization, spore numbers and shoot dry mass from the glasshouse experiment were analysed separately using the general linear model in Minitab Release 11 (Minitab Inc., State College, PA, USA) with soil type, fungal treatment (grown with a member of the same family only, or with members of the other family) and AM fungal species as factors. As the spores of our four AM fungal species differ in size, we also analysed spore production based on spore biovolumes using average spore diameters for each species given on the INVAM website (http://invam.caf.wvu.edu). Results of these analyses were similar to those based on spore numbers and therefore are not presented.
Transformations were performed, when necessary, to fulfil the underlying assumptions of normality and equal variance. Two means were compared using a two-sample t-test, and three or more means were compared using Tukey's HSD at α = 0.05.