Soil characteristics more strongly influence soil bacterial communities than land-use type

Authors

  • Eiko E. Kuramae,

    Corresponding author
    1. Institute of Ecological Science, Free University Amsterdam, Amsterdam, The Netherlands
    • Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands
    Search for more papers by this author
  • Etienne Yergeau,

    1. Biotechnology Research Institute, National Research Council of Canada, Montréal, QC, Canada
    Search for more papers by this author
  • Lina C. Wong,

    1. Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands
    Search for more papers by this author
  • Agata S. Pijl,

    1. Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands
    Search for more papers by this author
  • Johannes A. van Veen,

    1. Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands
    2. Institute of Biology, Leiden University, Leiden, The Netherlands
    Search for more papers by this author
  • George A. Kowalchuk

    1. Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands
    2. Institute of Ecological Science, Free University Amsterdam, Amsterdam, The Netherlands
    Search for more papers by this author

Correspondence: Eiko E. Kuramae, Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands. Tel.: +31 317 473 502; fax: +31 317 473 675; e-mail: e.kuramae@nioo.knaw.nl

Abstract

To gain insight into the factors driving the structure of bacterial communities in soil, we applied real-time PCR, PCR-denaturing gradient gel electrophoreses, and phylogenetic microarray approaches targeting the 16S rRNA gene across a range of different land usages in the Netherlands. We observed that the main differences in the bacterial communities were not related to land-use type, but rather to soil factors. An exception was the bacterial community of pine forest soils (PFS), which was clearly different from all other sites. PFS had lowest bacterial abundance, lowest numbers of operational taxonomic units (OTUs), lowest soil pH, and highest C : N ratios. C : N ratio strongly influenced bacterial community structure and was the main factor separating PFS from other fields. For the sites other than PFS, phosphate was the most important factor explaining the differences in bacterial communities across fields. Firmicutes were the most dominant group in almost all fields, except in PFS and deciduous forest soils (DFS). In PFS, Alphaproteobacteria was most represented, while in DFS, Firmicutes and Gammaproteobacteria were both highly represented. Interestingly, Bacillii and ClostridiumOTUs correlated with pH and phosphate, which might explain their high abundance across many of the Dutch soils. Numerous bacterial groups were highly correlated with specific soil factors, suggesting that they might be useful as indicators of soil status.

Introduction

Soilborne microorganisms are key to numerous biological processes such as nutrient cycling, plant nutrition, disease suppression, water purification, and soil structure maintenance (Filip, 2002). Even though our knowledge of the particular organism (species) involved in these key functions is limited, bacteria are known to be influenced by a range of biotic and abiotic factors, such as vegetation (Marschner et al., 2001; Kowalchuk et al., 2002; Weinert et al., 2011), soil characteristics (Hansel et al., 2008; Wu et al., 2008), soil texture (Schutter et al., 2001), land use (Kennedy et al., 2005; Yergeau et al., 2007), geographic distance (Yergeau et al., 2007), and pH (Fierer & Jackson, 2006; Lauber et al., 2008).

Soils are physically, chemically, and biologically heterogeneous, thereby providing a wide range of niches to sustain microbial diversity. More recently, it was shown that small changes in soil pH can have important impacts on microbial succession in abandoned Dutch chalk grasslands (Kuramae et al., 2010). In some ecosystems, soil properties overide the effects of land management (Schutter et al., 2001; Lauber et al., 2008).

The aforementioned soil characteristics are strongly affected by land usage. For instance, intensive agriculture causes physical degradation, such as erosion and compaction, and chemical changes, such as pollution from pesticides. Furthermore, differences in nutrient pools owing to fertilization may lead to organic matter depletion and induce changes in nutrient cycling (Pretty & Shah, 1997; Doran & Zeiss, 2000). All these factors, in turn, have profound effects on soil microorganisms, and it is therefore expected that soils with diverse management systems will contain dissimilar bacterial communities.

In order to increase crop yield, inorganic fertilizers and animal manure are often added to agricultural soils; especially in the Dutch agriculture systems including pasture, tremendous amount of fertilizers have been applied over the past decades. The addition of organic manure has been shown to promote soil microbial activities (Enwall et al., 2007), and the addition of mineral fertilizers has been shown to have an impact on microbial community diversity (Zhong et al., 2010). In the face of current anthropogenic pressure on soil ecosystems, for instance owing to agricultural intensification and climate change, there is a need to better understand the effects of these factors in order to predict and mitigate the impacts of such changes. However, reliable predictions of the potential consequences of perturbations on soil microorganisms and subsequent ecosystem feedbacks are hampered by the lack of baseline knowledge about the distribution, density, and ecology of soil-borne microbial communities. The objective of this study was to investigate the relative importance of various soil factors and land-use regimes on soilborne microbial community composition. To this end, we examined microbial community structure across a range of soils and land-use types throughout the Netherlands with a combination of real-time PCR, PCR-denaturing gradient gel electrophoreses (DGGE) approaches, and phylogenetic microarray analysis (PhyloChips). These three methods were chosen because they give different information on microbial community structure based on the SSU rRNA gene, namely quantification of bacterial and fungal abundance by real-time PCR, bacterial community profile by PCR-DGGE, and bacterial community composition by PhyloChips. Resulting community patterns were then related to soil and land-use characteristics in order to identify the most important drivers of soil-borne bacterial community structure and to identify the bacterial classes, orders, and families that best correlate with specific soil factors.

Material and methods

Experimental design, sampling, and soil analyses

Twenty-five fields, which represent six of the most important land usages in the Netherlands (two deciduous forests, three pine forests, two natural grasslands, six pastures, six conventional arable land, and six organic arable land), were sampled (Supporting Information, Fig. S1) in May 2007. All sampling was conducted within a 2-day period, in the absence of notable weather events, in order to minimize the effects of sampling time. In each field, a central point was selected, and subsequently, four sampling points at 20 m from the central point were chosen to obtain five samples per field (A, B, C, D, and E). Each sample (A, B, C, D, and E) was comprised of five subsamples (A1, A2, A3, A4, A5; B1, B2, B3, B4, B5, etc) consisting of soil cores (8 cm diameter × 30 cm deep) taken randomly within a two-meter radius of each of the five sample points A, B, C, D, and E (see Fig. S2). Soil samples were sieved through a 4-mm mesh to remove stones, roots, and plant material. Equal amounts of each of the five subsamples of a sample were pooled to obtain a composite sample replicate per field, thereby yielding five biological replicates per field (A, B, C, D, and E). Part of each sample was stored at −80 °C for DNA extraction, while the rest kept at 4 °C for physical and chemical analyses. For physical and chemical analyses, equal amounts of each of the five replicates per field were pooled.

Physicochemical characterization of total C, total N, phosphate, organic matter, pH, As, CaCO3, Cd, Cr, Cu, Hg, Ni, Pb, Zn, soil texture, and soil moisture was carried out by BLGG (Bedrijfslaboratorium voor Grond en Gewasonderzoek, Wageningen, the Netherlands, https://blgg.agroxpertus.nl/).

DNA extraction

DNA was extracted separately on each of the five replicates per field using the MoBio Power Soil Extraction kit (MoBio, Carlsbad, CA) with bead-beating (Retsch MM301; Retsch GmbH, Germany) at 5.5 m s−1 for 10 min. Total DNA concentration was quantified on a ND-1000 spectrophotometer (Nanodrop Technology, Wilmington, DE).

PCR-denaturing gradient gel electrophoresis

Bacterial 16S rRNA gene-specific PCR and subsequent DGGE were carried out as previously described, using a D-Code Universal Mutation Detection System (Bio-Rad, Hercules, CA) (Yergeau et al., 2007). Banding patterns were normalized with respect to standards of known composition as well as samples loaded across multiple gels. The validity between comparisons was tested by examining the grouping of samples run across multiple gels, which revealed tight grouping of replicates as opposed to grouping according to gel (not shown). Community profile banding patterns were analyzed using the image master 1D program (Amersham Biosciences, Roosendaal, The Netherlands) as described by Yergeau et al. (2007). The resulting binary matrices based on 66 DGGE-detected band positions were used in statistical analyses as ‘species’ presence/absence matrices.

Real-time PCR

Real-time PCR quantifications for bacteria were performed using primers and cycling conditions for 16S rRNA genes as previously described (Fierer et al., 2006), while 18S rRNA gene copies for fungi were quantified according to Lueders et al. (2004). Real-time PCR quantifications were carried out on soil DNA using ABsolute QPCR SYBR Green mixes (AbGene, Epsom, UK) on a Rotor-Gene 3000 (Corbett Research, Sydney, NSW, Australia) as previously described (Yergeau et al., 2007).

Sample preparation for PhyloChip analysis

For PhyloChip analyses, the five replicate DNA extractions from a single field were pooled, providing a single representative DNA sample per field. However, in order to examine within-site variation, seven fields (1F, 4F, 8F, 10F, 13F, 16F, and 25F) representing the different land-use types in this study were assessed in three of the five field replicates. 16S rRNA gene amplification was carried out by bacterial-specific primers, 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′). PCR amplifications were carried out with 1× Ex Taq buffer (Takara Bio Inc, Japan), 0.8 mM dNTP, 0.02 units mL−1Ex Taq polymerase, 0.4 mg mL−1 BSA, and 1.0 mM of each primer. Three independent PCR amplifications were carried out with annealing temperatures of 48, 51.9, and 58 °C, with an initial denaturation step at 95 °C for 3 min, followed by 25 amplification cycles with denaturation at 95 °C for 30 s, annealing for 30 s, and extension at 72 °C for 60 s, followed by a final extension at 72 °C for 7 min. PCR products were pooled, and a 2-μL subsample was quantified on 2% agarose gel. The volume of the pooled PCR products was reduced to less than 40 μL with micrometer YM100 spin filters (Millipore, Billerica, MA). Regression analysis confirmed that the quantity of PCR amplicon applied to the array was not correlated with any organism abundances as estimated by fluorescence intensity of hybridization (data not shown).

Phylochip processing, scanning, and normalization

The pooled PCR products described earlier were spiked with known concentrations of amplicons derived from yeast and prokaryotic metabolic genes. This amplicon mix was fragmented and biotin-labeled using the GeneChip DNA labeling reagent (Brodie et al., 2006). Subsequently, the labeled DNA was denatured at 99 °C for 5 min and hybridized to custom-made Affymetrix GeneChips (16S PhyloChips G2) (DeSantis et al., 2007) at 48 °C and 60 rpm for 16 h in a hybridization chamber. PhyloChip washing and staining were performed according to standard Affymetrix protocols (Brodie et al., 2006). Each PhyloChip was scanned and recorded as a pixel image, and initial data acquisition and intensity determination were carried out using affymetrix software (GeneChip microarray analysis suite, version 5.1).

To account for variation in scanning intensity from array to array, the intensities resulting from the internal standard probe sets were natural-log-transformed. Adjustment factors for each PhyloChip were calculated by fitting a linear model with the least-squares method. Resulting PhyloChip adjustment factors were subtracted from each probe set's natural log of intensity.

Background subtraction, noise calculation, and spot detection and quantification were carried out essentially as previously reported (Brodie et al., 2006). A probe pair was considered positive if the difference in intensity between the perfect match and mismatch probes was at least 130 times the square noise value (N). A taxon was considered present in the sample when 90% or more of its assigned probe pairs for its corresponding probe set were positive [positive fraction (PosFrac) ≥ 0.90]. Operational taxonomic unit (OTU) richness was simply the number of OTU considered positive for a given sample. Relative OTU intensities were calculated by dividing the average signal of the probes aiming at a given OTU by the total average signal for all the OTUs that were identified as present. The relative abundance values were used directly for most analyses or summed up to the Phylum level.

Statistical analyses

Nonmetric multidimensional scale (NMDS) was used to visualize the differences among different land-use types and soil physicochemical properties owing to microbial community composition. Data from each molecular approach for microbial community analysis, namely real-time PCR, DGGE-PCR, and PhyloChips, were analyzed separately with respect to soil properties. Analysis of similarity (ANOSIM) (primer v.5 software) was used to assess significant differences with respect to land use. This nonparametric permutation procedure uses the rank similarity matrix underlying an ordination plot to calculate an R test statistic. For real-time PCR and PhyloChip data, Bray–Curtis distance index was calculated, whereas for DGGE data, Hellinger distance index was used. Bray–Curtis is a popular similarity index for abundance data, whereas Hellinger distance is used to quantify the similarity between two probability distributions.

Bacterial community structure was related to soil factors using canonical correspondence analyses (CCA) in Canoco (ter Braak & Šmilauer, 2002). DGGE-PCR presence–absence data were used as ‘species’ data, while soil data were included in the analysis as ‘environmental’ variables. Variables having the most significant influence on the microbial community structure were chosen by forward selection with a P < 0.010 baseline. The variables selected this way were then included in a model for which significance was tested with 999 permutations. The CCAs for real-time PCR and PhyloChips were basically the same used for the CCA analysis of DGGE-PCR data, except that absolute values and normalized intensities were used as ‘species’ for real-time PCR and PhyloChips, respectively.

Pearson's correlations were calculated between soil factors (C : N ratio, As, CaCO3, Cd, Cr, Cu, Hg, Pb, Zn, phosphate, total C, total N, organic matter, soil moisture, clay, sand, silt, and soil pH) and normalized OTU hybridization intensities using the ‘multtest’ package in R (version 2.6.0; The R Foundation for Statistical Computing). P values were corrected for multiple testing, using the false discovery rate controlling procedure (Benjamini & Hochberg, 1995).

Results

Soil characteristics

Soil physicochemical characteristics varied according to land use (Table 1). Pine forest soils (PFS) showed lower pH and higher C : N ratios than other soils. Furthermore, arable fields and pastures had higher pH values and higher phosphate than deciduous forest and natural grassland fields. Natural grasslands and pastures had two times more total N than arable fields and pine forest and 1/3 more total N than deciduous forest soils. Natural grasslands field 26F had higher total carbon and organic matter than other fields. Pasture field 19F had more silt, clay, total N, Cr, Ni, and Zn than all other fields.

Table 1. Soil physical and chemical values of 25 fields representing six land-use types in the Netherlands
FieldpHTotal N (mg kg−1)Total P (mg P2O5 100 g−1)Total C (g 100 g−1)C : N ratioOrganic matter (%)CaCO3 (%)Moisture (%)Clay (%)Silt (%)Sand (%)Cd (mg kg−1)Cr (mg kg−1)Cu (mg kg−1)Ni (mg kg−1)Pb (mg kg−1)Zn (mg kg−1)As (mg kg−1)Hg (mg kg−1)
Conventional arable field
7F5.11309822.519.13.90.1110.56.193.4006.3001600
8F5.81267902.116.63.50170.610.988.40.27.611.0015.0281.40.03
9F5.51985813.819.18.30.2161.24.794.10.14.011.0035.0131.90.13
16F7.312171021.19.01.70.21611.627.261.20.119.006.716.0297.50.04
17F7.41741622.313.24.10.7155.615.678.70.113.006.66.1226.80.04
18F7.412921061.29.31.62.41516.137.946.00.129.021.014.014.04111.00.03
Organic arable field
10F5.81338152214.93.60131.810.288.00.27.25.3017.0252.50.03
11F6.2966761.919.72.90110.9693.103.66.406.1191.10
12F6.512752961.411.02.70.11713.429.457.40.219.015.08.421.07110.00.09
13F7.414401371.49.72.43.8188.239.852.00.226.08.210.015.04013.00.04
15F7.313671601.28.82.20.8157.535.257.20.219.026.09.921.0368.80.05
27F6.27141640.912.61.20134.18.887.00.25.30011.0243.60
Pasture
1F6.221423372.712.65.80.2144.116.179.70.35.521.0037.04326.00.05
2F6.124981972.39.24.00161.16.891.90.26.913.009.7282.20
3F7.618581302.815.16.10192.224.373.30.36.815.0014.0342.50.04
19F6.054893104.88.79.40.22936.751.112.10.552.027.033.032.013014.00.08
21F6.624911962.39.24.40172136.842.10.528.047.024.042.09414.00.18
28F5.946102344.18.98.302732.144.423.60.542.026.034.040.011214.00.1
Natural grassland
25F5.52029302.914.35.30251.86.491.90.27.26.23.59.2123.40.06
26F5.74990598.316.616.30559.522.668.00.617.06.48.026.0427.20.08
Pine forest
4F3.71703173.822.36.40170.34.395.40.100015.001.80.04
5F4.11281312.721.14.60141.59.888.80.23.50013.002.40.04
6F3.8814102.024.63.001415.594.0000010.001.60
Deciduous forest
23F3.72805495.017.89.00284.39.885.90.18.203.823.0125.00.09
24F6.31431781.611.23.00164.36.988.80.211.012.06.537.0794.40.12

Bacterial and fungal abundance

Real-time PCR targeting small subunit rRNA genes was used to quantify the relative abundance of bacteria and fungi across the range of soils sampled. In general, bacterial abundance was lower in PFS and higher in deciduous forest soils (Table 2). Fungal abundance was lower in natural grasslands, and relatively high in forest soils, especially in sample 23F. Fungal abundance was significantly correlated with phosphate, while bacterial abundance was not significantly correlated with any measured soil physicochemical factors (Table S1). The fungal abundance was not significantly correlated with total numbers of OTUs given in the PhyloChips (Table S2).

Table 2. Mean real-time PCR quantification of bacterial 16S rRNA genes and fungal 18S rRNA genes for 25 fields
Field #Land useBacterial 16S rRNAFungal 18S rRNA
108 copies g−1 soil FW*107 copies g−1 soil FW
  1. *, fresh weight. The real-time PCR quantification of bacteria and fungi of each field is a mean of five replicates per field.

  2. The means were calculated for each land-use type, and the standard error of the means (SEM) is indicated in parenthesis.

7FArable field conventional9.3110.82
8F9.276.83
9F12.028.71
16F7.575.32
17F22.407.06
18F8.393.19
10FArable field organic11.4016.17
11F10.067.74
12F9.697.14
13F8.734.63
15F11.789.78
27F11.8011.20
1FPasture15.1310.83
2F16.859.02
3F17.117.69
19F9.901.94
21F26.779.87
28F16.033.44
25FNatural grassland9.751.43
26F16.591.34
4FPine forest7.847.15
5F7.224.96
6F6.744.52
23FDeciduous forest32.1619.09
24F20.664.97
Mean (SEM)
Arable field conventional11.49 (2.27)6.99 (1.08)
Arable field organic10.58 (0.52)9.44 (1.63)
Pasture16.97 (2.24)7.13 (1.48)
Natural grassland13.17 (3.42)1.39 (0.04)
Pine forest7.27 (0.31)5.54 (1.10)
Deciduous forest26.41 (7.35)12.03 (4.78)

The total bacterial and total fungal abundance as determined by real-time PCR did not separate the fields according to land use by NMDS analysis (Fig. 1a*). The NMDS analysis gave a good representation of the data (stress = 0.06613, Bray–Curtis distance index). These data were analyzed by a second method, CCA, in order to confirm observed differences, and this additional analysis showed no clustering of the fields according to land use (Fig. 1a**).

Figure 1.

NMDS* and CCA** of sampling sites, soil factors and (a) bacterial and fungal real-time PCR quantification, (b) PCR-DGGE fingerprinting, (c) normalized intensities given in PhyloChips for 25 soil samples representing six different land uses across the Netherlands. The red arrows in CCA (a**, b**, c**) are significant soil factors.

DGGE analysis

A total of 66 band positions were detected across all samples, and almost no variability in banding patterns was observed between replicates (Fig. S3). NMDS analysis of DGGE data showed no clear separation of the fields according to land use (Fig. 1b*). The stress value, based on Jaccard similarity, was 0.21471, indicating that an additional analysis method is necessary for data interpretation. We therefore also performed CCA analysis (Fig. 1b**) with DGGE data and soil physicochemical data, the results of which clearly showed that the main differences in bacterial structure were between PFS and field 23F, and all other soils. C : N ratio, sand, and pH were the underlying soil variables most responsible for the variation as shown on the second axis of the ordination.

OTU richness detected by PhyloChips

PhyloChip analyses detected a total of 2869 different OTUs across the 25 soils examined. Of these OTUs, 362 were detected in all samples. The numbers of OTUs per site ranged from 2207 in the pasture 21F field to 867 OTUs in pine forest field 4F (Table 3). On average, arable soils (conventional and organic), natural grasslands, pasture and deciduous forest soils had 38% to 42% more OTUs than PFS (Table 3). This was mainly attributable to lower numbers of OTUs belonging to the phyla Bacterioidetes, Firmicutes, Nitrospira, and classes Beta proteobacteria and Gammaproteobacteria in PFS. Most OTUs detected belonged to Firmicutes (16–18% of the total) in almost all fields, except in pine forest and deciduous forest soils (Table 3). In pine forests, members of the Alphaproteobacteria (17%) were most detected, while in deciduous forest soils, members of Firmicutes and Gammaproteobacteria had similar contributions (17%) to the total bacterial communities. The total numbers of OTUs detected were positively correlated with pH and Zn and negatively correlated with C : N ratio (Table S2).

Table 3. Total numbers of OTUs detected in PhyloChips and the total numbers of OTUs per phylum or class of bacterial phylum in 25 fields
FieldTotal numbers of OTUAcidobacteriaActinobacteriaBacteroidetesChlorobiChloroflexiCyanobacteriaFirmicutesGemmatimonadetesNatronoanaerobiumNitrospiraPlanctomycetesAlphaproteobacteriaBetaproteobacteriaDeltaproteobacteriaEpsilonproteobacteriaGammaproteobacteriaSpirochaetesUnclassifiedVerrucomicrobia
Conventional arable field
7F15454418810562944305744132271198338192302521
8F13403216310142247254622101791167339172252415
9F14114017510062643259733122061187838178262218
16F2169762411699414937294112629816911148350363439
17F16774622512972745233957212351399039271352723
18F2040792301559455038994102826516710443268353731
Mean169752.8203.7126.56.831.746.3302.07.83.76.218.3235.0138.089.840.8238.531.228.224.5
Organic arable field
10F16055219810663046319844142091408539209302521
11F1590621929862844298947122261478837204272519
12F18166920913473749321946212361539540270312930
13F13954217810542833240757121951068136202212321
15F16005017912863245278946132181208439247332724
27F17656520412263551306843162511458640275312830
Mean1628.556.7193.3115.55.831.744.7293.78.34.25.514.7222.5135.286.538.5234.528.826.224.2
Pasture
1F19486522416873548351946212671549939285363228
2F1415411958762045250742142081097738206132216
3F16984420413373348318927182461358037230302827
19F15515420712062443198957152361308738225342824
21F2207752561641045504059593429218311542317354237
28F169965222113834453228491624013610238172343033
Mean1753.057.3218.0130.87.331.846.5307.38.54.06.719.7248.2141.293.338.7239.230.330.327.5
Natural grassland
25F1741571891248374729582121921915610133265353430
26F150559150109937402228414141981479937203283129
Mean162358169.5116.58.53743.5258.5831316.5208.5151.51003523431.532.529.5
Pine forest
4F867399632526281075311015859663597171810
5F123246152636314117473512194818438154292521
6F93622108484183012652212154795725143151713
Mean1011.735.7118.747.75.025.033.0135.75.72.72.711.3168.773.069.032.7131.320.320.014.7
Deciduous forest
23F1560491728773047259644152221227434279332527
24F2137732471521043493599482830517510742338373936
Mean1848.561209.5119.58.536.5483097.54621.5263.5148.590.538308.5353231.5
Total numbers of OTUs detected 8833122010536150895144640322014657454445748

The NMDS analysis of the PhyloChip data did not show clear clustering of samples based on land use (stress = 0.04673), except for pine forest and natural grassland (Fig. 1c*). CCA analysis clearly grouped pine forest sites and natural grassland sites into distinct clusters, but the remaining land-use fields did not form groups according to land-use type. The main factor that grouped the PFS was the C : N ratio, while natural grassland communities seemed to be most influenced by soil moisture (Fig. 1c**). Particular OTUs belonging to Acidobacteria-4, Acidobacteria-6, and Acidobacteria-7 subgroups (Acidobacteria), Actinomycetales (Actinobacteria), Flavobacteriales, Sphingobacteriales (Bacterioidetes), Bacillales, Clostridiales, Lactobacillales (Firmicutes), Rhizobiales, Rhodobacterales, Sphingomonadales (Alphaproteobacteria), Burkholderiales (Betaproteobacteria), Aeromonadales, Chromatiales, Enterobacteriales, Pseudomonadales, Xanthomonadales (Gammaproteobacteria), and Verrucomicrobiales (Verrucomicrobia) were found only in agricultural soils and not in pine forest or in natural grassland soils (Table 4).

Table 4. Taxa and numbers of OTUs detected in soils of pine forests and field 23F, of natural grasslands (fields 25F and 26F), and of the remaining fields (Others)
PhylumClassOrderPine forests, 23FaNatural grasslandaOthersa
  1. a

    OTUs detected at least in three fields of pine forests and field 23F (four fields in total), in two fields of natural grasslands (two fields in total), and in 14 fields of others (18 fields in total).

AcidobacteriaAcidobacteriaAcidobacteriales012
Acidobacteria-4 003
Acidobacteria-6 008
Acidobacteria-7 001
AcidobacteriaHolophagales010
ActinobacteriaActinobacteriaAcidimicrobiales303
Actinomycetales1036
Coriobacteriales011
Rubrobacterales001
Unclassified103
BD2-10 groupUnclassified001
BacteroidetesBacteroidetesBacteroidales054
FlavobacteriaFlavobacteriales0212
SphingobacteriaSphingobacteriales0017
UnclassifiedUnclassified001
BRC1UnclassifiedUnclassified100
ChloroflexiAnaerolineaeChloroflexi-1a100
Unclassified010
DehalococcoidetesUnclassified010
CyanobacteriaCyanobacteriaChloroplasts002
Oscillatoriales002
UnclassifiedUnclassified010
DeferribacteresDeferribacerUnclassified100
DictyoglomiDictyoglomiDictyoglomales100
DSS1UnclassifiedUnclassified010
FirmicutesBacilliBacillales0330
Lactobacillales1315
ClostridiaClostridiales31024
MollicutesAcholeplasmatales007
Anaeroplasmatales001
Mycoplasmatales001
UnclassifiedUnclassified100
NC10NC10-2Unclassified001
NitrospiraNitrospiraNitrospirales051
OP8UnclassifiedUnclassified010
OP9/JS1JS1Unclassified020
PlanctomycetesPlanctomycetaciaPlanctomycetales026
Proteobacteriaα-ProteobacteriaAcetobacterales101
Azospirillales001
Bradyrhizobiales103
Caulobacterales024
Consistiales030
Devosia001
Ellin314/wr0007002
Rhizobiales0112
Rhodobacterales0014
Sphingomonadales0114
Unclassified213
Verorhodospirilla001
β-ProteobacteriaBurkholderiales135
Hydrogenophilales010
MND1 clone group020
Neisseriales010
Nitrosomonadales001
Rhodocyclales023
δ-ProteobacteriaDesulfobacterales120
Myxococcales102
Desulfuromonadales001
Unclassified010
Desulfuromonadales040
Syntrophobacterales050
ε-ProteobacteriaCampylobacterales024
γ -ProteobacteriaAeromonadales009
Chromatiales115
Enterobacteriales049
Methylococcales001
Oceanospirillales011
Pseudomonadales015
Unclassified103
Vibrionales002
Xanthomonadales008
UnclassifiedUnclassified002
SpirochaetesSpirochaetesSpirochaetales022
TM7UnclassifiedUnclassified001
VerrucomicrobiaVerrucomicrobiaeVerrucomicrobiales016

More detailed analyses were conducted by correlating OTUs intensities given in the PhyloChip with 19 soil physicochemical characteristics (Table 1) measured across all soils. A total of 670 OTUs significantly correlated with at least one of the following soil factors: pH, C : N ratio, phosphate, soil texture (sand, silt, and clay), Cr, Zn, and soil moisture. These OTUs were distributed over most of the common phyla and classes detected in soil such as Acidobacteria, Actinobacteria, Bacterioidetes, Chloroflexi, Cyanobacteria, Firmicutes, Nitrospira, Planctomyces, Spirochaeta, Verrucomicrobia, Alphaproteobacteria, Betaproteobacteria, Deltaproteobacteria, and Gammaproteobacteria.

A total of 425 OTUs were correlated with pH (421 positively and three negatively), 150 OTUs positively correlated with phosphate, 147 OTUs with C : N ratio (124 negatively and 23 positively), 75 OTUs positively correlated with sand, 60 OTUs negatively correlated with silt, 78 OTUs negatively correlated with clay, 88 OTUs negatively correlated with Cr, and 10 OTUs negatively correlated with Zn. Numerous OTUs were co-correlated with pH, C : N ratio, and phosphate.

Because the C : N ratio and pH of PFS were very different from the other fields, we also carried out CCA analysis of microarray data and soil data excluding those samples. The CCA results did not group the fields according to land-use or soil types (Fig. S4). Based on this analysis, the main factor that appeared to influence bacterial community structure was the phosphate content, independently of the land-use type.

Discussion

Effects of land use on bacterial communities

A combination of molecular approaches was used in this study, namely general fungal and bacterial quantity (real-time PCR), fingerprinting (PCR-DGGE), and high-throughput phylogenetic microarrays (PhyloChips), to assess the bacterial community structure of six different common land usages in the Netherlands. Although the three different approaches have different levels of robustness and resolution, none of them revealed a grouping of fields according to the land use only, with the exception of the three PFS (PCR-DGGE and PhyloChips) and one natural grassland (PhyloChips). High C : N ratio, low pH, and low phosphate concentration in pine forest might explain this distinction, because these PFS were more acidic and had lower phosphate than the soils with other soil usages. Soil pH is known to affect microbial community structure (Fierer & Jackson, 2006; Baker et al., 2009; Lauber et al., 2009; Kuramae et al., 2010). The largest numbers of OTUs that correlated with soil physicochemical factors were positively correlated with pH; only three OTUs affiliated to Gammaproteobacteria and Actinobacteria were more abundant in soils with low pH. Indeed, 425 (63%) of the OTUs correlated with soil factors were correlated with soil pH; 154 (23%) of the OTUs that correlated with pH also showed a significantly negative correlation with C : N ratio, which are represented by some taxa belonging to Acidobacteria (subgroup Acidobacteria-4, subgroup Acidobacteria-6, class Acidobacteriales), Bacterioidetes (Bacterioidales, Flavobacteriales, and Sphingobacteriales), Alphaproteobacteria (Sphingomonadales), Gammaproteobacteria (Alteromonadales), and Verrucomicrobia (Verrucomicrobiales). When removing the pine forest fields from the analysis, due to their very different characteristics of C : N ratio, pH, and phosphate as compared to the remaining fields, the CCA analysis still did not group the fields according to land-use or soil types and the main factor that appeared to influence bacterial community structure independently of land use was the phosphate content. The similarity of bacterial community structures in some pasture fields, conventional arable fields, and organic arable fields is probably explained by the fact that these fields have similar pH and phosphate content. Strikingly, soil clay or sand textures did not have a large impact on soil bacterial community structure; phosphate content was even more important when the pine forest fields were excluded from the analysis. Phosphate was also observed to be an important driver of soilborne microbial community succession in chronosequences of abandoned chalk grasslands with neutral pH (Kuramae et al., 2011).

Fungal abundance was highest in deciduous forest soils but not in PFS. The higher abundance of fungi in forest soils is in line with results from Jangid et al. (2008). Interestingly, PFS had very low fungal abundance compared with other soils, including deciduous forests. It may be that the difference in the types of complex organic matter in pine forests contributed to the selection of distinct groups of fungi in these soils. Fungal abundance was also low in natural grasslands. This type of Dutch natural grassland called ‘blauwgrasslands’ is characterized by frequent water logging, creating episodic anaerobic conditions that might explain the low abundance of fungi. Our pine forest samples were rich in humic acids, which are known to inhibit PCR amplification steps in various molecular analyses. To eliminate the possible influence of such PCR-inhibiting compounds, extracts were tested for amplification efficiency by spiking test reactions with known quantities of tester DNAs and comparing product yield with and without environmental DNA extract. No PCR inhibition was observed. DNA yields (17–25 ng μL−1), DNA quality (260/280 nm ratio ranging from 1.7 to 1.9), and real-time PCR results were also highly consistent across the five replicates examined per field.

Effects of soil factors on bacterial community structure

PCR-DGGE and PhyloChips analyses revealed that the C : N was an important factor contributing to the differences in bacterial community structure between PFS and all other soils. Fierer & Jackson (2006) and Fields et al. (2006), who used PCR T-RFLP, found pH to be a highly significant predictor of soil microbial diversity and richness in samples across North America and South America. In contrast, we found C : N ratio to be a very strong predictor of bacterial community composition in the Dutch soil systems studied here. The high C : N ratio in PFS might indicate the presence of more complex organic compounds and N limitation, two factors that may be unfavorable for bacterial and fungal growth. High C : N ratio is typical of soil systems with a high amount of recalcitrant organic matter that is decomposed very slowly. Although these three fields of pine forest are located in different places in the Netherlands (North, Center, and South), they showed remarkably similar bacterial communities, perhaps due to the selective pressures elicited by pine leaf litter. Another remarkable result was that PFS had the lowest number of bacterial taxa (PhyloChips) and lower bacterial 16S rRNA gene copies (real-time PCR) than the other 22 fields across the Netherlands. This fact is explained by a combination of high C : N ratio and low pH. Similar to findings of Fierer et al. (2006) and Fierer & Jackson (2006), the total numbers of OTUs in Dutch soils were positively correlated with soil pH. Soil pH was not an independent variable, being significantly correlated with C : N ratio (Pearson's r = −0.76, P < 0.05).

Representative bacterial groups

Firmicutes, mainly Bacilli and Clostridia, represented the most dominantly detected group in most of the soils target in this study, with the noted exception of the PFS. These results are consistent with previous studies that reported a large proportion of Firmicutes in chalk and slightly acidic grasslands, as determined by this same PhyloChip platform (Kuramae et al., 2010) and quantitative dot blot hybridization (Felske et al., 1998, 2000), respectively. However, the high abundance of members of Firmicutes in Dutch soils is in stark contrast to most studies that have examined soil microbial communities, including soils from North America and South America, Europe, and Antarctica (Zarda et al., 1997; Chatzinotas et al., 1998; Kobabe et al., 2004; Caracciolo et al., 2005; Stein et al., 2005; Fierer & Jackson, 2006; Janssen, 2006; Yergeau et al., 2009). These other 16S rRNA gene-based analyses of soil communities have typically found Alphaproteobacteria, Acidobacteria, and Actinobacteria to be the most dominant groups, as opposed to the Bacteroidetes, Firmicutes, and Planctomycetes, which have typically been found to be less abundant. Firmicutes have, however, been observed as being dominate among bacterial communities in forest soil of Kashmir, India (Ahmad et al., 2009), and in ornithogenic soils in the Ross Sea region of Antarctica (Aislabie et al., 2009). Interestingly, many of the OTUs within the Bacilli and Clostridia classes of Firmicutes were correlated with pH and phosphate, which might explain their high abundance in Dutch soils, which are generally characterized by high phosphate levels that have accumulated over decades of external inputs of inorganic and organic fertilizers (van Bruchem et al., 1999) especially between 1950 and 1980.

The detailed analysis of PhyloChips and soil physicochemical properties may allow for the selection of candidate indicators of Dutch soil conditions at finer taxonomical level than phylum and class. Taxa belonging to Alphaproteobacteria (Acetobacterales, Bradyrhizobiales, and Rhizobiales), Deltaproteobacteria (Bdellovibrionaceae, Dessulfobulbaceae, Nitropiraceae, and Syntrophobacteraceae), Betaproteobacteria (Alcaligenaceae, Burkholderiacea, and Oxalobacteracea), Gammaproteobacteria (Chromatiales), Chloroflexi (Anaerolineae), and Actinobacteria might be possible indicators of soils with high C : N ratios (> 18) and low pH (< 4.0). Likewise, some taxa within the Alphaproteobacteria (Sphingomonadales and Rhodobacterales), Gammaproteobacteria (Alteromonadales), Betaproteobacteria (Comamonadaceae), Deltaproteobacteria (Synthrophaceae), Bacteroidetes (Flavobacteria, Sphingobacteria, and Prevotellaceae), and Acidobacteria (Acidobacteria subgroups 4, 6, 7, 9) are indicative of fields with high pH (> 6.5) and low C : N ratio (< 12.5). Several of these associations are highly coherent with previously published results. For example, Rhizobiales and Sphingomonadales have previously been observed to respond differentially to the rhizosphere inputs of different plants (Haichar et al., 2008), and we found that these groups were associated with either natural (Rhizobiales) or agricultural (Sphingomonadales) ecosystems. Furthermore, the abundance of Bacteroidetes was shown by Nemergut et al. (2008) to increase in fertilized soil, and Percent et al. (2008) found this group to be positively correlated with soil pH. In the present study, we also found that several taxa within the Bacteroidetes were more abundant in agricultural ecosystems, as compared to deciduous forest soils. Our results regarding Acidobacteria subgroups 4, 6, 7, and 9 showed these subgroups to be strongly correlated with soil pH, which is in agreement with the findings of Jones et al. (2009).

In summary, our study showed that soil physicochemical factors, in particular C : N ratio, phosphate, and pH, were the main factors explaining the variation in bacterial communities, as opposed to the independent impact of vegetation type and land-use practices. Exceptions were the pine forest and natural grassland sites. Furthermore, in comparing the molecular approaches used in this study, the high-throughput microarray approach proved to be more informative than real-time PCR and PCR-DGGE, as the PhyloChip approach allowed for community assessment at several taxonomical levels, facilitating a fine-scaled and detailed assessment of microbial community composition patterns. Using this approach, we were not only able to discern the important drivers of soilborne bacterial communities, but also able to detect numerous bacterial groups that were indicative of specific environmental conditions across a range of Dutch soils.

Acknowledgements

We thank Y.M. Piceno (Lawrence Berkeley National Laboratory, Berkeley, USA) for laboratory assistance, G.L. Andersen for valuable consultation and guidance with PhyloChip analyses (Lawrence Berkeley National Laboratory, Berkeley, USA), BLGG (Wageningen, The Netherlands) for soil physicochemical analysis, and Bart Pietersen (BDS-BioDetection System), Remy Hillekens (NIOO-KNAW), and Tjalf de Boer (Vrij University of Amsterdam) for help with soil sampling. This work was supported by the Bsik program of ‘Ecogenomics’ (http://www.ecogenomics.nl/) Publication number 5092 of the NIOO-KNAW, Netherlands Institute of Ecology.

Ancillary