Introduction

Biodiversity is generally regarded as a stabilizing parameter in the sustainable functioning of our ecosystems (Cardinale et al., 2006, 2012). The current loss of biodiversity (Mora et al., 2011) has evoked numerous studies focusing on biodiversity and ecosystem functioning (BEF) relationships in eukaryotic communities (Hillebrand and Matthiessen, 2009; Loreau, 2010; Cardinale et al., 2012). However, microorganisms (i.e., bacteria, archaea, protozoa, fungi and algae), comprising a major part of the total biomass of organisms inhabiting earth (Whitman et al., 1998) and representing the largest source of biodiversity (Gans et al., 2005), are absent in the ongoing BEF debates and conservation policy, despite various pleas to do so (Cockell and Jones, 2009; Griffith, 2012). Moreover, microorganisms play critical roles in natural biogeochemical processes (Falkowski et al., 2008) as well as in engineered or managed systems like wastewater treatment plants, agricultural fields and industrial bioreactors providing many services to our societies. A number of gaps in the knowledge of environmental microbial communities are fundamental to the absence of microbes on the global biodiversity conservation agenda (Bodelier, 2011), the inability to link microbial species to the processes they catalyze as well as the assumed high functional redundancy in microbial communities being the most crucial ones (Bodelier, 2011). These issues may be the reason why links between microbial diversity and ecosystem processes have been observed only for processes carried out by narrow groups of microbes carrying out a specific process (e.g., chitin and cellulose degradation) (Wohl et al., 2004; Peter et al., 2011) in diversity manipulation or artificial community experiments. The often-observed saturating species–function curves in these experiments, interpreted as functional redundancy, may be caused by including species not actively contributing to function. For microbial communities in soils, it has been demonstrated that up to 80% of all species present are inactive or in a dormant state (Lennon and Jones, 2011). Another explanation may be a lack of niches for species to be active due to the absence of spatial and temporal environmental heterogeneity in highly reductionist microbial BEF experiments. In fact, the lack of environmental heterogeneity and thus of suitable niches has been put forward as one of the crucial points to address in future BEF experiments in ecology in general (Cardinale et al., 2012).

The best available approach to link microbial species to substrate consumption (i.e., analog for microbial behavior) in complex environmental settings is the tracking of isotopic labels into compounds of taxonomic relevance (e.g., polar lipid-derived fatty acids (PLFAs), RNA, DNA and proteins). In that respect, aerobic methane-oxidizing bacteria (MOB) are a group of environmentally highly relevant microbes that form an ideal model system for performing microbial BEF experiments. MOB have a very restricted substrate range generating their energy and providing in their biomass carbon largely by the utilization of CH4 (Semrau et al., 2010; Stein et al., 2012). Next to this, the special C1 metabolism distinguishes MOB from the activity of other bacteria, which in combination with the possession of unique PLFAs (Bodelier et al., 2009) facilitates the use of stable isotope labeling approaches (Boschker et al., 1998; Bodelier et al., 2012) to capture the active MOB in the environment. The molecular detection of MOB is facilitated by the use of a functional marker gene (particulate methane monooxygenase (pmoA)) that is strongly linked to the phylogeny of aerobic proteobacterial MOB (Holmes et al., 1995; Kolb et al., 2003; Lüke and Frenzel, 2011). The phylogeny of the methane-oxidizing guild has been changed substantially in the last decade. The classical canonical assignment of MOB subgroups (i.e., types I, X and II), which was based on physiology, biochemistry, morphology and phylogeny, does not reflect the current situation any longer considering the discovery of facultative methane oxidizers (Dedysh et al., 2005; Belova et al., 2011) (Semrau et al., 2011), Verrucomicrobial methanotrophs (Dunfield et al., 2007; Pol et al., 2007; Op den Camp et al., 2009) and ‘aerobic methanotrophs’ generating their own oxygen by nitrite reduction (Ettwig et al., 2010) belonging to the NC10 phylum. Hence, currently it is more useful to refer to MOB subgroups based on phylogenetic assignments exclusively (Op den Camp et al., 2009). The well-studied γ- and α-proteobacterial MOB harbor the families of the Methylococcaceae (γ-Proteobacteria), the Methylocystaceae and Beijerinckiaceae both belonging to the α-Proteobacteria. Within these families, a large number of classified genera and species have been described that, on the basis of phylogenetic assignment, are referred to as types Ia (Methylomonas, Methylomicrobium, Methylobacter, Methylosarcina, Methylosoma, Methylovulum, Methylosphaera, Crenothrix and Clonothrix), Ib (Methylococcus, Methylocaldum, Methylogaea, Methylohalobius and Methylothermus) and II (Methylocystis, Methylosinus, Methylocella, Methylocapsa and Methyloferula) (Lüke and Frenzel, 2011; Stein et al., 2012). The application of molecular tools (McDonald et al., 2008) based on the pmoA gene in combination with stable isotope probing of PLFA (SIP-PLFA) and/or DNA/RNA has rendered type Ia, Ib and II MOB as the best-studied group of environmental microbes, even allowing for assigning of life-history strategies to this group of bacteria (Ho et al., 2013).

Aerobic MOB constitute the largest methane sink in terrestrial habitats, especially thriving at oxic–anoxic interfaces in wetlands (e.g., rice paddies, peat lands and floodplains) and surfaces of lake sediments, thereby directly contributing to the regulation of earth’s climate through modulation of atmospheric methane concentrations (Conrad, 2007; Semrau et al., 2010). Methane emission estimates from wetlands are characterized by high uncertainty (Bloom et al., 2010). Recent variation in atmospheric methane concentrations, including the resumed increase from 2007 onwards, has been linked to changes in wetlands (e.g., in surface area, fertilizer use in rice cultivation and microbial processes) (Bousquet et al., 2006; Kai et al., 2011; Spahni et al., 2011). Especially, wet mineral soils (e.g., riparian floodplains), which can serve as a sink as well as a source of methane depending on the hydrological conditions, are proposed to contribute to the 2007 increase (Spahni et al., 2011). The question is to what extent the diversity of MOB communities and the traits of the species therein modulate methane consumption, thereby contributing to variability in methane emission from wetland habitats.

In this study, we investigate the response of aerobic MOB in an irregularly flooded wetland and assess the role of diversity of MOB communities in the modulation of methane consumption in these soils. A combination of molecular biological tools (McDonald et al., 2008) and stable isotope labeling of lipids (Boschker et al., 1998; Neufeld et al., 2007) and proteins (Seifert et al., 2012) is applied to exploit the characteristics of aerobic MOB to serve as a model group in microbial BEF studies in the most optimal way. More specifically, we want to investigate the role of available niches on the diversity–functioning relationship by assessing the active MOB community members in a spatial and temporal investigation carried out in a riparian floodplain that is characterized by a natural gradient of hydrological conditions providing a variable amount of niches for MOB.

Materials and methods

Sampling site

The sampling site was located at Ewijkse Waard (51°88′ N, 5°73′ E), a riparian floodplain along the river Rhine in the Netherlands. The soil properties have been described previously in detail (Kemnitz et al., 2004; Steenbergh et al., 2010). The sampling site is located on the embankment of a small oxbow lake that is connected to the river on one side but is separated from the river by an outflow barrier, ensuring that the oxbow lake does not run dry completely. This outflow barrier results in slow retreat of water from the sampling plot after flooding, which results in a gradient of flooding days within the plot. The highest part is flooded up to 2 weeks per year, whereas the lowest part can be flooded up to 150 days per year. The flooding occurs very irregularly when the water level of the Rhine reaches a threshold level upon which the oxbow lake fills up with water (Supplementary Figure S1). A 10 × 10 m plot was established situated on a slope on the embankment of a small oxbow lake, resulting in an elevation difference of 1.1 m between the lowest and highest point of the plot (Figure 1) (Wang et al., 2012).

Figure 1
figure 1

Information on sampling design and moisture content in the experimental plot. Gravimetric water content (left panel) as function of the location in the plot. The right-hand panel depicts the elevation differences with the 10 × 10 m experimental plot. The cylinders (C1–C5) depict the spatial arrangement of the cores taken monthly. Samples for moisture content were taken in November 2006 in 73 samples collected within the plot.

Full size image

Soil methane emission

Methane emissions were measured in the field in May, July and October 2007 in vented closed flux chambers using a photoacoustic infrared gas analyzer with a multisampler (Bruel and Kjaer, Naerum, Denmark). The flux chambers had an inner diameter of 15.2 cm and a height of 24.2 cm and were attached to preinstalled 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.

Sample collection for in vitro experiments

In September 2006 and May, July and October 2007, five soil cores (7 cm in diameter and 40 cm in length), evenly distributed over the diagonal of the sampling plot, were collected (Figure 1) and transported to the laboratory for further analyses the next day. The soil cores are subsectioned into three layers (0–2, 7–9 and 27–29 cm) and sieved (2 mm) and mixed before further processing. The choice for these layers was based on pilot experiments where the most active methane-oxidizing horizons were determined using methane oxidation assays as described below (data not shown).

Soil physicochemistry

For the samples collected in September 2006, a range of soil physico chemical parameters were measured (i.e., moisture content, water-filled pore space (WFPS), pH (water and KCl), organic matter, total N, total P, granulometric composition, available cations (NH4+, Mg2+, Ca2+) and anions by CaCl2 extraction). These parameters were analyzed as described earlier (Bodelier et al., 2012). In the other sampling months, only moisture, WFPS, pH (water and KCl), organic matter and mineral N (water and 2 M KCl, i.e., NH4+ and NO3−) availability were measured.

Methane oxidation activity and kinetics

Methane oxidation activity was executed in soil slurries as described by Steenbergh et al. (2010) using a headspace methane concentration of 10 000 p.p.m.v. (parts per million volume). This concentration proved to be saturating the methane-oxidizing enzymes in the floodplain samples (data not shown). In short, for each sample, 5 g of the sieved and mixed soil was suspended in 10 ml of Milli-Q water (Milli-Q Reagent Water System, Millipore, Billerica, MA, USA) in a 150-ml flask capped with a rubber stopper. A portion of 1.4 ml pure CH4 was added to obtain a CH4-mixing ratio of approximately 10 000 p.p.m.v. The CH4 concentration in the headspace was monitored during the PMO assays by gas chromatography flame ionization detector (GC-FID) analysis (HP 5890 Gas Chromatograph, Hewlett-Packard, Wilmington, DE, USA). To assess methane oxidation kinetics of the methane-consuming communities in the floodplain samples, identical slurry incubations as above were carried out using seven different initial methane concentrations (100, 300, 600, 1000, 5000, 10 000 and 20 000 p.p.m.v.). Methane consumption rates were calculated based on headspace samples taken during 24 h of incubation. Apparent Km and Vmax were determined using Monod kinetics calculated by applying nonlinear fitting routine included in Sigmaplot 12.3 (Sigmaplot, Systat Software Inc., San Jose, CA, USA). Specific methane oxidation affinity was calculated as Vmax/Km.

Methane production activity

Potential methane production rates were determined as described earlier (Bodelier et al., 2000) following methane production in anaerobic slurry incubations using 10 g of fresh soil in 20 ml of Milli-Q water. Headspace methane measurements in all laboratory assays were measured by GC-FID analyses (HP 5890 Gas Chromatograph, Hewlett-Packard).

DNA and RNA extraction

DNA was extracted as described by Pan et al. (2010), based on the FastDNA spin kit for soil (MP Biomedicals LLC, Solon, OH, USA). RNA was extracted as described by Steenbergh et al. (2010). RNA extraction was executed for the samples collected in 2007 only. Quality and quantity of DNA and RNA were determined spectrophotometrically using the NanoDrop ND-1000 (Nanodrop Technology, Wilmington, DE, USA).

PCR amplification and reverse transcription

Amplicons of the particulate methane monooxygenase gene (pmoA) were produced by a two-step PCR protocol using the primer pairs A189–A682 followed by A189–Mb661 following the conditions as described by Pan et al. (2010). mRNA was reverse transcribed and cDNA was produced as described by Steenbergh et al. (2010) using the Super-Script III Reverse Transcriptase kit (Invitrogen, Carlsbad, CA, USA).

DNA- and mRNA-based microarray analyses

Diversity analyses of the methanotrophic community were executed using a diagnostic microarray approach based on the pmoA gene as described earlier (Pan et al., 2010). For every sample, three independent PCRs were carried out and pooled prior to the in vitro transcription step in the microarray procedure. The normalized signal intensities derived from the scanned microarrays of only the non-redundant probes were used to infer diversity parameters (i.e., Shannon index (log e based), evenness) using the ‘Diverse Routine’ implemented in the Primer-E software (Plymouth Marine Laboratory, Plymouth, UK).

Quantitative PCR analyses of methanotrophic bacteria

Gene copy numbers of three subgroups of methanotrophic bacteria (types Ia, Ib and II) (Supplementary Figure S3) were determined using quantitative PCR (qPCR) of the pmoA gene as described earlier (Pan et al., 2010).

Clone library construction and phylogenetic analyses

Clones were generated from PCR products using a seminested approach (primers A189–A682 followed by A189–mb661) (clones assigned WP1AN, WP1BN). Cloning was performed using the TAs cloning kit (Invitrogen), using a mix of three replicate PCR products for ligation. Of both libraries (WP1AN and WP1BN), 75 clones carrying the insert were randomly selected for sequencing. Clones were sequenced using an ABI 3130 capillary sequencer (Applied Biosystems, Foster city, CA, USA). Clone sequences were compared with the sequences available in public databases, using the BLAST software from the National Center for Biotechnology Information (available at http://www.ncbi.nlm.nih.gov/BLAST/), to determine their phylogenetically closest relatives. pmoA sequences have been submitted in the GenBank database under accession numbers KF256139–KF256263.

Intact cores: ‘in situ’ methane consumption

In July 2009, triplicate soil cores were collected (1.8 cm in diameter and 5 cm in length) at the same five elevation levels within the 10 × 10 m plot as were sampled for the ‘in vitro’ experiments. The cores were transported to the laboratory and incubated intact in the same bottles as used for slurry incubations (see the Methane oxidation activity and kinetics section). In these static incubations, methane was added amounting to 10 000 p.p.m.v. headspace concentration. Methane decrease in the headspace was monitored by GC-FID as described by Steenbergh et al. (2010) to determined actual methane consumption by intact cores. For SIP-PLFAs and stable isotope probing of proteins (protein-SIP) (see below), three times 10 000 p.p.m.v. (i.e., 4.2 ml) methane was consumed before stopping incubation. In case of SIP-PLFA, 30 μl 99% 13C-CH4 (Campro Scientific BV, Veenendaal, The Netherlands) was added to increase the signature of the methane-C to approximately 1400‰. In case of protein-SIP, 70% of the methane in the headspace (7000 p.p.m.v.) was 13C-CH4. For both SIP-PLFA and protein-SIP, five unlabeled control cores were analyzed. Soil cores were freeze dried before extraction of lipids and proteins.

SIP-PLFA

PLFAs were extracted and analyzed as described by Mohanty et al. (2006) using 4 g of freeze-dried soil with a Bligh and Dyer extraction procedure. Compound-specific gas chromatography isotope ratio mass spectrometry (MS) was applied as described by Mohanty et al. (2006) to determine 13C incorporation into individual PLFA. Labeling profiles of soil samples were compared with those of a PLFA database (Mohanty et al., 2006; Bodelier et al., 2009) to determine the active methanotrophic species in these samples.

Protein-SIP

Protein Extraction Protocol and In-Solution Digestion

Two protein extraction protocols were used. First, 2.5 g of soil sample was used for the protocol described by Benndorf et al. (2007), which uses an NaOH extraction step followed by a phenol extraction and an ammonium acetate precipitation. The extracted proteins from the water and inter phase as well as from the phenol phase were operated separately in the following steps. The proteins were washed with an acetone and a subsequent ethanol step and suspended in Tris buffer. An in-solution digestion procedure was applied to all protein fractions. Either all or at least 30 μg of the protein extracts were incubated with 8 M urea, reduced with DTT and alkylated with iodoacetamide. Trypsin was added in a ratio of 1:100, and the digest was incubated overnight at 37 °C. The peptide mixture was desalted using C18-Zip Tips columns. A second protocol using an SDS buffer extraction was processed (Chourey et al., 2010). In this protocol, 5 g of soil was resuspended in a 2% SDS buffer and boiled for 20 min. The proteins in the supernatant were precipitated with 20% TCA overnight. After washing the protein pellet with acetone, the proteins were denatured using guanidine-HCl+DTT. The ammonium bicarbonate-buffered protein extract was then directly used for an in-solution digestion with trypsin overnight. The peptide mixture was desalted using C18-Zip Tips columns.

Mass Spectrometric Analyses

Peptides were reconstituted in 0.1% formic acid and concentrated on a trapping column (nanoAcquity UPLC column, C18, 180 μm × 2 cm, 5 μm, Waters, Eschborn, Germany) with water containing 0.1% formic acid at flow rates of 15 μl min−1. The complex peptide mixture was fractionated using a nanoAcquity UPLC (Waters) equipped with a C18 column (75 μm × 250 mm, 1.7 μm). Chromatography was performed at a flow rate of 0.3 μl min−1 by using 0.1% formic acid in solvents A (100% water) and B (100% acetonitrile), with peptides eluted over 147 min (samples phenol extraction) or 90 min (samples SDS extraction) with a 8–40% solvent B gradient using a nano-HPLC system (nanoAcquity, Waters) coupled to an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, Schwerte, Germany). Continuous scanning of eluted peptide ions was carried out between 300 and 2000 m z−1 (samples phenol extraction) or 400 and 1400 m z−1 (samples SDS extraction), automatically switching to an MS/MS CID mode on ions exceeding an intensity of 2000 (samples phenol extraction) or 3000 (samples SDS extraction).

MS raw data were processed for database search using Thermo Proteome Discoverer software (version 1.2.0.208, Thermo Fisher Scientific). Search was performed by tandem MS ion search algorithms from the Mascot house server (version 2.3.01) (Perkins et al., 1999). The following parameters were selected: taxonomy ID 403 (Methylococcaceae), 206350 (Methylophilales), 31993 (Methylocystaceae), 120652 (Methylocella), 184923 (Methylocapsa) 82115 (Rhizobiaceae) and 74201 (Verrucomicrobia) of NCBI nr (National Center for Biotechnology Information, Rockville Pike, Bethesda, MD, USA; state time 6 January 2013) as criteria for taxonomy, tryptic cleavage, maximum two missed cleavage sites. A peptide tolerance threshold of ±10 p.p.m. and an MS/MS tolerance threshold of ±0.8 Da were chosen. Carbamidomethylation at cysteines was given as static and oxidation of methionines as variable modification. Peptides were considered to be identified by Mascot when a false-positive probability <0.05 (probability-based ion scores threshold >40) was achieved. The level of 13C incorporation was calculated as described by Taubert et al. (2011) using the Excel spreadsheet (Taubert et al., 2011).

Results and discussion

The methane flux measurements from the studied floodplain were highly variable, with negative (methane uptake) as well as positive (methane emission) values well within the normal range for this type of ecosystem (Supplementary Figure S2). Methane consumption (initial rates as well as Vmax) displayed a distinct distribution with highest rates in between the low (prolonged wet) and the high (predominantly dry) ends of the flooding gradient (Figures 2a and b). This is the opposite trend compared with the emitted methane flux (Supplementary Figure S2), suggesting that the MOB community is capable of scavenging all methane produced at the intermediate elevation levels.

Figure 2
figure 2

(a) In vitro methane consumption, (b) maximum methane oxidation capacity (Vmax) and (c) pmoA gene copy numbers of type Ia (closed circles), type Ib (open circles) and type II (closed triangles) MOB in the surface 0–2 cm of soil collected at different elevations in a riparian floodplain. Values are seasonal averages of four sampling months plus standard deviation.

Full size image

MOB genera and species, belonging to the lineage of the Proteobacteria (α- and γ- Proteobacteria subdivision), are divided in type Ia, Ib and II subgroups (see also Supplementary Figure S3) based on pmoA phylogeny, which is largely congruent with the phylogeny based on 16S rRNA (Lüke and Frenzel, 2011). qPCR assays targeting these subgroups individually in the 0–2 cm layer of the soil revealed clear differentiated distributions, with type II increasing in number with increasing elevation with respect to the lowest point of the plot (Figure 2c). In contrast, numbers of types Ia and Ib reached highest values in the mid-elevation levels, reflecting the overall pattern of methane consumption (Figure 2). Initial methane consumption activity as well as specific affinity (i.e., specific affinity for methane defined as Vmax/Km) was highest at elevations of 0.1–0.3 m corresponding to moisture content of the soil between 35% and 40% (Supplementary Figure S4), indicating the niche boundaries for highest methane consumption in this soil. qPCR results showed that type II MOB are numerically more abundant than type Ia at all elevations and than type Ib at elevations >0.3 m (Figure 2c). Clone libraries obtained from samples from both ends of the flooding gradient confirmed the dominance of type II (Supplementary Figure S3).The orders of magnitude difference between types II and Ia in the floodplain in this study was in line with parallel studies on the same site (Steenbergh et al., 2010; Wang et al., 2012), confirming the observed relative abundances.

MOB community composition (i.e., Shannon diversity index) as determined by pmoA-based microarray analyses was positively correlated with methane consumption when considering the data obtained from all sampling months (Figure 3a). One has to consider that the Shannon index is applied using microarray probes as ‘species’ analogs. In reality, however, the highest resolving probes still cover multiple species and sequence types. Hence, the Shannon index in this study reflects presence and relative abundance of taxonomical units and cannot be used to compare diversity with other prokaryotic or eukaryotic communities. The observed correlation, however, only occurred in the upper soil layer (0–2 cm) (Figure 3a) and not in the lower layers (7–9 and 27–29 cm) (Figures 3c and e). The short-term in vitro methane oxidation assays are an analog for the potential active community at the time of sampling (Steenbergh et al., 2010). The congruence between abundance of specific MOB subgroups, community composition and activity in the 0–2 cm layer, exclusively, suggests that these subgroups contribute predominantly to methane oxidation in the soil itself.

Figure 3
figure 3

Relationship between in vitro methane consumption and diversity components (Shannon index and evenness) of the methane-oxidizing community in floodplain soil samples collected in September 2006 and May, July and October 2007. Diversity of the MOB community was assessed using a pmoA-based microarray. Data points represent the results from five soil cores taken per month, which were subsectioned in (a and b) 0–2 cm, (c and d) 7–9 cm and (e and f) 27–29 cm. Equations above panels are linear regression results (d.f.=degrees of freedom, F=ANOVA F-value; P=significance level of the regression).

Full size image

The observed correlation between diversity index and methane consumption was determined by the evenness of the MOB community and, hence, changes in relative abundance of the active species in line with change in activity (Figure 3). Richness was not related with methane oxidation activity (linear regression: R2=−0.02, F=0.58, P<0.46). Other activity components were also positively correlated with diversity index. Specific affinity was correlated with Shannon index on DNA basis and Vmax on RNA basis (Supplementary Table S1). Dynamics of signals of individual microarray probes revealed that only type Ia- and Ib-related general and species-specific probes correlated to methane measures of methane-oxidizing activity (i.e., initial consumption, Vmax and specific affinity), whereas type II probes did not (Supplementary Figure S5 and Supplementary Table S1), suggesting that changes in types Ia and Ib to be responsible for the change in evenness of the community. qPCR data also demonstrated strong correlations between abundance of type Ib and both initial methane oxidation and Vmax, whereas numbers of type Ia are correlated with specific affinity (Supplementary Table S1). A principal component biplot of the most relevant components analyzed in this study also clearly demonstrates the exclusive association of type I MOB with all relevant methane consumption parameters, whereas the variation found in type II MOB-related parameters, with Methylocystis species in particular, correlates with the elevation (Supplementary Figure S6). The relationship between mRNA transcript numbers, which can be used as a proxy for enzymatic activity (Freitag et al., 2010), also contributes to the line of correlative evidence for type I MOB as being the active species in this floodplain (Supplementary Figure S7).

What has to be kept in mind with respect to the results described above is the limitations of the pmoA-based methods we have used in this study. The genera containing only soluble methane monooxygenase (i.e., Methylocella and Methyloferula) as well as the MOB belonging to the Verrucomicrobia and the NC10 cluster cannot be detected. More recent described genera (Methylovulum, Methylogae, Methylosoma, Crenothrix and Clonothrix) are not included in the version of the microarray used. The genus Methylocapsa is not detected by the type II qPCR assay but is present on the microarray. Although this genus was detected by micro array probes, the relative abundance did not correlate with activity components but only with elevation, similar to other type II MOB (data not shown). Hence, we cannot exclude that the observed methane consumption is carried out partly by non-detected MOB. However, this does not alter the relationships between abundance, community composition and activity of the genera we could reliably detect.

To gather more direct evidence for the active species and to exclude the possibility that the correlations between in vitro methane oxidation and type Ia and Ib MOB, as observed above, are the result of a negative bias against type II MOB caused by the slurry and shaking procedure of the in vitro assays, we used an SIP approach on intact soil cores. These cores were collected in the field in July 2009 and incubated intact without shaking in an atmosphere spiked with 13C-CH4, which can be traced in taxonomically relevant compounds (e.g., PLFAs and proteins). The intact-core methane consumption pattern revealed higher methane consumption in the lower parts of the flooding gradient, without a distinct optimum (Supplementary Figure S8). This could be related to the long period without flooding preceding the sampling (Supplementary Figure S1). Stable isotope labeling of PLFAs clearly showed that the majority of the label ends up in C16 PLFA, which is known to be mainly associated with type I MOB (Bodelier et al., 2009) (Figure 4). Using multivariate ordination techniques, the labeling pattern, as displayed in Figure 4, can be compared with PLFA profiles of cultivated species (Mohanty et al., 2006; Bodelier et al., 2012), which clearly shows that a combination of both type Ia- and Ib-related MOB is actually consuming the methane in these cores (Figure 5). The labeling of the MOB-specific PLFA (i.e., not present in other microbes; C16:1ω5t and C16:1ω8c) is direct evidence for the methane consumption by representatives of the genera Methylomonas/Methylomicrobium/Methylobacter/Methylosarcina (type Ia) (Bodelier et al., 2009), which were also linked to methane consumption in vitro by microarray analyses (Supplementary Figures S5 and S6 and Supplementary Table S1). The PLFA C16:1ω8c is also present in two type II strains, isolated from acidic habitats (Dedysh et al., 2007); however, significant incorporation by these species would be mirrored by an increase of labeling of the PLFA C18:1ω8c, which they also posses in large amount. In our samples, this did not occur, strengthening the conclusion that type Ia MOB are responsible for incorporation of label in C16:1ω8c.

Figure 4
figure 4

Incorporation of 13C-CH4 using SIP-PLFAs in intact soil cores collected at different elevations within a riparian floodplain. Intact cores were incubated with 13CH4 to label PLFA associated to MOB communities. Each data point is the average of three soil cores per elevation level. Error bars indicate standard deviation. The grey-shaded area indicates PLFA associated to type I MOB, whereas the open box depicts PLFA indicative for type II MOB.

Full size image
Figure 5
figure 5

Multivariate statistical analyses of SIP 13CH4-PLFA profiles of the MOB community (type I only) in intact soil retrieved from a riparian floodplain in July 2009. The inputs of the multidimensional scaling (MDS) analyses were PLFA profiles of MOB cultures (expressed as percentage of total PLFA content) and of SIP profiles of soil samples (expressed as percentage of 13C incorporated in separate PLFAs of the total PLFA 13C uptake). The two-dimensional distances between samples in the MDS graph show the relative similarity between samples. Stress value of the MDS plot was 0.14. For the MDS analyses, the profiles were transformed [log(x+1)] before the Bray–Curtis similarity matrix was established.

Full size image

However, PLFA analysis also gives evidence for methane consumption by Methylocaldum species (type Ib) based on a high label signal in C16:0 and C16:1ω7c relative to the aforementioned PLFAs. These type Ib MOB were detected by cloning, but Methylocaldum-specific microarray probe signals did not correlate with methane consumption. Considering the fact that the used probes cover the clone sequences detected, this suggests the presence of yet uncultivated MOB with a similar PLFA as Methylocaldum species. The non-generic type Ib-related probe (501-286, Supplementary Figure S5 and Supplementary Table S1) that correlates most significantly with methane oxidation activity is in fact a probe targeting a cluster of uncultivated type 1b-related sequences that are most closely related to Methylococcus capsulatus (Supplementary Figure S5).

The type II MOB detected by cloning as well as by microarray were all Methylocystis-related species that contain the MOB-specific PLFA C18:1ω8c as a major component (Bodelier et al., 2009). Label incorporation into this PLFA was insignificant compared with the total incorporation, which also did not allow for comparison of labeling profiles with known cultures as has been performed for type I (Figure 5). However, the displayed incorporation and distribution over the flooding gradient was similar to type II qPCR numbers and microarray probe signals. Type II label incorporation was higher towards higher elevations, whereas incorporation in the specific type I PLFA showed an optimum (Figure 6), similar to the optimum in methane consumption and dynamics of numbers of types Ia and Ib as determined by in vitro assays.

Figure 6
figure 6

13C-CH4 incorporation into specific PLFA for type I and II MOB in intact soil cores, retrieved from a riparian floodplain in July 2009. Bars represent the average of three replicate cores, with error bars showing standard deviation.

Full size image

Similar to the analysis of stable isotope-labeled PLFAs, 13C-labeled proteins can be used to identify the organisms that have assimilated the labeled substrate (Jehmlich et al., 2010). For this study, protein-SIP was used for the first time on MOB communities. The shift in mass spectra resulting from the incorporation of 13C into proteins can be assessed by MS (Supplementary Figure S9) and used to calculate label incorporation by comparison with the spectra of the corresponding unlabeled peptides, which have to be identified in a control sample in parallel.

Protein identification was focused towards members of α-, β- and γ-Proteobacteria and Verrucomicrobia as they include representatives of MOB. The protein hits were mainly affiliated to Methylobacter tundripaludum SV96, Verrucomicrobium spinosum, Methylacidiphilum infernorum V4, Rhizobium spp. and Sinorhizobium spp. (Supplementary Tables S2–S6). Although several genome sequences of MOB (Methylomonas methanica MC09 (Boden et al., 2011), Methylomicrobium album strain BG8, Methylococcus capsulatus strain Bath and Texas (Ward et al., 2004; Kleiveland et al., 2012), Methylocystis strain Rockwell (Stein et al., 2011), Methylocystis strain SC2 (Dam et al., 2012), Methylocystis parvus OBBP (del Cerro et al., 2012), Methylosinus trichosporium OB3b (Stein et al., 2010)) were deposited in the consulted NCBI nr database, almost no other labeled peptide sequences affiliated to MOB were identified besides peptides of Methylobacter tundripaludum SV96. In addition, most of the identified peptide sequences had to be excluded according to the stringent Mascot scoring parameters, which may be related to the soil matrix. Protein extraction from soil is still a great challenge for environmental proteomics. Several protocols are available (Benndorf et al., 2009; Chourey et al., 2010; Keiblinger et al., 2012) and two were also tested here, but none improved the number of identified and validated peptides in a sufficient manner. Nevertheless, the prominent identified protein was PmoB. The preferred detection of PmoB may be caused by the localization in the cytosol compared with PmoA, which is membrane bound and hard to solubilize during protein extraction. In a recent study (Paszczynski et al., 2011), the dominance of peptides belonging to PmoB of Methylosinus trichosporium OB3b was also described by the analyses of groundwater samples from a site contaminated with chlorinated ethenes. Peptides of other subunits were either not identified or identified in minor numbers.

The calculation of stable isotope incorporation in the proteins clearly indicated methane consumption by type Ia-related species (Table 1). Label incorporation was not detected at the highest elevation, whereas at intermediate elevations (0.20 m), proteins (PmoB, guanosine triphosphate-binding protein, methanol dehydrogenase) associated with the type Ia MOB Methylobacter tundripaludum SV96 could be detected (Table 1). This points to highest amounts of labeled proteins in the intermediate samples, corroborating the results found with PLFA and molecular analyses. No 13C incorporation was found in proteins of both type Ib and II MOB and of Verrucomicrobia (Supplementary Tables S2–S6). Either none of the corresponding masses were found in the mass spectra of the control samples, or just the unlabeled (i.e., 12C) masses were detected.

Table 1 RIA of proteins showing 13C incorporation into peptides
Full size table

The evaluation of the mass spectra during the calculation of the relative isotope abundance (RIA) showed two 13C incorporation features, a high incorporation of about 90% 13C and a medium incorporation of about 62% 13C. Both belonged to the same peptide sequence. This occurrence may point to either de novo enzyme synthesis of existing cells, growth of novel cells or differences in enzyme turnover between cells. Another explanation would be that there is still internal production of unlabeled methane within the cores, resulting in methanotrophs consuming unlabeled methane in the inside of the core and labeled methane at the outer parts, resulting in a mix of differently labeled proteins in the final mixed sample used for extraction.

Theoretically, some type Ib and all type II MOB can incorporate carbon from CO2 next to methane (Jahnke et al., 1999; Templeton et al., 2006). Very recently, an elegant study using a stable isotope-labeling approach in cultures of Methylosinus trichosporium OB3b demonstrated that even up to 60% of the biomass carbon can be derived from CO2 (Yang et al., 2013). Basically, this could also have led to the fact that both types are not detected in protein-SIP and to low incorporation in PLFA by type II. When assuming that 60% of the methane is respired to CO2, the ratio of labeled CO2 to unlabeled atmospheric CO2 in our incubations will be 32 (12 600/390 p.p.m.v.), which makes it highly unlikely that the observed results have been influenced by this mechanism.

Very recently, it has been shown that Verrucomicrobial methanotrophs are autotrophs and do not utilize methane but CO2 for generating cell carbon (Khadem et al., 2011; Sharp et al., 2012). Hence, possible activity of these MOB would be missed in our incubations, but as explained above, labeled CO2 is produced during incubation. However, the amount produced in combination with the incubation time may not have been sufficient for effective labeling of Verrucomicrobial proteins. Therefore, we cannot exclude activity of these MOB normally found only in geothermal environments (Sharp et al., 2012). If the presence of Verrucomicrobial MOB proteins detected in our floodplain soil will be confirmed by molecular detection techniques as recently developed (Sharp et al., 2012), then this would be the first evidence of the presence of these methanotrophs in non-extreme habitats. Future work will reveal whether this indeed is the case.

Synthesis

By using a polyphasic approach combining biogeochemical, molecular biological methods and stable isotope labeling, we were able to connect microbial methane oxidation with the active species in the environment. Collectively, the data obtained give support for a linear relationship between biodiversity components and microbial functioning in natural complex soil ecosystems. The investigated natural flooding gradient provided a combination of environmental drivers (e.g., methane, oxygen and nitrogen) leading to proliferation of subgroups of MOB (i.e., type Ia/Ib), resulting in highest methane consumption coinciding with highest community evenness. The congruence between community evenness and function is probably observed in this study because we assessed the species actually responsible for the measured function in a range of conditions allowing them to be active. Thereby, this study highlights three important aspects of BEF research in general and of microbial BEF studies more specifically:

(i) Spatial and temporal environmental heterogeneity giving rise to variability in the number of niches has been neglected so far in BEF studies (Hillebrand and Matthiessen, 2009; Caliman et al., 2010). In comprehensive meta-analyses on BEF studies of the past decades, the incorporation of temporal and spatial scales, representing environmental heterogeneity and the number of niches, in BEF studies is designated as one of the most important issues to address in future studies (Cardinale et al., 2012). Our study incorporated both temporal and spatial scales, which led to a linear relationship between community composition and function, which seems in contrast with the often-saturating species–function curves in artificial community experiments. However, what occurs in our study is that the number of active species, which we are obviously detecting, is increasing with time and environmental variability. The distribution of MOB species and methane consumption in the studied environmental gradient can be represented conceptually in a three-dimensional space (Figure 7), where the relationship between function and biodiversity (i.e., richness, evenness) is determined by the number of niches (defined as the Hutchinson’s ‘ecological niche’; i.e., a multidimensional space where the dimensions are environmental conditions and the resources that define the requirements of an individual or a species to practice ‘its’ way of life) present allowing species to be active and contribute to function. Many BEF studies do not include environmental heterogeneity (i.e., variation in niches), leading to a high number of niche-less (i.e., niche–species mismatch as shown in Figure 7) microbial species that remain inactive and therefore do not contribute to function. This is represented by the solid line in Figure 7, which can be misinterpreted as functional redundancy. The results obtained in our study are more similar to the dashed line as displayed in Figure 7, where the environmental gradient dictates the number of niches available for the large natural diversity of microbial species to emerge from seed banks and become active and contribute to function. Environmental historical contingencies can determine the size of the species pool present in the seed banks and the species emerging. For MOB in rice fields, it has already been shown that similar environmental conditions over thousands of years lead to a rather stable seed bank and to a selection in the species emerging from the seed bank under conditions of methane and oxygen availability (Ho et al., 2011). The rice cultivation over centuries has led to a restricted number of niches and also to a selection of type I MOB emerging from the seed bank. Why type II MOB are abundant in many wetland soils, also in our floodplain, but rarely are detected as dominating active MOB is still a very open question (Ho et al., 2013). Following the logic of Figure 7, no suitable niches are available in our system or alternatively we fail to detect their contribution to methane consumption for methodological reasons.

Figure 7
figure 7

Graphical representation of the three dimensions involved in BEF relationships. The curve of the species–function relationship will depend on the number of niches present, allowing species to be active and contribute to function. Many BEF studies do not include environmental heterogeneity (i.e., variation in niches), leading to a high number of niche-less microbial species that remain inactive and therefore do not contribute to function. This is represented by the solid line, which can be misinterpreted as functional redundancy. The results obtained in our study are more similar to the dashed line; the environmental gradient dictates the number of niches available for the large natural diversity of microbial species to become active and contribute to function.

Full size image

More supporting evidence for the importance of habitat heterogeneity for the BEF has recently been generated in highly controlled stream biofilms, where algal species richness positively correlated to nitrate assimilation under variable flow regimes leading to heterogeneous habitat conditions by which the number of niches could be varied experimentally (Cardinale, 2011). Hence, BEF studies in artificial as well as manipulated natural ecosystems using a fixed number of niches will give an incomplete and even biased picture of BEF relationships in natural ecosystems, especially when results are used for future predictions of ecosystem functioning.

(ii) Functional redundancy in microbial communities as explanation of saturating species–function curves may very well be an invalid concept because the biodiversity axis consists of many species that do not contribute to the measured function, leading to the solid line shown in Figure 7. Next to this, the species that do contribute to function can often not be assigned. Microbial BEF studies use DNA-based molecular diversity assessment techniques bearing no diagnostic means to determine which of the species is contributing to the observed function (Bell et al., 2009). Positive BEF relationships are most often observed with narrow functions (e.g., methane oxidation (Levine et al., 2011), polymer degradation (Wohl et al., 2004; Peter et al., 2011) or pesticide degradation (Monard et al., 2011)), where species are brought together to determine which all contribute to the specific function in the respective experimental setup. Hence, BEF studies in microbial ecosystems where species that are actively contributing to ecosystem functioning cannot be pinpointed will give biased results, often concluding functional redundancy to be occurring which actually is either a lack of niche heterogeneity or cumulative counting of ‘seed bank’ species that are not actively contributing to the measured function.

(iii) Globally highly important processes are regulated by minor components of microbial communities. The active methane-oxidizing species (types Ia and Ib) in the studied wetland constitute a numerically subordinate group within the functional guild of methanotrophs, which is already a minority on itself. When assuming the total amount of bacteria to be approximately 1.109 cells per gram of dry soil (Wessen et al., 2010; Ho et al., 2011; Daniell et al., 2012; Kuramae et al., 2012), which is even the lower limit of the range typically observed, then type Ia and Ib MOB constitute 0.001–0.05% and 0.08–0.5% of the total community, respectively. The individual species carrying out the observed activity that are collectively targeted by the qPCR assays constitute an even lower percentage of the community. Type Ia MOB related to the ones found in our study have been demonstrated to be responsible for methane consumption in many important methane-emitting habitats such as rice paddies (Bodelier et al., 2000), arctic wetlands (Graef et al., 2011), landfills (Chen et al., 2007), lake sediments (Dumont et al., 2011) and floodplains (Bodelier et al., 2012), which has been ascribed to their specific traits enabling them to be very responsive to the periodic availability of methane and other nutrients (Steenbergh et al., 2010; Bodelier et al., 2012). Recent studies assigning species-specific contributions to important biogeochemical cycles using stable isotopes also indicated a disproportionate role of single rare microbial species to globally important processes (Musat et al., 2008; Pester et al., 2010; Peter et al., 2011), strongly suggesting that the traits of the organisms involved will be fundamental to the variability and dynamics in the biogeochemical process catalyzed. Recently, a strong link between diversity of microbes consuming atmospheric methane and soil methane uptake was demonstrated (Levine et al., 2011) for high-affinity methanotrophs in upland soils. In combination with our results on low-affinity MOB in wetland habitats, it can be concluded that microbial diversity and microbial traits are important regulating factors in the global methane budget. The relevance of microbial minorities and their respective niches for globally important ecosystem functions calls for consideration of conservation policy with respect to microbial diversity, especially in light of climate change and other anthropogenic disturbances.