Spatial patterns of methanotrophic communities along a hydrological gradient in a riparian wetland


Correspondence: Paul L.E. Bodelier, Department of Microbial Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Droevendaalsesteeg 10, NL-6708 PB Wageningen, The Netherlands. Tel.: +31(0)317473485; fax: +31(0)317473476; e-mail:


Microbial communities display a variety of biogeographical patterns mainly driven by large-scale environmental gradients. Here, we analysed the spatial distribution of methane-oxidizing bacteria (MOB) and methane oxidation in a strongly fluctuating environment. We investigated whether the spatial variability of the MOB community can be explained by an environmental gradient and whether this changes with different plot sizes. We applied a pmoA-specific microarray to detect MOB, measured methane oxidation, methane emissions and soil properties. All variables were measured in a 10 × 10 m, 1 × 1 m and 20 × 20 cm plot and interpreted using a geostatistical approach. Methane oxidation as well as MOB displayed spatial patterns reflected in the underlying flooding gradient. Overlapping and contrasting spatial patterns for type I and type II MOB suggested different ecological life strategies. With smaller plot size, the environmental gradient could not explain the variability in the data and local factors became more important. In conclusion, environmental gradients can generally explain variability in microbial spatial patterns; however, we think that this does not contribute to a mechanistic explanation for microbial diversity because the relevant scales for microorganisms are much smaller than those normally measured.


Microorganisms are the largest source of biomass and diversity on earth, catalysing biogeochemical processes fundamental to ecosystem functioning (Falkowski et al., 2008). In contrast to animals and plants, the spatial distribution of environmental microorganisms has not been as comprehensively investigated (Hughes Martiny et al., 2006). The paradigm that ‘everything is everywhere but the environment selects’, originally termed by Baas Becking in 1934 (reviewed in de Wit & Bouvier, 2006), was at the basis of the widely accepted assumption that microorganisms are omnipresent and not limited by dispersal. Recent work on microbial biogeography demonstrates that spatial patterns for microbial communities can be observed in the range of millimeters to hundreds of kilometers (Grundmann & Debouzie, 2000; Fierer et al., 2009; Bru et al., 2011; Wessen et al., 2011; Hanson et al., 2012; Lindstrom & Langenheder, 2012; Monroya et al., 2012).

A large fraction of the spatial variability of microbial communities can be ascribed to soil physico-chemical properties and their gradients in the environment (Ritz et al., 2004; Fierer & Jackson, 2006; Bru et al., 2011), but also biological factors (e.g. inter- and intraspecific competition) may play an important role in structuring the distribution of microorganisms (Horner-Devine et al., 2007). The scale where these processes shape microbial communities can be very different, that is, some act primarily at the microscale (e.g. soil density, competition), while others are more influential over larger distances (e.g. precipitation, vegetation). Recent studies have found effects of soil properties on microbial communities at multiple scales (Franklin & Mills, 2009; and references therein); however, it is still not fully understood to what extent environmental gradients contribute to explain spatial variability of microbial communities at different spatial scales.

Aerobic methane-oxidizing bacteria (MOB) are key players in the degradation of methane contributing significantly to the regulation of the concentration of this greenhouse gas in the atmosphere (Semrau et al., 2010). Their physiology and diversity are well described (Trotsenko & Murrell, 2008; Lüke & Frenzel, 2011b), and they can be detected by the pmoA gene, which encodes a subunit of a protein catalyzing the first step in methane oxidation. They belong to the Proteobacteria and Verrucomicrobia. Based on physiological, biochemical and phenotypical properties, MOB within the Proteobacteria have been classically separated into type I and type II (Hanson & Hanson, 1996). Today this classification mainly corresponds to the families Methylococcaceae (type 1,Gammaproteobacteria), Methylocystaceae and Beijerinckiaceae (type II, Alphaproteobacteria; Semrau et al., 2010).

In wetlands, which are the largest biogenic source of methane (Ringeval et al., 2011), MOB are typically active at the oxic–anoxic boundary layers, where methane and oxygen gradients overlap. They oxidize methane before it is released to the atmosphere thereby mitigating the release of methane from wetlands on average by 40 % (Frenzel, 2000; Denman et al., 2007). Despite their vital role in the reduction in greenhouse gas emissions, information about spatial patterns of MOB is rather limited. In rice fields and landfill cover soils representing relatively homogenous systems, MOB communities displayed no recognizable large-scale spatial patterns (Krause et al., 2009; Kumaresan et al., 2009). In contrast, in stratified environments, such as a littoral wetland of a boreal lake with a stable environmental gradient, the MOB community displayed clear spatial patterns (Siljanen et al., 2011), strongly driven by the water table height.

In the present study, we investigated MOB community patterns in a riparian floodplain with a temporally and spatially irregular flooding regime. From previous studies, it is known that MOB communities in these soils are very responsive to irregular flooding (Bodelier et al., 2012) and that methane oxidation rates, as well as different groups of MOB, are correlated with an underlying hydrological gradient (Wang et al., 2012). However, an in depth analysis on the spatial distribution of methane oxidation rates in relation to the environmentally important group of MOB at a high phylogenetic resolution has not been performed. In particular, we were interested in the following key questions: (1) Do MOB communities and methane oxidation rates display spatial patterns even in highly dynamic environments? (2) Are there differences in the spatial distribution of different groups of MOB? (3) At what spatial scale is the hydrological gradient insignificant in determining the spatial variability of MOB?

We measured MOB communities, methane oxidation rates and soil parameters in a riparian wetland at three different plot sizes (10 × 10 m, 1 × 1 m, 0.2 × 0.2 m). We analysed the MOB community structure with a pmoA-specific microarray and applied geostatistics to visualize spatial patterns of MOB, distribution of methane oxidation rates, methane emissions and soil properties.

Materials and methods

Soil and experimental design

In November 2006, soil was sampled from the ‘Ewijkse Waard’ (51°88′N, 5°73′E), a riparian floodplain at the River Waal, the largest tributary of the River Rhine in the Netherlands. The study site is the river bank of a small oxbow lake that is still connected to the river. From the water front, the study site gradually increases in elevation to 1.1 m, which results in a gradient of different flooding intensities within the plot. While the higher parts can be flooded up to 2 weeks, the lowest part can be flooded up to 150 days per year (Bodelier et al., 2012). Soil characteristics have been described earlier (Kemnitz et al., 2004; Steenbergh et al., 2010; Bodelier et al., 2012; Wang et al., 2012). In brief, the soil texture was sandy loam at the lowest part, silty clay loam at the middle part and sandy loam at the higher parts. The average pH–H2O values did not differ among sites, with 7.7, 7.7 and 7.8, respectively.

We collected 73 soil cores of 1.8 cm diameter and 5 cm depth from the top soil in a nested plot design (Franklin & Mills, 2003). First, 24 soil cores were sampled from a 10 × 10 m plot (Fig. 1a). Second, another 24 soil cores were collected from a 1 × 1 m plot in the centre of the first plot (Fig. 1b). Finally, we sampled 25 soil cores from a 20 × 20 cm plot in the centre of the second plot (Fig. 1c). In each plot, samples were collected at equal distances along the sides and diagonals. In the following, plots are termed large, medium and small, respectively. We measured the soil parameters, soil water content based on dry weight, soil particle density and total soil organic matter using standard techniques.

Figure 1.

Sampling scheme, illustrating the location of sampling points at the study site.

Methane oxidation and flux measurements

As methane depletion curves in soils often display biphasic kinetics, initial and induced methane oxidation rates have been determined (Steenbergh et al., 2010). The initial methane oxidation rate has been used as a proxy for the active methanotrophic community, while the induced methane oxidation rate has been taken as an indicator for the methane oxidation capacity of the whole methanotrophic community (Steenbergh et al., 2010) present in the soil seed bank (Krause et al., 2010).

Methane oxidation assays have been measured as described in Steenbergh et al. (2010). Briefly, in a 150-mL bottle capped with a rubber stopper, 5 g of homogenized soil was mixed with 10 mL of Milli-Q water (Milli-Q Reagent Water Systems; Merck Millipore, Billerica, MA). To the headspace, methane (1.4 mL) was added to a final concentration of approximately 7.15 μg mL−1. Soil slurries for each sample were incubated on a shaker (120 r.p.m.) at room temperature in the dark. Depletion of methane was measured during the incubation by subsampling followed by analyses using a gas chromatograph (HP 5890 Gas Chromatograph, Hewlett-Packard) equipped with an FID detector.

Fluxes of CH4 were measured at the large scale in vented closed flux chambers (Supporting Information, Fig. S1) using a photoacoustic infrared gas analyzer with a multisampler (Bruel and Kjaer, Denmark) prior to soil sampling. The flux chambers had an inner diameter of 15.2 cm and a height of 24.2 cm and were attached to pre-installed frames to minimize disturbance of the soil during measurement. Within each flux chamber, at least five gas samples were taken over a measurement period of 1 h.

Nucleic acid extraction and microarray analysis

DNA was extracted from 0.3 g of freeze-dried soil following a modified protocol, based on the FastDNA spin kit for soil (MP Biomedicals, Solon, OH) described in detail in Wang et al. (2012). The DNA concentration was quantified using an ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE).

The community composition of MOB was determined with a pmoA-specific microarray (Bodrossy et al., 2003). As it is a high-resolution fingerprint technique, the division of type I MOB can be further subdivided into type Ia comprising along with others the genera Methylomonas, Methylobacter, Methylosarcina and Methylomicrobium, and type Ib with Methylococcus and Methylocaldum (Bodrossy et al., 2003).

The procedure has been described in detail (Stralis-Pavese et al., 2004, 2011). pmoA genes were amplified using the forward primer A189f and the reverse primers T7-A682r and T7-mb661 in a two step-touchdown PCR as described earlier (Pan et al., 2010). From the samples L1, L2, L4, S49, S54–58, S63, S67, S68 and S73, the pmoA gene could not be amplified and analysed with the microarray. By design, the microarray is hierarchical, containing multiple probes covering the same group of operational taxonomic units (Bodrossy et al., 2003), which has been proven very useful in a wide range of studies (Bodrossy et al., 2006; Kumaresan et al., 2009; Siljanen et al., 2011). However, including these probes into the analysis, correlations with other factors may be over or underweighted. Therefore, only selected MOB microarray probes (Table S1) have been used for statistical analyses as suggested in previous studies (Lüke et al., 2011a; Krause et al., 2012).

Statistical analyses

Prior to modelling and mapping soil and MOB community variables, we applied a Mantel test to check the presence of autocorrelation in measured variables (Legendre & Fortin, 1989), that is, to assess whether samples located closer together are also more similar than those separated by larger geographical distances (Ettema & Wardle, 2002). Subsequently, all variables with a statistical significant autocorrelation have been processed for a variogram analysis and kriging (spatial interpolation). First, a semi-variogram was calculated using the Hawkins and Cressie's modulus estimator. In brief, semi-variances between samples were calculated and graphed versus their spatial separation distance. Second, the geostatistical appropriate correlation function was selected. We used weighted least squares (WLS) estimations with the Matérn model to estimate the parameters of the geostatistical model from the observations. Goodness-of-fit of the model parameters was tested by cross-validation. Finally, parameters from the geostatistical model were applied to perform ordinary kriging using global neighbourhood. Normal distribution of data has been tested with the Shapiro–Wilk test. When data deviate from normal distribution indicator, kriging has been applied (Journel, 1983). It follows the same procedure from above but instead of real values thresholds are used, that is, if value is above threshold, then it is replaced by one, else it is replaced by zero. Spatial maps take values between zero and one displaying the probability that the threshold is exceeded or that the variable is present.

To evaluate the relationships between measured environmental variables, methane oxidation rates and selected microarray probes, Spearman rank correlations were calculated. Differences between soil parameters, methane oxidation rates and the MOB community at the small, medium and large sampling plot have been evaluated using analysis of variance (anova) or Kruskal–Wallis rank sum test when data deviated from a normal distribution. Tukey's post hoc test or Mann–Whitney U-test has been used to test individual differences between plot sizes. To demonstrate how variability changed over different sampling scales, we calculated the coefficient of variation, that is, the ratio between the standard deviation and the mean. All analyses were carried out with the vegan package, geoR, ecodist and as implemented in the statistical software r (Ribeiro & Diggle, 2001; Goslee & Urban, 2007; Oksanen, 2011; R Development Core Team, 2012).


Spatial distribution of soil properties

Prior to any spatial analysis, we tested each variable for the presence of spatial autocorrelation at all sampling locations and at each plot size individually. Spatial structure was detected in moisture content, soil organic matter and soil density (Table 1). Analysing each plot size individually resulted in significant spatial structure for soil moisture content at the large scale and organic matter content at the medium scale (Table 1). We further looked into the variability of the data and identified that the coefficient of variation from moisture content and organic matter content decreased from 10 × 10 m to 20 × 20 cm, while it did not for soil density (Table 2). Next, we fitted geostatistical models to soil properties at all sampling locations and calculated the structural variance to indicate the degree of spatial dependency (Table S2). Here, all variables showed strong spatial dependency over 80 %. Finally, interpolated spatial maps (kriging maps) were produced to visualize spatial patterns (Fig. 2). The kriging map of moisture content followed the elevation gradient from the lower part to the higher part of the plot, which was additionally confirmed by a strong negative Spearman correlation between soil moisture and elevation (Table 3). When comparing this distribution with the actual aboveground situation (Fig. S1, Table S3), it reflected different vegetation types at the study site very convincingly (Fig. 2). In contrast, spatial maps of soil density and soil organic matter revealed significantly higher values within the middle of the plot (Fig. 2, Table S4).

Table 1. Mantel r statistics of measured variables over all spatial scales. A significant Mantel r statistic indicates that samples located more closely together are more similar than those further apart. r Values are between −1 (strong negative correlation) and +1 (strong positive correlation). An r value of 0 indicates no correlation. Euclidean distances were used for matrix calculation. To test for significance, we used 10 000 permutations. Values in bold are significant at P < 0.05
VariableMantel r statistic
All plots (n = 73)Large plot (n = 24)Medium plot (n = 24)Small plot (n = 25)
Soil factors
Soil moisture (dry weight) 0.36 0.27 0.020.05
Organic matter content 0.11 0.16 0.28 0.05
Soil density 0.14
Methane oxidation/emission
Initial rate 0.12 0.36 0.090.01
Induced rate 0.19 0.14 0.010.05
Emissiona 0.37 n.a.n.a.n.a.
 All plots (n = 61)Large plot (n = 22)Medium plot (n = 24)Small plot (n = 15)
  1. a

    For technical reasons, methane emission measurements were measured only at the large sampling scale (n = 15).

MOB community
Shannon index0.
Evenness 0.19
Type Ia
Methylobacter sp. LW120.
Methylobacter sp. LW12 related 0.19
Methylomonas 0.18
Methylosarcina sp. LW140.06 0.18 0.070.03
Type Ib
Methylococcus related marine and freshwater sediment clones0.
Type II
Methylocystis strain M and related0.
Methylocystis B (parvus/echinoides/strain M)0.05 0.20 0.030.09
Methylosinus 0.04 0.22 0.040.08
pmoA-20.01 0.18 0.010.04
Methylocapsa and related clones 0.31
RA21 clone0.
Table 2. Coefficient of variation (standard deviation/arithmetic mean) for different measured variables expressed in percentage. Only variables were tested, which were significantly autocorrelated. Methane emissions were only measured at the large scale, and therefore, no values are available for smaller scales
Soil factorsLarge scale (n = 22)Medium scale (n = 24)Small scale (n = 15)
Soil moisture (dry weight)11.624.763.85
Organic matter content15.9714.2510.14
Soil density12.6511.1012.43
Methane oxidation
Initial rate108.9455.6546.25
Induced rate35.4123.4328.24
MOB community
Table 3. Spearman correlations of soil variables, methane oxidation rates, methane emission and evenness. Only variables were tested, which were significantly autocorrelated. Values in bold are significant at P < 0.05
  1. a

    Correlations from emission (n = 15).

Soil factors
Moisture −0.57       
Organic matter0.22 0.35      
Soil density0.2 0.26 0.25    
Methane oxidation
Initial rate −0.52 0.44 0.230.11   
Induced rate0.  
Emissiona 0.56 −0.43 0.070.28 −0.62 0.25 
Diversity of MOB
Evenness 0.28 0.16 −0.26 −0.72
Figure 2.

Kriging maps of soil parameters, methane oxidation and methane emission. Kriged maps have been created applying ordinary kriging using global neighbourhood (OK) or indicator kriging (IK) when variable deviated from normal distribution. Colour bar to the right of each map indicates the range of given variable for OK. In case of IK, colour bars indicate the probability of a variable to exceed the threshold value (above the mean or below the mean). Geostatistical model parameters are summarized in Table S2.

Spatial patterns of methane oxidation rates and methane emission

Both methane oxidation rates and emission measurements were autocorrelated (Table 1). While spatial dependencies of methane oxidation rates were comparable to soil properties, methane emissions showed only 7 % spatial dependence (Table S2). When comparing the coefficient of variation, we detected a continuous decrease in the variation of methane oxidation rates with decreasing plot size (Table 2). Kriging maps indicated that highest initial methane oxidation rates were located where moisture content values were highest (Fig. 2). In addition, the NE part of the plot displayed the strongest methane emissions but simultaneously low or absent initial and induced methane oxidation rates (Fig. 2). Initial methane oxidation followed the spatial distribution of the moisture gradient, which was supported by a strong positive Spearman correlation of initial methane oxidation and moisture content (Table 3). Induced methane oxidation rates followed the same trend (Fig. 2), but were not significantly correlated (Table 3). Where initial methane oxidation rates were high, methane emissions were low, and vice versa. This contrasting pattern was underlined with a negative Spearman correlation (Table 3).

Spatial patterns of MOB

The MOB community was assessed with a pmoA-specific microarray (Fig. S2). To describe the MOB diversity, we focused on richness, evenness and the Shannon index. Only evenness was spatially correlated over all sampling scales with a spatial dependence of 89 % (Table S2), but not at any plot size individually (Table 1). The spatial interpolation map revealed two strong drops in evenness in the centre and the NE corner of the large plot (Fig. 3).

Figure 3.

Kriging maps of species/genus/cluster and evenness of total MOB communities detected by the microarray (Fig. S2). Only spatial maps are displayed for MOB, which showed spatial autocorrelation in the Mantel test (Table 1). Kriged maps have been created applying ordinary kriging using global neighbourhood (OK) or indicator kriging (IK) when variable deviated from normal distribution. Colour bar to the right of each map indicates the range of given variable for OK. In case of IK, colour bars indicate the probability to exceed the threshold value (presence or absence). Geostatistical model parameters are summarized in Table S2. For the probes representing Methylobacter sp. LW12 related and Methylomonas, the spatial variability in the data could not fully be described by the geostatistical model and spatial maps should rather be interpreted as presence/absence maps of these groups of organisms.

We also wanted to know the spatial distribution of several MOB as detected by microarray probes (Table S1), which resulted in autocorrelation from both type I and type II MOB representative groups. No autocorrelation was identified for the medium and small plot size (Table 1). Calculated spatial dependencies were comparable to those of soil properties and methane oxidation rates (Table S2). Kriging maps of type Ia MOB represented by Methylobacter sp. LW12 related and Methylomonas depicted a major absence to the middle of the large plot. In contrast, Methylosarcina sp. LW14 displayed a spatial pattern with four abundance hotspots (Fig. 3). The spatial patterns of genera/species of MOB containing an isoenzyme of pmoA (pmoA-2) were only located at the borders of the large sampling plot and displayed highest abundances in the area where other MOB representatives generally were lower (Fig. 3). Spearman correlations only revealed significant positive relationships of Methylosinus with soil parameters and initial methane oxidation rates (Table S5). In addition, there was a negative relationship of Methylocapsa with initial methane oxidation (Table S5). Not all genera/species of MOB were spatially structured but the Spearman correlation pointed out type Ia (Methylobacter) and type Ib (Methylocaldum, LW21) MOB which were either positively or negatively affected by soil moisture content. Similarly, there were correlations of type Ia (Methylobacter) and Ib (LW21, FW1) and RA21 (a clone of uncertain identity) with initial methane oxidation rates (Table S5).


Spatial patterns of environmental properties, methane oxidation and emission

In the present study, we investigated the spatial distribution of MOB and methane oxidation in a riparian floodplain. In particular, we were interested in identifying spatial patterns of MOB and methane oxidation rates in such a dynamic environment.

We first analysed soil properties and showed that moisture content reflected a hydrological gradient as result from slope, distance to the surface water, differences in flooding intensities and vegetation, whereas deviating spatial patterns of organic matter content and soil density led us to conclude that those factors are not directly linked to the flooding gradient. It has been shown earlier that moisture content affects the oxygen and methane availability in the soil and thus is an important factor regulating methane oxidation (Bender & Conrad, 1995).

Additionally, moisture has a direct effect on the soil matrix connectivity with consequences for resource availability and species interactions (Treves et al., 2003). Hence, the congruence in spatial patterns between initial methane oxidation rates and moisture content underlined that this factor, or one associated to it, such as oxygen availability, significantly affects methane oxidation in this soil. In previous work at the same study site, it was demonstrated that methane oxidation was strongly correlated with moisture content as result of different flooding intensities (Bodelier et al., 2012). But also in the littoral zone of a boreal lake, Siljanen et al. (2011) found that both the activity and distribution of MOB were strongly dependent on an underlying hydrological gradient.

Highest methane emissions in the NE corner of the study site were most likely caused by higher plant coverage and roots resulting in a different soil structure with more open pore space where methane, produced in deeper layers by methanogens, passed through. Next to this, the total MOB community in this area showed a rapid change in diversity as shown by a drop in evenness. Here, the community composition mainly differed in type I MOB as shown by the microarray results. Hence, local conditions, for example moisture content, were most probably unfavourable for certain MOB in this part of the study site leading to increased methane emissions. Induced methane oxidation rates also supported the idea that the present MOB could not thrive under the environmental conditions at that time.

Spatial patterns of type I and type II MOB

In general, spatial patterns of individual MOB subgroups did not follow the flooding gradient. This is in contrast to previous spatial analysis using quantitative PCR assays, collectively targeting the abundance of all genera within MOB subgroups from the same soil samples. Abundance of all type Ia and type Ib MOB, as assessed by separate assays, was found to be negatively correlated with the flooding gradient and followed roughly the distribution of the moisture content (Wang et al., 2012). The observed differences in the two studies may be due to the restriction of quantitative PCR to the analysis of the main phylogenetic groups of MOB, while the microarray analysis has a much higher phylogenetic resolution. Hence, fine-scale patterns of genera or species contributing in a minor way to the overall quantitative PCR signal cannot be elucidated by this technique.

Another example of patterns missed by quantitative PCR analyses is the broad occurrence of Methylosarcina and Methylosinus, which may be caused by the capability to produce resting stages or in the case of Methylosarcina a diffusive slime layer (Hanson & Hanson, 1996; Wise et al., 2001). This might have an advantage under changing environmental conditions as in an irregular and fluctuating flooding gradient. With respect to Methylosinus, being the only probe positively correlated with the observed methane oxidation rates, a possible explanation lies in locally favourable conditions for specific genera/species of type II MOB. However, in a similar study at the sampling site using SIP-PLFA, it has been demonstrated that type I MOB contributed most to the measured methane oxidation in this soil (Bodelier et al., 2012).

We also observed deviating spatial patterns of probes representing methanotrophs possessing the pmoA-2 gene and the genus Methylocapsa with probes representing the genera Methylosarcina, Methylocystis and Methylosinus. In this context, the microarray signals for Methylosinus and Methylocapsa were not supported by specific probes for Methylosinus or Methylocapsa, which suggests that next to these organisms, potentially, other unknown MOB may also be responsible for the observed spatial patterns.

The pmoA-2 gene was detected in a variety of type II MOB including Methylocystis and Methylosinus (Tchawa Yimga et al., 2003). Hence, microarray signals from the probe for pmoA-2 (NmsiT-271) may have originated from the same species pool as Methylocystis (McyB304) and Methylosinus (Msi232) signals making an ecological interpretation difficult. However, in a study about the recovery of MOB from disturbance, Ho et al. (2011) observed a rapid increase in MOB containing the pmoA-2 gene after severe disturbance. They suggested that traits other than the high affinity of the enzyme (Baani & Liesack, 2008) may explain their success. This may also be applicable to our case where the occurrence of this probe was restricted the certain areas at the study site.

The overlapping spatial patterns of type I MOB (Methylosarcina) with type II MOB (Methylocystis, Methylosinus) as well as the deviating spatial patterns within type II MOB (Methylocapsa) do not fully support the general idea that both types coexist and inhabit different niches (Henckel et al., 2000). In this context, it has been suggested that types I and II can be ecologically classified according to the r scheme and k scheme (Dobzhansky, 1950) with type I as r strategists investing in reproduction and type II as k strategist investing in longevity by maintaining population size at the carrying capacity of the environment (Steenbergh et al., 2010). However, the spatial patterns observed cannot be captured in these broad two-dimensional life-strategy categories (Bodelier et al., 2012). Recently, Ho et al. (2012) adopted the competitors–stress tolerators–ruderals (C-S-R) framework (Grime, 1979) to conceptualize ecological characteristics of MOB as life strategies. In brief, there are three primary life strategies (competitors, stress tolerators and ruderals) summarizing the response to stress (all factors limiting biomass formation), response to disturbance (all factors leading to biomass destruction) and competitive abilities. This framework offers more ecological depth accommodating mixed strategies and allows classifying MOB with more nuanced distinction between different genera of MOB. In this framework, Methylosarcina can be classified as a strong competitor, while Methylosinus and Methylocystis can be classified as stress tolerators, competitors–stress tolerators or stress tolerators–ruderals (Ho et al., 2012).

The scale effect of the gradient on MOB diversity and function

The results of this study supported the idea that with decreasing plot size, the methane oxidation as well as community structure is governed by factors acting more locally. We demonstrated that the effect of the hydrological gradient on MOB diversity was lowest at the small sampling plot, but differences in the MOB diversity did not diminish. This corresponds well with results from Wang et al. (2012), who used quantitative PCR and showed significantly lower abundance of type Ia in comparison with type Ib and type II in the small sampling plot.

Similarly, initial methane oxidation rates were not spatially structured as indicated by the Mantel r statistic and did not correlate with the flooding gradient at smaller plot scales (data not shown). This means that the variability in the MOB community and methane oxidation rates at smaller scales could not be explained by the environmental gradient, and differences were also a result of other factors not directly linked to the flooding gradient.

The deviating distribution of MOB in the middle of the plot may have been caused by biotic interactions, for example inter- or intraspecific competition. Competition for resources such as oxygen with other functional groups, for example iron-oxidizing bacteria (Wang et al., 2012), is a plausible explanation leading to the observed species-specific distribution of MOB. But it has also been hypothesized that indirect species interaction, for example resource competition of iron reducers with methanogens, can affect the methane availability for MOB (Wang et al., 2012) and consequently their performance in the environment. Next to this, grazing by protists might induce local scale effects on the distribution of MOB (Murase et al., 2006; Murase & Frenzel, 2007). Murase & Frenzel (2010) demonstrated a strong grazing pressure of type I, especially type Ia, by protists. Assuming that the environment is highly structured with different pore space and a different connectivity through the soil pore water, this might have led to fine-scale favourable microenvironments for protists, which in turn might result in strong small-scale differences of type Ia MOB abundances.

In this study, the MOB community and methane oxidation were structured by an unpredictable variability in unknown factors or entangled factors at the metre and centimetre scale. We demonstrated earlier that even within 0.7 mm vertical soil profiles at a resolution of 100 μm different genera of MOB inhabited different niches (Reim et al., 2012). The genus Methylosarcina was mostly detected at the surface layers with high oxygen and low methane concentration and the genus Methylobacter at the oxic–anoxic boundary layer of methane and oxygen. Comparing these results with large-scale spatial studies showing strong dependencies with environmental drivers such as pH (Fierer & Jackson, 2006; Bru et al., 2011), it may be that these common environmental factors explain the distribution of microbial communities at larger scales. However, it remains unclear to what extend this contributes to understand the mechanisms generating and maintaining microbial diversity at the scale relevant for microorganisms.


We demonstrated that even in a highly irregular and fluctuating environment, MOB displayed spatial patterns at scales ranging from centimetres to metres. Groups within the type I and type II MOB differed in their spatial distribution patterns. We were able to provide evidence that the environmental variability observed in this study cannot explain MOB microbial community patterns at smaller sampling scales (< 1 m). Other biotic or abiotic factors play a role; however, the identity of those factors needs to be determined in future studies. We suggest that future studies on the spatial distribution of microorganisms need to be performed on multiple scales to identify the relevant scale for microbial interactions to proceed from rather descriptive to mechanistic explanations for the vast biodiversity.


We thank the Foundation ‘De Ark’ for permitting us to take samples at ‘Ewijkse Waard’. We thank anonymous reviewers for constructive comments that improved this manuscript. This study was part of the ESF-Eurodiversity program (ERAS-CT-2003-98049, 6th EU-framework program) and was financially supported by grants from the Netherlands Organization for Scientific Research (NWO; Grant no. 855.01.108). This work was also part of the European Science Foundation EUROCORES Programme EuroEEFG and was supported by funds from the Netherlands Organization for Scientific Research (NWO; Grant number 855.01.150). This is publication no. 5416 of the Netherlands Institute of Ecology.