Introduction

Nitrous oxide (N2O) is considered as a major greenhouse gas and is a significant contributor to ozone layer destruction (Zumft and Kroneck, 2006). N2O is mainly produced by denitrification, a microbial respiratory process in which nitrate/nitrite is reduced to gaseous forms (NO, N2O and N2); however, other microbial processes, such as nitrification and dissimilatory nitrate reduction to ammonium (DNRA), can also produce N2O (Conrad, 1996).

Agricultural fields are one of the main sources of N2O emission (Philippot et al., 2007; Minamikawa et al., 2010). In contrast to upland crop fields, little N2O is emitted from rice paddy soils, even though paddy fields are known to have strong denitrification activity (Akiyama et al., 2006). Dissolved N2O is occasionally detected in surface and ground water in rice paddy fields, whereas little or no N2O is emitted to the atmosphere above these fields (Xiong et al., 2006; Minamikawa et al., 2010). This indicates that water-dissolved N2O is possibly reduced by N2O-reducing microorganisms in rice paddy fields.

N2O can serve as an electron acceptor for microbial respiration. The standard reduction potential (E0′ at pH 7) of the reaction (N2O+2H++2e− → N2+H2O) is 1.35 V with ΔG0′ of −339.5 kJ mol−1 (Zumft and Kroneck, 2006). Phylogenetically diverse bacteria and archaea have the ability to reduce N2O. Although the reduction of N2O to N2 gas is part of denitrification, some denitrifiers do not have the ability to reduce N2O (Tiedje, 1994). Both N2O-reducing strains and non-reducing strains may be present within the same species (Sameshima-Saito et al., 2006). In addition, some DNRA bacteria have the ability to reduce N2O (Conrad, 1996). Therefore, it is difficult to use 16S rRNA gene sequences alone to identify N2O reducers. Instead, the gene encoding N2O reductase (nosZ) has been used to detect potential N2O reducers in various environments (Rich et al., 2003). Although nosZ phylogeny is generally in agreement with 16S rRNA gene phylogeny, horizontal gene transfer may have occurred among closely related microorganisms (Dandie et al., 2007; Jones et al., 2008) and we therefore cannot identify N2O reducers on the basis of nosZ sequence information alone.

One approach to link microbial identity to a specific function is stable isotope probing (SIP) of nucleic acids (Radajewski et al., 2000; Gutierrez-Zamora and Manefield, 2010). In the SIP approach, microbes that have incorporated heavy stable isotopes (for example, 13C, 15N, 18O) into their DNA (or RNA) can be identified by analyzing the heavy DNA fractions separated by density gradient ultracentrifugation. Using the SIP approach, we can analyze the 16S rRNA and functional gene diversities of microbial populations involved in specific functions. Previously, 13C-assimilating populations under denitrifying conditions were analyzed by DNA-based SIP analysis (Ginige et al., 2004; Osaka et al., 2006, 2008; Saito et al., 2008). However, microbial populations responsible for N2O reduction have not been examined to date.

Another approach to identifying such populations is to isolate and analyze N2O reducers that are active and dominant in the environment. We previously developed a single-cell isolation technique to obtain actively growing microorganisms from environmental samples and designated it the functional single-cell (FSC) isolation method (Ashida et al., 2010). In this method, individual cells growing in response to certain conditions (for example, denitrification-inducing conditions) are elongated or enlarged, and can be individually captured with a micromanipulator. Single-cell isolation techniques provide an environment without resource competition, thereby allowing microbes, including slow-growing microorganisms, to multiply without interference from fast-growing ones (Ishii et al., 2010a). The FSC isolation method allowed us to obtain denitrifiers that were shown to be active and dominant by culture-independent analyses (Ishii et al., 2011). By analyzing the isolated strains, we were able to directly link the 16S rRNA gene and functional gene phylogenies. In addition, various cell properties, such as denitrification and N2O reduction rates, could also be measured (Tago et al., 2011).

Consequently, the objectives of the current study were (1) to identify 13C-assimilating populations under N2O-reducing conditions by SIP, (2) to isolate N2O-reducing microorganisms from rice paddy soil by using the FSC isolation method, (3) to examine the N2O reduction rates of the isolated strains and (4) to compare the results obtained by SIP and the FSC isolation method.

Materials and methods

Soil microcosm

Soil samples were collected from rice paddy fields at the Institute for Sustainable Agro-Ecosystem Services, The University of Tokyo, Nishitokyo City, Tokyo, Japan (Saito et al., 2008). A soil microcosm setup was established based on the previous reports (Saito et al., 2008; Ishii et al., 2009b), except N2O was used as an electron acceptor instead of nitrate. Succinate was used as an electron donor for N2O reduction in this study. As succinate is a member of the TCA cycle and is considered as a non-fermentable carbon substrate, it can be used by various N2O reducers, but not by fermenting microbes. The optimum concentrations of electron acceptor and donor (N2O and succinate, respectively) were determined by adding several combinations of N2O (0%, 0.5%, 1%, 2%, 5% and 20% in Ar base) and succinate (0 mg, 0.01 mg, 0.025 mg, 0.05 mg, 0.1 mg, 0.25 mg and 0.5 mg C per g soil), and were set at 5% N2O and 0.1 mg succinate C per g soil. For SIP, 13C-labeled succinate (Cambridge Isotope Laboratories, Andover, MA, USA) was used (0.1 mg (=8.3 μmol) of 13C per g soil). For FSC isolation, cell-division inhibitors (20 μg nalidixic acid, 10 μg pipemidic acid, 10 μg piromidic and 10 μg cephalexin) (Joux and Lebaron, 1997) were added to the vial together with N2O and succinate. The vial was then incubated under an Ar:N2O (95:5) atmosphere and static conditions at 30 °C for 24 h.

CO2 and N2O gases in the headspace of the vial were quantified by gas chromatography (GC) as described previously (Saito et al., 2008). When 13C-labeled succinate and 15N-labeled N2O (15N, 99 atom. %; Cambridge Isotope Laboratories) were used, 44CO2, 45CO2, 44N2O, 46N2O, 28N2 and 30N2 were separately quantified by GC–mass spectrometry (GC/MS) using the gas chromatograph/mass spectrometer QP5050 (Shimadzu, Kyoto, Japan) as described elsewhere (Miyahara et al., 2010). Succinate was extracted from soil with 5 ml water and quantified by high-performance liquid chromatography as described previously (Saito et al., 2008). Fe2+ was anaerobically extracted from soil with 1 M ammonium acetate solution (pH 3) and quantified colorimetrically as described previously (Ishii et al., 2009b).

SIP

DNA was extracted from the soil microcosms (n=10) amended with N2O and 13C-succinate (sample 13SN) using ISOIL for bead beating (Nippon Gene, Tokyo, Japan). As controls, DNA was also extracted from the soil microcosms (n=10) amended with N2O and 12C-succinate (sample 12SN), 13C-succinate only (sample 13Su), and 12C-succinate only (sample 12Su). Community structures among the replicate samples were analyzed and compared by PCR–denaturing gradient gel electrophoresis (DGGE) targeting 16S rRNA gene as described previously (Ishii et al., 2010b). After confirming that the DGGE profiles looked similar among the replicate samples, purified DNA from 10 replicate samples was pooled to ensure a sufficient amount of DNA for ultracentrifugation. Cesium chloride density gradient ultracentrifugation was performed as described by Neufeld et al. (2007) at an average of 177 000 × g (53 200 rpm) using a P100VT rotor (Hitachi Koki, Tokyo, Japan). After 40 h centrifugation, gradients of density-resolved DNA were fractionated and purified as described elsewhere (Neufeld et al., 2007). The copy number of the 16S rRNA gene in each fraction was determined by quantitative PCR as described previously (Fierer et al., 2005; Ishii et al., 2009b). Community structures among the DNA fractions were analyzed by the 16S rRNA gene-based PCR–DGGE and principal component analysis as described previously (Ishii et al., 2009a) and clone library analysis as described below.

Single-cell isolation

Metabolically active cells were stained with 5-carboxyfluorescein diacetate acetoxymethyl ester as described previously (Ashida et al., 2010). Fluorescing cells were observed under a fluorescent microscope (Diaphot 300, Nikon, Tokyo, Japan) with 400–1000 × magnification.

Single cells were isolated using a micromanipulator (MTA-31, Daiwa Union, Iida, Nagano, Japan) equipped with a microinjector (UJI-A, Daiwa Union) as described previously (Ashida et al., 2010). After a single cell was captured in the capillary of the micromanipulation system, the tip of the capillary was soaked in 70% ethanol for 30 s to disinfect its outside. The captured cell was then ejected into a test tube containing 100-fold diluted nutrient broth (Hashimoto et al., 2009) supplemented with 4.4 mM succinate (DNB–S medium) and incubated at 30 °C under N2O-reducing conditions for 2 weeks. To obtain purified isolates, the cultures in the DNB–S media were streaked onto DNB–S agar and incubated at 30 °C for 2 weeks.

N2O reduction and denitrification activities of the strains

Each strain was inoculated into 5 ml of DNB–S medium in a 10 ml glass serum vial, and the headspace air was replaced with Ar:15N-labeled N2O (95:5) gas. After incubation at 30 °C for 1 week, the amounts of 15N-labeled N2 and N2O were measured by GC/MS as described above. Denitrification activities of the strains were also measured in duplicate (two vials for each strain) by the acetylene block method (Tiedje, 1994) described previously (Ishii et al., 2011; Tago et al., 2011).

For selected strains, the N2O-reducing rate was also measured. Cells were grown in DNB–S medium in a vial with 5% non-labeled N2O gas in Ar base. After 1-week incubation, cells were harvested and inoculated, in triplicate (three vials for each strain), into fresh, 5 ml of DNB–S medium at 105 cells ml−1. The headspace air was replaced with Ar:15N-labeled N2O (95:5) gas, and the vial was then incubated at 30 °C. Amounts of 15N-labeled N2 and N2O were measured at 3, 6, 9, 12, 18 and 24 h after inoculation by GC/MS as described above.

PCR, cloning and sequencing

For culture-independent clone library analysis of the microbial community in the heavy fractions from 13SN and 13Su samples, the 16S rRNA gene and nosZ were PCR amplified using primers m-27F and m-1492R (Tyson et al., 2004) and nosZ-F-1181 and nosZ-R-1880 (Rich et al., 2003), respectively. PCR was performed using a Veriti 96-well thermal cycler (Applied Biosystems, Foster City, CA, USA) under conditions described elsewhere (Rich et al., 2003; Ishii et al., 2009b). After removing excess primers and dNTP by using a Wizard DNA Cleanup system (Promega, Madison, WI, USA), PCR products were cloned into a pGEM-T Easy vector (Promega) and transformed into Escherichia coli JM109 high-efficiency competent cells (Promega) according to the manufacturer's instructions. DNA inserts from randomly selected clones were amplified by PCR with vector primers T7-1 and SP6, and sequenced as described previously (Saito et al., 2008).

For isolated strains, DNA was extracted from cells as described previously (Ashida et al., 2010). PCR was performed to amplify the 16S rRNA gene and nosZ as described above. In addition, the nitrite reductase gene (nirK or nirS) was amplified using primers F1aCu and R3Cu (Throbäck et al., 2004) or cd3aF and R3cd (Throbäck et al., 2004), respectively, as described previously (Yoshida et al., 2010). PCR products were purified using a Wizard DNA Cleanup system (Promega) and directly sequenced as described previously (Ashida et al., 2010).

DNA fingerprinting analysis

Repetitive element palindromic-PCR DNA fingerprinting was performed using the BOXA1R primer according to the protocol described by Rademaker et al. (2008) to examine the relatedness of the strains (Ishii and Sadowsky, 2009). The amplified DNA fragments were separated by electrophoresis on 1.5% agarose gel at 80 V for 8 h, and the image was visualized under UV light. The image was digitalized and analyzed as described previously (Ishii et al., 2009a). Strains with >80% DNA fingerprint similarity were considered identical.

Phylogenetic analysis

The nucleotide sequences were trimmed and assembled as described previously (Ishii et al., 2009b; Ashida et al., 2010). Taxonomic assignment of the clones or strains was performed based on their 16S rRNA gene sequences by using the Ribosomal Database Project classifier program (Wang et al., 2007) with 80% as the bootstrap cutoff. Operational taxonomic units were determined at 97% nucleotide sequence similarity by using MOTHUR program (Schloss et al., 2009). The nucleotide or deduced amino acid sequences from multiple strains were aligned with reference sequences obtained from the DDBJ/EMBL/GenBank databases. A phylogenetic tree was constructed based on the maximum likelihood method by using MEGA version 5 (Tamura et al., 2007).

Nucleotide sequence accession numbers

The nucleotide sequences of the 16S rRNA gene and nosZ from the isolated strains were deposited in the DDBJ/EMBL/GenBank databases under the accession numbers AB545618–AB545660 and AB545661–AB545698, respectively (Supplementary Table S1). The nucleotide sequences of the 16S rRNA gene and nosZ from the culture-independent analysis were also in the databases under the accession numbers AB608638–AB608703 and AB608704–AB608729, respectively.

Results

Evaluation of the soil microcosm

Based on the preliminary experiments, all of the added N2O disappeared within 24 h of incubation when <2% N2O was added (data not shown). As N2O should always be present to minimize utilization of succinate by metal reducers, the concentration of N2O should be >2%. Based on the Bunsen absorption coefficient and Henry's law, the concentration of the water-dissolved N2O would be 1 mM when 5% N2O was added to a 10 ml vial containing 1 g soil submerged in 1 ml water. This concentration is 10- to 100-fold less than the N level found in the rice paddy field right after the fertilizer application (Saito et al., 2008).

Preliminary experiments also showed that the addition of <0.1 mg succinate C did not significantly enhance N2O reduction (Supplementary Table S2). Addition of >0.1 mg succinate C significantly enhanced N2O reduction (Supplementary Table S2), but 33% and 58% of the added succinate remained unused when 0.25 and 0.5 mg succinate C was added, respectively. In the presence of 5% N2O, all of the added succinate (0.1 mg C) was consumed within 24 h, whereas 32% of the added succinate remained unused in the absence of N2O. Concentrations of Fe2+ in the soil significantly increased (P<0.05) after 24 h anaerobic incubation with 0.1 mg succinate C, but not after anaerobic incubation with 0.1 mg succinate C and 5% N2O nor after anaerobic incubation without succinate addition (Supplementary Table S3). These results suggest that succinate is likely used by N2O reducers when N2O is present, but it can be used by metal reducers when N2O is absent. Based on these results, we considered 0.1 mg succinate C per gram soil to be sufficient and the minimum required for enhancing N2O reduction.

Figure 1 shows time-course changes in N2O, N2 and CO2 in the soil microcosms amended with the optimum concentrations of N2O and succinate (5% and 0.1 mg C, respectively). The quantity of 15N-labeled N2 increased along with the decrease in 15N-labeled N2O in the microcosm, suggesting that N2O was reduced to N2. The N2O decrease was larger than the amount of N2 produced. This unbalanced N2O mass may be attributed to N2 fixation or other N2O metabolism such as N2O oxidation. The amount of 13C-labeled CO2 gradually increased and reached a plateau after 18 h, whereas non-labeled CO2 continued to increase after 24 h. About 10% of the added succinate (0.1 mg C=8.3 μmol) was oxidized to CO2. As all of the added succinate was consumed within 24 h, the remaining ca. 90% of the added succinate was assumed to be used as a C source by actively growing microbes.

Figure 1
figure 1

Time-course changes in 46N2O (), 30N2 (), 45CO2 () and 44CO2 () in soil microcosms amended with 46N2O and 13C-labeled succinate. Mean±s.e. (n=3) is shown. An arrow indicates the time when DNA was extracted (24 h).

Full size image

Identification of N2O reducers by SIP

SIP was performed to study succinate-assimilating populations under N2O-reducing and non-reducing conditions. Figure 2 shows the relative amount of the 16S rRNA gene in DNA fractions separated by CsCl density gradient ultracentrifugation. All four samples had peaks in the light DNA fractions (L fraction) with buoyant densities of 1.70–1.715 g cm−3. An additional peak was also observed in the 13SN sample in the heavy DNA fraction (H fraction) with buoyant densities of 1.73–1.75 g cm−3. A small amount of DNA was also seen in the H fraction from the 13Su sample.

Figure 2
figure 2

Cesium chloride density gradient centrifugation of DNA extracted from soil. Buoyant densities of the light (L), middle (M) and heavy (H) density fractions were 1.70–1.715, 1.715–1.73 and 1.73–1.75 g cm−3, respectively. 13SN sample (), 13Su sample (), 12SN sample () and 12Su sample ().

Full size image

PCR–DGGE analysis showed that the community structure differed between the H and L fractions within a sample (Figure 3). The community structure also differed among the H fractions originating from the 13SN, 13Su and 12SN samples. Bands specific to each fraction were excised and sequenced (Supplementary Table S2). While most bands originated from bacteria belonging to the orders Burkholderiales (class Betaproteobacteria) and Rhodospirillales (class Alphaproteobacteria) in the H fraction of the 13SN sample (Figure 4a), many bands were from bacteria belonging to the order Desulfuromonadales (class Deltaproteobacteria) in the H fraction of the 13Su sample (Figure 4b). Bands appearing in the H fraction of the 12SN sample were similar to that of the 16S rRNA gene sequence of bacteria belonging to the orders Bacillales and Clostridiales (phylum Firmicutes) and the order Rhodospirillales (Supplementary Table S4).

Figure 3
figure 3

Community structure assessed by DGGE analysis. (a) DGGE banding profile from each fraction separated by CsCl density gradient centrifugation. Gel region shown is between 44% and 54% denaturant concentrations, as estimated by the DGGE marker II (Nippon Gene). L, M and H correspond to the light, middle and heavy DNA fractions as shown in Figure 2. Bands specific to the H fractions of each sample (indicated by arrows) were excised, cloned and sequenced (Supplementary Table S2). (b) Principal component analysis plot based on the DGGE profile. The normalized location and intensity of each DGGE band were used (Ishii et al., 2009a). The numbers in the plot correspond to the lanes in panel a. The percentages in parentheses are the percentages of variation explained by the components.

Full size image
Figure 4
figure 4

Taxonomic classification of the (a) DGGE band excised from the H fraction of the 13SN sample, (b) DGGE band excised from the H fraction of the 13Su sample, (c) clones obtained from the H fraction of the 13SN sample, (d) clones obtained from the H fraction of the 13Su sample and (e) strains obtained by the FSC isolation method. Taxonomic assignment was performed using the Ribosomal Database Project classifier program (Wang et al., 2007) at the order and genus level for the DGGE results (ca. 180 bp) and clone library results (ca. 1450 bp), respectively. Relative intensities of the DGGE bands (see Supplementary Table S2) correspond to the fraction of the assigned taxon.

Full size image

In order to examine the community structure in the H fractions of the 13SN and 13Su samples in detail, we performed clone library analysis based on the near-full length 16S rRNA gene. Similar to the PCR–DGGE results, most clones were related to the orders Burkholderiales and Rhodospirillales in the H fraction of the 13SN sample (Figure 4c). Among these, clones closely related to the genus Herbaspirillum (order Burkholderiales) were most frequently obtained. In contrast, clones related to the genus Geobacter (order Desulfuromonadales) dominated the H fraction of the 13Su sample (Figure 4d).

Isolation of N2O reducers

In addition to the culture-independent analyses, culture-based analysis was also performed in this study. During FSC isolation, 61 elongated single cells were captured from the soil microcosm incubated under N2O-reducing conditions. No elongated cells were observed in the sample without cell-division inhibitors. After single-colony isolation and GC/MS analysis, 33 N2O-reducing strains were obtained.

Similar to the results obtained by clone library analysis, strains closely related to the genus Herbaspirillum were most frequently obtained (20 strains; Figure 4e). 16S rRNA gene sequences of the isolated Herbaspirillum strains were >98% similar to the SIP clones obtained in this study (Figure 5a). Strains related to the genera Azospirillum (seven strains) and Burkholderia (three strains) were the second and third most frequently obtained, respectively.

Figure 5
figure 5

Phylogenetic relationships between SIP clones and FSC isolates. The phylogenetic trees were constructed based on (a) the 16S rRNA gene sequences and (b) deduced nosZ amino acid sequences, by using the maximum likelihood method. Clones obtained from the H fractions of the 13SN and 13Su samples are shown in green closed circle and blue open circle, respectively; strains obtained by the FSC isolation method are shown in red square. Taxonomic assignment of the strains obtained by the FSC isolation method was performed using the Ribosomal Database Project classifier program (Wang et al., 2007). The numbers in parentheses are the numbers of clones in the operational taxonomic units (for SIP) or the number of strains that have the identical DNA fingerprinting patterns as the representatives (for FSC isolation). The accession numbers of the reference strains in the DDBJ/EMBL/GenBank databases are indicated in brackets. The bootstrap values (>70%) from 500 replicates are indicated next to the branches.

Full size image

N2O reductase gene

nosZ was detected in all N2O-reducing strains. Diverse nosZ sequences were also obtained from the clone library constructed based on the H fraction of the 13SN sample. Figure 5b shows the phylogenetic tree constructed based on the nosZ sequences obtained in this study. With some exceptions, similar nosZ sequences were obtained from phylogenetically closely related strains. For example, nosZ sequences of most Burkholderiales bacteria (Burkholderia spp., Herbaspirillum spp. and Massilia spp.) were clustered together (cluster I). The nosZ sequences of some Herbaspirillum strains were distantly related to these sequences and were more closely related to the nosZ of Azospirillum spp. (cluster II).

Figure 5b also shows the relatedness between nosZ sequences obtained from SIP and FSC analyses. From the H fraction of the 13SN sample, nosZ sequences in cluster I were most frequently obtained (78%), and these sequences were >76% similar to those of Burkholderiales. We did not find 100% match in the nosZ sequences between isolated strains and SIP clones. This may be due, in part, to the formation of chimeric sequences in SIP analysis.

N2O reduction and denitrification activities

Based on the nosZ sequence information and repetitive element palindromic-PCR DNA fingerprinting, three Herbaspirillum strains (TSO23-1, TSO35-1 and TSO37-1), three Azospirillum strains (TSO5, TSO22-1 and TSO41-3) and two Burkholderia strains (TSO10-2 and TSO47-3) were selected for measurement of N2O reduction activity. As both an electron acceptor and an electron donor were abundantly present under the experimental conditions, the reaction (N2O+2H++2e− → N2+H2O) followed zero-order kinetics. The N2O reduction rates of Herbaspirillum spp., Burkholderia spp. and Azospirillum spp. were 1.72±0.13, 1.39±0.24 and 0.65±0.06 pmol h−1 cell−1, respectively; these differed by genus (P<0.05) but not by nosZ cluster (cluster I vs II+III).

The majority (76%) of the N2O-reducing strains carried nirS, which encodes cytochrome cd1 nitrite reductase, and were able to perform denitrification (Supplementary Table S1). No strains were detected with nirK, which encodes copper-containing nitrite reductase. We could not amplify the nitrite reductase gene from the denitrifying Azospirillum strains TSO5, TSO7, TSO9 and TSO35-2. Azospirillum sp. strain TSO41-3, Burkholderia sp. strain TSO11-3, and Massilia and Bacillus strains did not show denitrification activity and nitrite reductase genes were not detected.

Discussion

Although denitrifying and nitrate-reducing communities have been studied in various environments including rice paddy soils (Philippot et al., 2007; Ishii et al., 2009b), microbial communities responsible for N2O reduction have not been well characterized. In the present study, we employed both culture-independent (SIP) and culture-dependent (FSC isolation) techniques to analyze N2O reducers in rice paddy soil. Populations that assimilated succinate under N2O-reducing conditions were examined by SIP analysis. The FSC isolation method was used to isolate microbes that were ready to grow under the same N2O-reducing conditions used for the SIP analysis. Combined analysis of the results obtained by SIP and FSC isolation allowed us to assess the phylogeny, function and physiology of the microbes responsible for N2O reduction.

In the present study, succinate was used as an electron donor by the N2O reducers. Previous studies have shown that anaerobic incubation of soil with nitrate and succinate greatly enhances denitrification activity (Saito et al., 2008; Ishii et al., 2009b). Under such conditions, succinate could be used by various denitrifiers, whereas there would be little utilization of succinate for other functions, such as fermentation, DNRA, and metal and sulfate reduction (Saito et al., 2008; Ishii et al., 2009b). Oxidation of succinate (E0′=+33 mV) can also be coupled with reduction of N2O (E0′=+1355 mV). According to the thermodynamic theory (Thauer et al., 1977), N2O is the preferred electron acceptor to Mn4+, Fe3+ and sulfate. Our results support this notion as the production of Fe2+ was suppressed by the addition of N2O.

Succinate-assimilating populations under N2O-reducing and non-reducing conditions were examined by SIP. Under non-N2O-reducing conditions (sample 13Su), clones related to the genus Geobacter (order Desulfuromonadales) were most frequently (42%) obtained (Figure 4d). As production of Fe2+ was observed in the 13Su sample, bacteria identified in the 13Su clone library may be involved in metal reduction with succinate as an electron donor. The Geobacter species is well known for its capacity to reduce metals (Lovley et al., 2004). Similar to our study, RNA-based SIP analysis has revealed that Geobacter, Anaeromyxobacter, and a novel Betaproteobacteria closely related to the order Rhodocyclales were acetate-assimilating iron reducers in Italian rice paddy soil (Hori et al., 2009). In the present study, Herbaspirillum spp. and other Burkholderiales bacteria were also detected in the H fraction of the 13Su sample. As these bacteria were not detected in the control sample (H fraction of the 12SN sample), they are most likely enriched under succinate-assimilating and metal-reducing conditions. Similar to our study, clones related to Herbaspirillum have been frequently obtained in a sediment sample (collected at Oak Ridge, TN, USA) incubated without nitrate (Li and Krumholz, 2008) and in a sediment sample (collected at Hanford, WA, USA) incubated with organic acids (Lee et al., 2010).

In contrast to the results obtained from the 13Su sample, clones related to Herbaspirillum spp. and other Burkholderiales bacteria dominated the population in the clone library constructed from the H fraction of the 13SN sample (N2O-reducing conditions) (Figure 4c). Involvement of these bacteria in N2O reduction was also supported by culture-dependent FSC isolation (Figure 4e). Herbaspirillum strains obtained by FSC isolation carried nosZ and reduced exogenous N2O to N2. The majority of the SIP nosZ clones were similar to nosZ of Herbaspirillum and other Burkholderiales N2O reducers (Figure 5). Similar nosZ clones have also been obtained from other paddy fields (for example, GenBank Accession No. ACI48848) and maize rhizospheric soils (Mounier et al., 2004; Dambreville et al., 2006; Henry et al., 2008). Considering the general agreement between the 16S rRNA gene and nosZ phylogenies (Jones et al., 2008; Palmer et al., 2009), these results suggested that Herbaspirillum and other Burkholderiales bacteria may be important players in N2O reduction, not only in rice paddy soils but also in other environments. Herbaspirillum strains were previously shown to be involved in nitrate reduction of rice paddy soil (Ishii et al., 2009b, 2011), but the present study showed that they are also important players in N2O reduction. Although some Herbaspirillum species (for example, Herbaspirillum seropedica) can colonize rice roots and stems and fix atmospheric N2 (Baldani et al., 1986; Elbertagy et al., 2001), almost all strains obtained in this study did not show N2-fixing ability (S. Ishii, unpublished data). In addition, 16S rRNA gene similarities between the Herbaspiriilum strains obtained in this study and other Herbaspirillum species were <97%. These results suggest that the N2O-reducing Herbaspirillum strains obtained in this study may constitute a new species.

The SIP and FSC isolation results also suggested that Azospirillum spp. and other Rhodospirillales bacteria were responsible for N2O reduction. The N2O reduction rates suggested that Azispirillum spp. reduced N2O more slowly than Herbaspirillum spp. Although their in situ N2O reduction rates are not known, these results indicated that the relative contribution of Azospirillum strains to N2O reduction might be smaller than that of Herbaspirillum strains. Similar to other Azospirillum strains (for example, Azospirillum brasilense and Azospirillum sp. B510; Isawa et al., 2010), our Azospirillum strains also showed N2-fixing ability (S. Ishii, unpublished data). Relatively close phylogenetic relationship between the Azospirillum strains obtained in this study and other Azospirillum strains (Figure 5a) also suggested that they may be able to colonize plant roots and fix N2.

Some Azospirillum, Burkholderia, Massilia and Bacillus strains did not have a detectable nitrite reductase gene and did not show denitrification ability. The lack of detection of a nitrite reductase gene may be attributed to the primers used in this study, as there are no annealing sites for the currently available PCR primers on the nirK sequence of Azospirillum sp. B510 (Ishii et al., 2011). However, it is also possible that these strains lack a nitrite reduction pathway as nirK of several Azospirillum strains is located on plasmids (Pothier et al., 2008; Kaneko et al., 2010).

In conclusion, our results suggest that most N2O reducers are denitrifiers under the present study conditions, although some DNRA bacteria are known to reduce N2O (Conrad, 1996). Among the N2O reducers, Burkholderiales bacteria, especially those belonging to the genus Herbaspirillum, may have an important role in N2O reduction in rice paddy soil. As Herbaspirillum bacteria are potential key players in nitrate reduction (Ishii et al., 2009b), these bacteria can be used for the removal of contaminated nitrate from environments (for example, groundwater) while minimizing the emission of N2O. Our study also identified several N2O reducers lacking denitrification activity. These bacteria could be used to mitigate N2O emission from agricultural fields while minimizing the loss of fertilizer N.