Correspondence: Sven Marhan, Institute of Soil Science and Land Evaluation, Soil Biology Section, University of Hohenheim, Emil-Wolff-Strasse 27, 70599 Stuttgart, Germany. Tel.: +49 711 459 22614; fax: +49 711 459 23117; e-mail: firstname.lastname@example.org
A geostatistical approach using replicated grassland sites (10 m × 10 m) was applied to investigate the influence of grassland management, i.e. unfertilized pastures and fertilized mown meadows representing low and high land-use intensity (LUI), on soil biogeochemical properties and spatial distributions of ammonia-oxidizing and denitrifying microorganisms in soil. Spatial autocorrelations of the different N-cycling communities ranged between 1.4 and 7.6 m for ammonia oxidizers and from 0.3 m for nosZ-type denitrifiers to scales >14 m for nirK-type denitrifiers. The spatial heterogeneity of ammonia oxidizers and nirS-type denitrifiers increased in high LUI, but decreased for biogeochemical properties, suggesting that biotic and/or abiotic factors other than those measured are driving the spatial distribution of these microorganisms at the plot scale. Furthermore, ammonia oxidizers (amoA ammonia-oxidizing archaea and amoA ammonia-oxidizing bacteria) and nitrate reducers (napA and narG) showed spatial coexistence, whereas niche partitioning was found between nirK- and nirS-type denitrifiers. Together, our results indicate that spatial analysis is a useful tool to characterize the distribution of different functional microbial guilds with respect to soil biogeochemical properties and land-use management. In addition, spatial analyses allowed us to identify distinct distribution ranges indicating the coexistence or niche partitioning of N-cycling communities in grassland soil.
The published data are contradictory and depend mainly on the investigated scale. For example Franklin & Mills (2003) postulated that bacterial distribution patterns can be highly structured, even within a habitat that appears to be relatively homogeneous at the plot and field scale. At larger scales, however, Fierer & Jackson (2006) reported microbial biogeography as spatially independent at continental scales.
Grassland soils are of special interest in studying the spatial distribution of microbial communities because of the strong relationships between plants and microorganisms (Niklaus et al., 2006; Bremer et al., 2007) and the heterogeneous influence of grazers on soil properties (Bardgett & Wardle, 2003). Indeed, Ritz et al. (2004) reported that the spatial distribution of microbial communities was influenced by soil nutrient status. Similarly, the spatial distribution of several bacterial phyla in a pasture was correlated to soil properties such as pH, soil moisture or ammonium and nitrate concentration (Philippot et al., 2009a). These studies investigating the spatial distribution of microorganisms in grassland soil have mostly focused on total microbial communities. However, both herbivores and fertilization regimes have been shown to affect N-cycling and the corresponding microbial communities in grassland soils (Frank & Groffmann, 1998; Le Roux et al., 2003; Patra et al., 2005). Within the N-cycle, nitrifying and denitrifying communities are responsible for N-losses through nitrate (NO3−) leaching or greenhouse gas emissions in the form of nitrous oxide (N2O) (Philippot et al., 2007). The first step in nitrification, the aerobic oxidation of ammonium (NH4+) to nitrite (NO2−), can be performed by both archaea and Proteobacteria having amoA ammonia-oxidizing archaea (AOA) or amoA ammonia-oxidizing bacteria (AOB) genes, respectively (Rotthauwe et al., 1997; Treusch et al., 2005). Denitrification, the anaerobic reduction of NO3− to NO2− and to the gaseous N species NO, N2O and N2, is performed by a more diverse group of bacteria and archaea (Tiedje et al., 1989; Zumft, 1997; Philippot, 2002). The spatial distributions of nitrifying and denitrifying microorganisms have been investigated at scales ranging from millimeters (Grundmann & Debouzie, 2000) to the landscape level (Bru et al., 2010). A recent study of Enwall et al. (2010), exploring the spatial patterns of community structure, size and activity of denitrifying bacteria in both an integrated and an organic crop production system, again underlined the role of soil properties in shaping denitrifying communities. Differential habitat selection was observed for the denitrifiers having the NirS- and NirK-type nitrite reductases, with copper being a stronger driver of the abundance of the nirK- than nirS-type denitrifiers, while soil nitrate and clay were unique drivers for the nirS denitrifier community structure. These results suggested niche differentiation between denitrifiers having the two types of nitrite reductases to avoid competitive exclusion.
Although published data clearly indicate the role of plant community composition as well as soil properties as drivers for microbial distribution patterns in soil, little is known about the extent to which the land-use intensity (LUI) of grasslands influences the spatial distribution of microbial communities involved in N-cycling. For grazed grassland ecosystems, Ritz et al. (2004) showed that the intensity of land use, such as the application of fertilizer, affected soil properties and subsequently plant species composition and diversity of soil microorganisms. More recently, Philippot et al. (2009b) found that the intensity of cattle grazing together with soil properties strongly affected the spatial patterns of both the relative abundance and the activity of denitrifying bacteria. The studies mentioned above were conducted on a single site per treatment only. In order to investigate whether the spatial distribution of microbial communities is significantly affected by grassland management, a geostatistical approach on replicated sites is needed.
The objective of this study, therefore, was to investigate whether LUI changes the spatial distribution of microorganisms involved in nitrogen cycling at the plot scale (10 m × 10 m) using replicated grassland sites subjected either to low (unfertilized pastures with sheep grazing) or to high LUI (fertilized meadows, mown 2–3 times per year). We hypothesized that management practices such as fertilizer application and mowing at high LUI sites result in reduced spatial heterogeneity of soil biogeochemical properties and subsequently reduced spatial heterogeneity of microorganisms in comparison with grassland sites of low LUI. We assessed the abundances of total bacterial genes (16S rRNA gene), archaeal and bacterial ammonia oxidizers (amoA AOA, amoA AOB) and denitrifiers (napA, narG, nirK, nirS and nosZ) using quantitative real-time PCR (qPCR). In addition, soil biogeochemical properties potentially controlling microbial abundance were determined. Data were analyzed using a linear mixed model approach with a geostatistical covariance structure. Ordinary kriging was used to map the spatial distribution of ammonia-oxidizing and denitrifying communities of the study sites.
Materials and methods
The sites investigated in this study are located in a limestone middle mountain range, the UNESCO Biosphere region ‘Schwäbische Alb’ in Southwestern Germany. The climate is moderate, with an average annual precipitation of 700–1000 mm a−1 and mean annual temperatures of 6–7 °C. Sites are located between 690 and 810 m above sea level. Soil at the sites is identified as Rendzic Leptosol. An overview of the sites is presented in Table 1. We investigated six grassland sites at two different land-use intensities: (1) unfertilized pastures (low LUI) and (2) fertilized meadows that are mown 2–3 times per year (high LUI). These sites are part of the ‘German Biodiversity Exploratories’ and have been named as AEG 1–3 (high LUI) and AEG 7–9 (low LUI) (Fischer et al., 2010).
Sampling took place in spring (April, 2008), before the beginning of active plant growth. At each of the six grassland sites, bulk soil cores from 0 to 10 cm depth were taken using core augers (Ø 58 mm), and surface vegetation was removed. Samples were collected from a total area of 10 × 10 m per site. A grid mesh with 2.5 m distances was laid over each of the six sites and soil samples were taken starting at each grid point (Supporting Information, Fig. S1). Spatially randomized sampling distances, starting from each grid point and diminishing from 150, 100, 50, 25 to 12.5 cm, resulted in 54 soil cores per site for laboratory analyses. An additional sample directly adjacent to each soil core was collected to determine bulk density. Soil cores were packed in plastic bags and stored at −24 °C for further analysis. Before analyses, roots, stones and soil macrofauna were removed, and soils were sieved (<5 mm) and homogenized.
Soil biogeochemical properties
Soil pH was determined in 0.01 M CaCl2 [soil to solution ratio (w/v) 1 : 2.5]. The soil water content (SWC, expressed as % soil dry weight) was determined gravimetrically for each sample after drying at 105 °C for 24 h. Soil organic C (Corg) and total N (Nt) contents were measured using an elemental analyzer (Leco C/N 2000, Leco Corporation, St. Joseph). Ammonium (NH4+) and nitrate (NO3−) were extracted with 1 M KCl [soil to extractant ratio (w/v) of 1 : 4], shaken on a horizontal shaker for 30 min at 250 r.p.m. and centrifuged for 30 min at 4400 g. The concentrations of NH4+ and NO3− were measured on an autoanalyzer (Bran & Luebbe, Norderstedt, Germany). Soil microbial biomass carbon (Cmic) and nitrogen (Nmic) were determined using the chloroform fumigation extraction method (Vance et al., 1987) in 5 g soil subsamples and extracted with 20 mL 0.5 M K2SO4 on a horizontal shaker for 30 min at 250 r.p.m. and centrifuged for 30 min at 4400 g. A second sample remained nonfumigated, but was treated identically otherwise. C and N in supernatants were measured on a Dimatoc 100 DOC/TN-analyzer (Dimatec, Essen, Germany) and Cmic and Nmic were estimated using the conversion factors 0.45 (Joergensen, 1996) and 0.54 (Joergensen & Müller, 1996), respectively. Extractable organic C (EOC) and N (EN) were calculated from the C and N concentrations in the supernatants of the nonfumigated samples.
DNA was extracted from a homogenized soil subsample (0.2 g) using the FastDNA® SPIN for Soil Kit (MP Biomedicals, LLC, Solon, OH) according to the manufacturer's instructions. Quantification and quality evaluation of the extracted DNA were specified using the NanoDrop® ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA).
qPCR was applied to assess the abundance of 16S rRNA genes as a proxy for the total bacterial community, the ammonia-oxidizing community by targeting genes encoding the catalytic subunits of archaeal and bacterial ammonia monooxygenase enzymes (amoA AOA and amoA AOB) and the denitrifier community by quantification of genes encoding the catalytic subunits of enzymes involved in the denitrification pathway (napA, narG, nirK, nirS and nosZ). PCR reactions were performed according to published protocols. Details are given in Table 2. For all qPCRs, before the experiments, optimal dilutions were tested to avoid the inhibitory effects of coextracted humic acids and other substances.
For a realistic estimation of soil biogeochemical properties in terms of geostatistical analyses and microbial abundances, all data were expressed on an area basis of soil for the top 10 cm soil layer (m−2) (Bolton et al., 1990, 1993; Doran & Parkin, 1996). Geostatistical analyses were performed to interpret the spatial distribution of environmental properties as well as the abundance of ammonia oxidizers and denitrifiers in soil. We analyzed 18 different soil chemical and microbial parameters in total. Differences in heterogeneity among the two LUIs according to the above parameters were investigated by fitting two linear mixed models with geostatistical covariance. The full linear mixed model that we considered fits a separate spatial covariance structure to each level of LUI, i.e. a separate nugget effect, a separate sill and separate range. Thus, there are three additional parameters fitted compared with the reduced model. We used a restricted maximum likelihood approach to fit both mixed models. The fixed effect part of both mixed models is equal. We were therefore able to use a likelihood ratio test (LRT) to infer whether the covariance structure of the full model fitted significantly better than the reduced model (Schabenberger & Pierce, 2002).
The linear mixed model for one specific variable can be written as:
where yijk represents the kth measurement of a soil sample coming from the jth grassland site, which was managed at the ith intensity class. The term δi references the ith fixed effect for treatment (LUI), pij represents the ijth random plot effect, which is distributed as , tijk is the spatial trend effect, and the residual errors terms ɛijk are distributed as . In spatial modeling, the error variance δR2 is usually denoted as a nugget effect. The spatial trend effect tijk is used to model the covariance of each pair of observations (m, n) with locations (xm, ym) and (xn, yn) depending on their separation distance
The covariance function can be written as
where σV2 denotes the variance parameter, which is usually called the sill or the scale parameter, and ρ(h) corresponds to the correlation function, which determines the spatial dependency among observations as a function of distance h. We considered the Gaussian, exponential and spherical models (Schabenberger & Pierce, 2002) and confined the analysis to that spatial model for which we obtained the best-fitting full model.
We used the statistical software sas version 9.2, specifically the MIXED procedure, to fit both models for each variable. The resulting P-values for the LRT were adjusted for multiple testing using the Bonferroni-correction method, which controls the family-wise error rate. It is rather conservative, i.e. differences have to be greater in order to be declared significant (SAS Institute Inc., 1999).
The sampling locations were randomly chosen and revealed a rather coarse picture of the existing conditions regarding a specific variable. In order to smooth the data, we used the covariance parameter estimates obtained from fitting the linear mixed models, and used them as parameters for an ordinary kriging (OK) procedure (PROC KRIGE2D of the sas system), which yielded estimates of yijk−eijk.
This procedure is equivalent to the best linear unbiased prediction of the same quantity based on mixed model (1) (Robinson, 1991). We used KRIGE2D instead of the MIXED procedure for kriging because of computational speed. The spatial covariance model determines the way in which optimal weights are calculated for kriging (Isaaks & Srivastava, 1989). Spatial dependence and autocorrelation were described by a distinct set of spatial parameters that were estimated for each LUI when the model was able to separate between the two LUI. The ratio of partial sill to total sill (p-Sill/Sill) was expressed as percentage and used to classify spatial dependence. A ratio of <25% indicated weak spatial dependence, 25–75% indicated moderate spatial dependence and >75% indicated strong spatial dependence (Cambardella et al., 1994). The practical range (p-Range) is expressed in meters and was used as an indicator for the scale of spatial autocorrelation, i.e. high spatial autocorrelations indicate reduced spatial heterogeneity of a certain property at the investigated scale.
Spearman correlations were calculated including both LUIs and separately for low and high LUI to determine pairwise relationships between the parameters. The resulting P-values were adjusted for multiple testing using the Bonferroni-correction method as described for the LRT. Both correlation analysis and plotting of kriged maps were performed using the software r (R Development Core Team, 2008), version 2.9.1. To test differences between the two LUI, one-way anova were performed for soil biogeochemical properties and gene abundances with the fixed factor ‘LUI’ using statistica 6.0 (StatSoft Inc., Tulsa, OK). Transformation of data has been carried out with appropriate means to meet the requirements of a normal distribution and homogeneity of variance if required.
Soil biogeochemical properties
The average values of soil biogeochemical properties for each plot are given in Table S1. Bulk densities ranged from 0.64 to 0.83 g soil dry weight cm−3, with both minimum and maximum values found in low LUI sites, and soil pH values ranged from 6.4 to 7.1. SWC did not significantly differ between low and high LUI (P=0.878). Neither bulk density nor pH differed significantly between the two LUI. In high LUI, total nitrogen tended to be higher (P=0.088) and NO3− amounts were significantly higher (P=0.023), while Cmic was significantly higher (P=0.029) and Nmic tended to be higher (P=0.081) in the high LUI sites.
Abundances of total bacteria, ammonia oxidizers and denitrifiers
Bacterial 16S rRNA gene copy numbers, which were used as a proxy for estimating the number of total bacteria, ranged between 5.8 × 1015 and 1.7 × 1016 copies m−2 (Fig. 1) (equivalent to 9.5 × 109 and 2.1 × 1010 copies g−1 dry soil, Table S2). Comparison of the abundance of total bacteria between LUIs using the data expressed on an area basis did not show any significant differences (Fig. 1). In contrast, AOA had a significantly higher abundance (P=0.022) in the high LUI sites (Fig. 1). The same trend was observed for AOB, but the difference was not significant (P=0.065). Among the denitrification genes, LUI had a significant effect (P=0.036) only on napA gene abundance, with 8.0 × 1014 and 4.1 × 1014 gene copies m−2 in the low and high LUI sites, respectively.
The proportions of the different N-cycling bacterial communities within the total bacterial community expressed as a percentage of a specific gene to 16S rRNA gene copy numbers did not show significant differences between LUIs (data not shown).
Spatial distribution of biogeochemical properties and N-cycling communities
The results of geostatistical analyses calculated per area are presented in Table 3. For most of the parameters, either the exponential or the spherical spatial model was applied, while Corg was best explained by the Gaussian model. The percentage of structural variance indicated that the majority of the data displayed at least moderate spatial dependency at different scales. The full model significantly separated low from high LUI for pH, Corg, Nt, NO3−, Cmic, Nmic, Cmic/Nmic ratio, total bacteria and ammonia oxidizers. For the denitrifying community, the full model best characterized the abundance of nirS, while the reduced model provided the best fit for napA, narG, nirK and nosZ. Spatial autocorrelations (p-Range) of the different N-cycling communities ranged from 0.31 m for nosZ to 159 m for nirK (Table 3). In most cases, spatial autocorrelations were within the sampling area (largest possible distance within the grid=14 m). In general, biogeochemical parameters (pH, Corg, Nt and NO3−) displayed similar (NO3−) or higher spatial autocorrelations in high than in low LUI sites. In contrast, the distribution of the communities studied explained by the full model displayed higher spatial autocorrelations in low LUI. This was reflected by the kriged maps of ammonia oxidizers showing that the distribution of both AOA and AOB was patchier in the high than in the low LUI (Fig. 2). Furthermore, comparison of the kriged maps for AOA and AOB indicated similar distributions of the two groups of ammonia oxidizers in some of the sites. A similar pattern was observed for the spatial distributions of the napA and narG genes (Fig. 3). The spatial distributions of the highest abundances of nirK and nirS denitrifiers in low LUI did not overlap (Fig. 4). The maps of nirS in high LUI (Fig. 4) as well as of nosZ (maps not shown) denitrifiers in both LUIs showed high patchiness, reflecting spatial dependence at a small scale.
Table 3. Results of geostatistical analyses for soil biogeochemical properties and gene copy numbers related to area (m−2)
Structural variance (%)
p-range (m) P<0.05
The full model geostatistically separated low from high LUI. No separation between LUI was possible with reduced (red) models.NA, not applicable; p-Sill, partial Sill; sph, spherical model; gau, Gaussian model; exp, exponential model.
Correlation between biogeochemical soil properties and abundances of ammonia oxidizers and denitrifiers
Spearman correlations were calculated including both LUIs together (Table S3) as well as separately for low (Table S4) and high LUI (Table S5) in order to determine whether different LUIs affect the relationships between parameters. When the two LUIs were considered together, a high correlation (Spearman's correlation rs>0.7 or rs<−0.7) was observed between the abundance of AOA with the concentration of Nt, AOA with NO3−, and of AOB with NO3−. The abundances of AOA and AOB were also highly correlated (rs=0.905) (Table S3). When separately computing Spearman correlations for low and high LUI, correlations between the abundances of ammonia oxidizers and total nitrogen and nitrate concentrations were still evident (Tables S4 and S5). The abundances of both AOA and AOB were also correlated to pH in high (Table S4), but not in low LUI (Table S4). Both AOA (rs=0.738) and AOB (rs=0.739) were positively correlated to SWC only in low LUI. In low LUI, nirK was negatively correlated with pH and nirS was positively correlated. Denitrifiers were positively correlated to Nt (nirS) and EN (nosZ) only in high LUI (Table S5).
The sites differed in mowing practices and in whether or not they received fertilizer (Table 1), representing low (unfertilized pastures) and high LUI (fertilized mown meadows) grasslands. The inclusion of three independent sites per LUI enabled us to determine whether LUI significantly affects the spatial patterns of soil biogeochemical properties and of nitrifying and denitrifying microorganisms in soils. Because sites characterized by the same soil type (Rendzic Leptosol) and climatic conditions were selected, management practices were likely the main factor responsible for the differences between treatments.
For most soil biochemical properties, a higher range of spatial autocorrelation, indicated by the p-Range, was observed in the high LUI sites, which indicated reduced spatial heterogeneity. Although bulk density itself did not differ significantly between the two LUIs, both the lowest and the highest bulk densities were found in low LUI sites. Together, these findings support our hypothesis that fertilizer applications and mowing practices at high LUI sites reduced spatial heterogeneity.
Despite high variations between the sites, both archaeal and bacterial ammonia oxidizers were more abundant in the sites of high LUI (Fig. 1 and Table S2). This is in accordance with a study of Hermansson & Lindgren (2001), who reported AOB as being two to three times more abundant in fertilized arable soils than in unfertilized soils. In contrast to soil biogeochemical properties, the ranges of spatial autocorrelation for both AOA and AOB were larger in low than in high LUI sites, resulting in higher patchiness in high LUI sites. In the case of the AOA and AOB, we therefore reject the second part of our hypothesis that high LUI reduces the spatial heterogeneity of soil microorganisms, as results indicated increased spatial heterogeneity at high LUI sites.
Our findings suggest that a combination of factors such as soil structure, microclimate and oxygen status is driving the spatial distribution of AOA and AOB rather than the soil biogeochemical properties investigated.
At the six grassland sites, the average abundance of archaeal ammonia oxidizers was 44 times higher than that of bacterial ammonia oxidizers. Similarly, most studies on ammonia oxidizers in terrestrial ecosystems have reported that AOA were more abundant than AOB (Leininger et al., 2006; He et al., 2007; Nicol et al., 2008). An increasing body of literature has suggested niche partitioning between AOA and AOB, with ammonia concentrations and soil pH as the main environmental factors shaping the ecological niches of ammonia oxidizers (Erguder et al., 2009; Bru et al., 2010; Gubry-Rangin et al., 2010; Schleper, 2010). In the present study, most of the sites investigated showed similar spatial distributions of AOA and AOB (Fig. 2, AEG 2, 3, 8 and 9). This indicated the coexistence of the two groups of ammonia oxidizers, which was further supported by significant positive correlations between the abundances of AOA and AOB (Spearman's rs=0.905 without separation of LUI). The six grassland sites were characterized by high ammonia concentrations (1.26–3.97 g N m−2) with small variations in pH (6.38–7.09). Because the factors cited above did not separate the niches of AOA and AOB, we suggest that factors, which are otherwise masked by gradients in ammonia concentration or pH, induce the coexistence of the two populations.
The reduction of NO3− to NO2− can be performed by bacteria harboring either napA, narG or both genes (Zumft, 1997; Philippot & Højberg, 1999). Only the abundance of the napA gene, encoding the periplasmatic nitrate reductase, was significantly higher in high LUI sites. The abundance of the napA gene was also one to two orders of magnitude higher than the abundance of narG, which encodes the membrane-bound nitrate reductase. This suggests that the abundance of bacteria possessing napA may not simply resemble abundances possessing narG (Bru et al., 2007), but can be even more abundant under certain conditions. Similar to AOA and AOB, kriged maps showed a similar distribution of the nitrate-reducing microorganisms having the napA and narG genes (Fig. 3), which was supported by a positive correlation between the two genes (Spearman's rs=0.638 without separation of LUI). Although the correlation between the two genes was higher in high (rs=0.738, Table S5) than in low LUI sites (rs=0.524, Table S4), no differentiation between low and high LUI was possible according to their spatial distribution (reduced model, Table 3). However, it is not possible to determine from our results whether the napA and narG genes co-occurred in the same organisms or whether bacteria harboring either the napA or the narG genes coexisted at the studied sites.
Similar to the ammonia oxidizers, the range of spatial autocorrelation for nirS was larger in low than in high LUI sites, resulting in a higher patchiness in high LUI sites. In contrast to nirS, the distributions of nirK and nosZ were not affected by LUI. The different spatial distributions of nirK- and nirS-type denitrifiers (Fig. 4), together with the negative correlation between the abundances of nirK and nirS (Spearman's rs=−0.356), indicate a niche differentiation of organisms having the nirK and nirS functional genes in low LUI sites. The nitrite reductases encoded by nirK and nirS are functionally homologous (Glockner et al., 1993) and no denitrifying organism harboring both types of nitrite reductases has been reported as yet (Jones et al., 2008). Previous studies (Hallin et al., 2009; Philippot et al., 2009b; Enwall et al., 2010) have discussed the maintenance of two types of nitrite reductases as a result of niche differentiation, thus avoiding competitive exclusion. Furthermore, Hallin et al. (2009) discussed the distribution patterns among nirK- and nirS-type denitrifiers through habitats created by the absence/presence of plants, and both Enwall et al. (2010) and Bru et al. (2010) reported copper as being a strong driver for nirK-type denitrifiers.
Spatial analysis of microbial habitat characteristics and soil microbial communities is a powerful tool to understand not only links between environmental drivers and microbial abundance but also to reveal coexistence or niche partitioning of soil microorganisms. Whereas these links are well established in distinct microhabitats (e.g. rhizo- and detritusphere) (Haase et al., 2008; Poll et al., 2008), our study is one of the first to show that grassland management (e.g. LUI) differentially changes spatial patterns of soil biogeochemical properties and N-cycling microorganisms at the plot scale. Spatial heterogeneity decreased with higher LUI for biogeochemical properties, but increased for N-cycling microorganisms, indicating that spatially structured abiotic or biotic factors that were not taken into account are driving the microbial distribution in our study. Independent of LUI, we also found similar spatial distributions of the bacterial and archaeal ammonia oxidizers, while contrasting distributions were observed for nirS and nirK denitrifiers. This suggests that niche partitioning occurred only between the denitrifiers harboring either the copper or the cd1 heme nitrite reductase. Because LUI also changes plant community composition and diversity, future studies will have to evaluate the impact of different plant species and the quality of their rhizodeposits on the spatial distribution of N-cycling microorganisms.
The work has been funded by the DFG Priority Program 1374 ‘Infrastructure-Biodiversity-Exploratories’ (KA 1590/8-1). Field work permits were given by the responsible state environmental offices of Baden-Württemberg. We thank Kathleen Regan for English spelling corrections.