The diversity of the N2O reducers matters for the N2O:N2 denitrification end-product ratio across an annual and a perennial cropping system

Agriculture is the main source of terrestrial emissions of N2O, a potent greenhouse gas and the main cause of ozone layer depletion. The reduction of N2O into N2 by microorganisms carrying the nitrous oxide reductase gene (nosZ) is the only biological process known to eliminate this greenhouse gas. Recent studies showed that a previously unknown clade of N2O-reducers was related to the capacity of the soil to act as an N2O sink, opening the way for new strategies to mitigate emissions. Here, we investigated whether the agricultural practices could differently influence the two N2O reducer clades with consequences for denitrification end-products. The abundance of N2O-reducers and producers was quantified by real-time PCR, and the diversity of both nosZ clades was determined by 454 pyrosequencing. Potential N2O production and potential denitrification activity were used to calculate the denitrification gaseous end-product ratio. Overall, the results showed limited differences between management practices but there were significant differences between cropping systems in both the abundance and structure of the nosZII community, as well as in the [rN2O/r(N2O+N2)] ratio. More limited differences were observed in the nosZI community, suggesting that the newly identified nosZII clade is more sensitive than nosZI to environmental changes. Potential denitrification activity and potential N2O production were explained mainly by the soil properties while the diversity of the nosZII clade on its own explained 26% of the denitrification end-product ratio, which highlights the importance of understanding the ecology of this newly identified clade of N2O reducers for mitigation strategies.


Introduction
Nitrous oxide (N 2 O) is one of the six gases subject to restriction by the Kyoto Protocol, which aims at reducing anthropogenic greenhouse gas (GHG) emissions. N 2 O is both directly and indirectly of importance for the Earth's climate. Firstly, it is a potent greenhouse gas with a long life time of 110 years and a global warming potential 298 times that of carbon dioxide on a 100-year time scale and per unit of weight. Thus, N 2 O is the third most important GHG contributing to about 10% of annual global warming (Bates et al., 2008;Thomson et al., 2012). Secondly, after the success of the Montreal Protocol for phasing out chlorofluorocarbons (CFCs), N 2 O is today the dominant ozone-depleting substance (Ravishankara et al., 2009). The atmospheric concentration of N 2 O has been rising over the past 100 years resulting in a concentration 19% higher than preindustrial levels (Montzka et al., 2011) with an estimated increase of N 2 O emissions of up to 60% by 2050 (relative to 1900 values) (Bouwman et al., 2013).
N 2 O emissions are, to a great extent, the result of microbial processes such as nitrification and denitrification. However, denitrification is also the only known sink for N 2 O. Denitrification is a microbial respiratory pathway through which soluble forms of nitrogen, i.e., nitrate (NO − 3 ) and nitrite (NO − 2 ), are sequentially transformed into NO, N 2 O and N 2 gases via four enzymatic systems (Tiedje et al., 1982;Zumft, 1997). The reduction of soluble NO − 2 into NO and N 2 O is catalyzed by copper-or cd 1 -nitrite reductases and nitric oxide reductases, respectively (Zumft, 2005). Nitrous oxide reductase, whose catalytic subunit is encoded by the nosZ gene, is the last enzyme of the pathway. It converts the GHG N 2 O into inert N 2 , which accounts for 78% of the atmospheric gases, and is, therefore, the key enzyme involved in the N 2 O sink. It is now recognized that denitrification is a modular process (Graf et al., 2014). Thus, while some microorganisms harbor all denitrification enzymes and can potentially perform the complete pathway, others either lack the nitrous oxide reductase gene and produce only N 2 O as the denitrification end product , or are only able to reduce N 2 O (Sanford et al., 2002).
Recent studies have identified a previously undescribed nosZ clade, herein after named nosZII, which is diverse and widespread in the environment (Sanford et al., 2012;Jones et al., 2013;Orellana et al., 2014). Genome analyses showed that an important fraction of the microorganisms that possess this nosZII gene also harbor a highly truncated version of the denitrification pathway without any nitrite reductase or N 2 O-producing nitric oxide reductase, and, therefore, can only consume N 2 O (Graf et al., 2014). The abundance and phylogenetic diversity of nosZII microorganisms was found to mediate the soil N 2 O sink capacity in European soils , showing the importance of understanding the ecology of this microbial guild for mitigating N 2 O emissions.
Agriculture accounts for about 60% of all N 2 O emissions from global anthropogenic sources (Syakila and Kroeze, 2011). A compilation of more than 1215 measurements of N 2 O emissions from agricultural and natural soils showed that agricultural practices, such as the N application rate, type of crop and type of fertilizer, affected the emissions (Stehfest and Bouwman, 2006). More recently, Shcherbak et al. (2014) reported that the response of soil N 2 O emissions to nitrogen fertilizer was nonlinear for synthetic fertilizers and most crop types. The effect of agricultural practices has often been described in terms of changes in soil substrates or environmental conditions, which can also affect soil microbial communities in various ways. For example, soil amendment with peat stimulated the relative abundance of the Alphaproteobacteria, but reduced the relative abundance of Firmicutes (Wessen et al., 2010). Organic farming increased richness, decreased evenness and shifted the structure of the soil microbial community when compared with conventionally managed soils amended with mineral fertilizers (Hartmann et al., 2014) Shifts in abundance and structure of the denitrifier community have also been reported in response to the fertilization regime (Hallin et al., 2009;Clark et al., 2012;Tatti et al., 2014) or in response to land use intensity (Meyer et al., 2013). However, very little is known about how soil management could affect microorganisms belonging to the newly described nosZII clade while agricultural practices that foster these microorganisms are of interest for mitigating N 2 O emissions.
This study was therefore set out to determine how the nosZI and nosZII N 2 O-reducing communities responded to various agricultural practices in two different arable farming systems. It also assessed the relationships between diversity, composition, and abundance of the N 2 O-reducing microbial communities and N-gas production (N 2 O and N 2 ) by denitrification. The study was based on two randomized block experiments localized at the same site, one with an annual rotation with 5 different management practices (ORE) and one with a perennial crop system with 4 different management practices (BE).

Abundance of Total Bacteria, N 2 O-producers and N 2 O-reducers
The genes encoding catalytic enzymes involved in N 2 O production (nirK and nirS) and N 2 O reduction (nosZI and FIGURE 1 | Potential N 2 O emissions (A), potential denitrification activity (B) and denitrification gaseous end-product ratio [rN 2 O/r(N 2 O+N 2 )] (C). Means ± sem per treatments within each experimental block are given. BE and ORE represent the respective cropping systems. Significant differences between treatments are indicated with different letters (anova followed by Tukey HSD test, P < 0.05). nosZII) were quantified by Real-Time quantitative PCR (qPCR) and used as proxies for the abundances of the corresponding functional communities. The 16S rRNA gene copy number, which was used to estimate the abundance of the total bacteria community, varied from 1.3 × 10 9 to 5.1 × 10 9 copy numbers g-1 oil DW without any significant difference between treatments (data not shown). The relative abundances of nirS and nirK communities were similar, ranging from 2.5 to 7.6% of the total bacterial community ( Figure S1). The nosZI community was significantly (P < 0.05) more abundant than the nosZII community, ranging from 4.9 to 8.5% and 0.28 to 2.9% of the total bacteria, respectively ( Figure S1). No significant differences in N 2 O-producers and N 2 O-reducers abundances were found between treatments ( Figure S1). However, nosZII abundance was higher in BE than in ORE (P < 0.01).

N 2 O-reducer Diversity
To assess whether the agricultural practices could drive the composition and structure of the N 2 O-reducer community, the diversity of both nosZI and nosZII clades was characterized by 454 pyrosequencing. 123,130 nosZI and 121,500 nosZII sequences were obtained from samples after quality filtering. Similarity-based clustering of sequences gave an average of 162 (CI 95% = 133-191) and 312 (CI 95% = 279-345) OTUs for nosZI and nosZII respectively in BE, and 158 (CI 95% = 136-180) and 355 (CI 95% = 330-380) OTUs for nosZI and nosZII respectively in ORE. In both BE and ORE, the nosZII community had a significantly higher richness than nosZI (P < 0.0001). Within BE and ORE, the agricultural practices had no significant effect on the α-diversity of the N 2 Oreducing communities (Table S1). An analysis of nosZI and nosZII phylogenies showed that the most abundant sequences in ORE and BE were similar ( Figures S2, S3), with nosZII sequences affiliated mainly to nosZ from Bacteroidetes while nosZI sequences were affiliated to nosZ from Alphaproteobacteria and Betaproteobacteria. Further examination of the β-diversity by non-metric multidimensional scaling (NMDS) and analysis of similarity (ANOSIM) showed (Figures 2A,B) no clustering of samples according to the agricultural practices. However, differences in both nosZ communities between annual rotation and perennial cropping systems were significant, but stronger for nosZII communities, (R = 0.43, P < 0.0001 and R = 0.77, P < 0.0001 for nosZI and nosZII respectively). Fitting the environmental variables onto the ordination plot showed that pH and calcium were significant explanatory variables (P < 0.05) for the community structure of both guilds. The nosZI community structure was also related to the water content while the abundance of nitrite reductase genes (nirK) and sand content were related to the nosZII community structure ( Figure 2B). The [rN 2 O/r(N 2 O+N 2 )] surface was fitted onto the ordination plot and showed a strong relationship between the nosZII community structure, its diversity and the denitrification end products with a lower proportion of N 2 O produced when the nosZII diversity increased ( Figure 2B).

Denitrification Activity and End Product Ratio as a Function of Soil Properties, Abundance and Diversity of the N 2 O-reducing Community
We used variance partitioning technique to quantify the relative contribution of the different groups of variables to the variation in N 2 production by denitrification across samples ( Figures 3A-D). The physical and chemical characteristics of the soil, the abundance of N 2 O producers and reducers, and the diversity of N 2 O reducers were used as explanatory variables. After model selection using multiple linear regressions (Table S2), the physical and chemical properties of the soil were found to be the variables that contributed most to the potential N 2 O production and PDA, explaining up to 29 and 45% of the variance, respectively (Figures 3B,C). In contrast, the [rN 2 O/r(N 2 O+N 2 )] was mostly explained by the diversity of the N 2 O-reducers (26%). Interactions between physical and chemical properties of the soil and the diversity of the N 2 O-reducing communities accounted for 26 and 17% of potential N 2 O production and [rN 2 O/r(N 2 O+N 2 )], respectively ( Figures 3B,D). The importance of nosZII diversity for the end product ratio of denitrification was also suggested by the strong negative correlation between the (rN 2 O/r(N 2 O+N 2 )) and the nosZII diversity (r = -0.70, P < 0.0001) ( Figure  S4). The abundance of the communities studied made only a marginal contribution, explaining 2% of the variance in [rN 2 O/r(N 2 O+N 2 )]. ]. Significant explanatory variables are represented as blue vectors (P < 0.05), Ca (calcium g kg −1 dw soil), q.nosZII and q.nirK correspond to the quantification by qPCR of nitrous oxide reductase community (clade II) and copper nitrite reductase community, respectively (copy number g −1 dw soil); sand and water content are expressed in percentage. The lengths of the arrows are proportional to the strength of the correlation. Stress values are indicated at the bottom right of each panel.

Discussion
The recent identification of a previously unknown clade of N 2 O reducers (Sanford et al., 2012;Jones et al., 2013), whose abundance and phylogenetic diversity are critical for the soil N 2 O sink capacity , raised the question of whether it would be possible to mitigate N 2 O GHG emissions by selecting agricultural practices that favor these microorganisms.
A large body of literature shows that land management can drive microbial communities (Clark et al., 2012;Lauber et al., 2013;Hartmann et al., 2014). Here, our intent was not to compare the impact of individual agricultural practices between the two cropping systems but rather to assess how practices for a particular system, such as tillage, residue management, quantity and nature of N inputs in ORE and biomass crop species and harvest date in BE, affect N 2 O-reducing communities and N 2 production. We did not observe any significant effect of the practices for either BE or ORE on either the diversity or the abundance of the N 2 O-reducing communities (Table S1 and Figure S1). In contrast, previous results showed differences in the denitrifier community under different fertilization regimes (Hallin et al., 2009) or tillage systems (Melero et al., 2011). Significant effects of the cropping systems on the total bacterial communities were also reported (Hartmann et al., 2014). One common feature of these studies is that they were based on long term experiments that had been running for up to 50 years. This discrepancy with our results could be due to the fact that the ORE and BE systems were established only in 2010 and 2006, respectively, which has not allowed a strong differentiation Frontiers in Microbiology | www.frontiersin.org in soil properties between practices within a cropping system ( Table 1). On the other hand, the comparison of ORE and BE did show differences in the N 2 O reducing communities with a higher nosZII abundance for ORE than for BE (P < 0.01). NosZII community richness estimated as OTU numbers was also significantly higher in ORE than in BE (Table S1). Furthermore, NMDS also showed significant clustering according to perennial and annual rotation cropping systems for nosZII and to a lesser extent for nosZI (Figures 2A,B). Since it is not possible to distinguish the effect of practices from soil legacy or age effects between ORE and BE, it cannot be concluded that the observed changes in the nosZII community were due to differences in agricultural practices. However, the stronger response of the nosZII community structure and abundance indicates that this recently identified clade is more sensitive to environmental changes than the nosZI clade and is therefore more likely to be driven by the land use type.
The agricultural practices studied also had little impact on PDA and N 2 O production. However, larger differences were observed when calculating the denitrification gaseous endproduct ratio (rN 2 O/r(N 2 O+N 2 )) with, in particular, early harvest of M. giganteus giving a significantly higher proportion of N 2 O than late harvest. Previous studies at BE showed that late harvest of M. giganteus gave a significant net input and accumulation of organic matter due to senescence and leaf fall between early and late harvest (Amougou et al., 2011(Amougou et al., , 2012Cadoux et al., 2014). This type of mulch is known to improve soil moisture by reducing soil surface evaporation, resulting in a lower partial pressure of oxygen (pO 2 ) . This is consistent with our results since previous studies have shown that production of N 2 O relative to N 2 during denitrification in soils is strongly influenced by carbon availability and pO 2 (Firestone et al., 1980;Murray and Knowles, 2004;Walker et al., 2008;Giles et al., 2012;Saggar et al., 2013) with, for example decreasing proportion of N 2 O produced during denitrification with decreasing O 2 concentrations (Firestone et al., 1980). However, the harvesting date had no significant effect on the denitrification gaseous endproduct ratio for switchgrass. This might be due to the fact there is no leaf deposition between late and early harvest for switchgrass. Similarly, to the N 2 O reducer community structure, significant differences in the N 2 O ratio were observed between the BE and ORE cropping systems with a lower end-product ratio in ORE (P < 0.001). In agreement with previous results , a strong negative correlation was found between the (rN 2 O/r(N 2 O+N 2 )) ratio and nosZII diversity (r = -0.70, P < 0.0001) ( Figure S4), highlighting the importance of nosZII N 2 O reducers in the N 2 O-reducing capacity of a soil.
We found that the abundance of nosZII N 2 O reducers was significantly correlated with several soil properties such as pH, Ca concentration, soil moisture and total N. This was not the case for nosZI which was correlated solely with the C:N ratio, confirming that the nosZII clade is more sensitive to changes in environmental conditions than the nosZI clade. The analysis of the community structure showed that two N 2 O-reducer clades shared a certain number of explanatory variables but there were also distinct explanatory variables for each clade. The structure of both communities was driven by pH and calcium, while water content was related to clade I and sand to clade II (Figures 3A,B). To our knowledge, the difference in the response of the two N 2 O reducer clades to soil properties has only been assessed in one previous study . Using structural equation modeling, Jones et al. (2014) showed that soil texture was a more important driver of the abundance of the nosZI community whereas soil pH affected the abundance of the nosZII community only. Overall, our results, which showed that different factors influenced the nosZI and nosZII clades, confirm niche partitioning between the two N 2 O reducing communities. They also indicate that agricultural practices could affect nosZI and nosZII communities in different ways with consequences for N 2 O reduction.
Variation partitioning analysis was applied to disentangle the contribution of soil properties, the abundance of N 2 O producers and reducers, and the diversity of N 2 O reducing microbial communities to the production of N 2 O and N 2 as the denitrification end products. Potential N 2 O production (29%) and PDA (45%) were mostly explained by soil properties. Water content, C/N ratio, Cation Exchange Capacity (CEC) and pH were the major soil properties explaining N 2 O production, while sand content, total nitrogen and the C/N ratio explained PDA. Accordingly, nitrate and carbon availability, pO 2 and related variables such as water content and soil texture are well known proximal factors regulating N 2 O and PDA rates (Baggs, 2011;Giles et al., 2012). pH is also known to be a master regulator of biological processes in soils (Enwall et al., 2005;Giles et al., 2012;Petersen et al., 2012;Saggar et al., 2013;Liu et al., 2014) including the reduction of N 2 O into N 2 by nitrous oxide reductase (Firestone et al., 1980). Unlike other studies (Hallin et al., 2009;Petersen et al., 2012;Jones et al., 2014), the abundance of the N 2 O producing and reducing communities did not explain variations in PDA nor in potential N 2 O production (Figures 3B,C). According to Petersen et al. (2012) the abundance of denitrifiers was correlated with PDA only when large variations in fluxes were observed, which was not the case in this study. In contrast to potential N 2 O production and PDA, the variance in the proportion of N 2 O emitted by denitrification [rN 2 O/r(N 2 O+N 2 )] was mostly explained by the diversity of the nosZII clade (26%), and the interaction between this diversity and soil properties (17%) (Figure 3D). This is in agreement with previous research showing a positive correlation between the ability of the soil to consume N 2 O and the phylogenetic diversity of the nosZII clade . It also illustrates the importance of the nosZII community in determining the nature of the denitrification gaseous endproducts through their capacity to act as a N 2 O sink. Overall, our results showed that the newly described nosZII clade is the strongest predictor of the [rN 2 O/r(N 2 O+N 2 )] ratio while soil properties are the main drivers of potential denitrification and N 2 O production. They also showed that the two clades of N 2 O reducers were not affected by the same soil properties, suggesting niche partitioning. The nosZII clade was more sensitive to environmental changes than the nosZI clade, which may make it easier to foster this group using agricultural practices as a new strategy for mitigating N 2 O emissions. Further studies are required to determine the effect of different agricultural practices on the abundance and diversity of the nosZII clade, in sites with different pedoclimatic conditions to provide more information on the ecology of this recently described functional guild. Moreover, due to the increasing evidence that fungi can produce N 2 O and N 2 by denitrification and co-denitrification (Laughlin and Stevens, 2002;Wei et al., 2014;Maeda et al., 2015), respectively, the response of these microorganisms to agricultural practices should also be considered to circumvent any tradeoff.

Experimental Design and Sampling
Soil samples were collected in October 2013 from two randomized field experiments ORE (49 • 52 25.615 N, 3 • 1 53.914 E) and BE (49 • 52 19.29 N, 3 • 0 47.267 E) located at the same site near Estrées-Mons, France, which has both annual rotation and perennial crop systems, as well as various agricultural practices. A description of the practices can be found in Table 1. Replicated plots with the same practices were randomly distributed within each block experiment. Briefly, the ORE experiment, which consisted in 5 treatments (T1-T5), was set up in 2010 to study the effect of soil tillage, crop residue management, fertilization rate and substitution of mineral N input by fixation by legumes on biogeochemical cycles and soil biodiversity (www.soere-acbb.com). The BE experiment was set up in 2006 to compare the productivity and environmental impacts of various energy crop systems including perennial crops such as Miscanthus giganteus and Panicum virgatum and differences in management practices such as early or late harvest (ML, ME, SL, and SE). Three replicate samples were collected for each combination of cropping system and management practices, each being a composite sample of five subsamples (soil cores of 2.5 cm by 20 cm) from each plot. Samples were frozen (−20 • C) until further analysis. The physical and chemical soil characteristics were measured for all samples (INRA Laboratory of Soil Analysis, Arras, France) ( Table 1).

Potential Denitrification Activity (PDA) and Potential N 2 O production
Potential denitrification activity (N 2 O + N 2 ) and potential nitrous oxide production (N 2 O) were measured using the acetylene inhibition technique (Yoshinari et al., 1977). For each sample 10 g of fresh weight soil was wetted with 20 ml of distilled water and was amended with a final concentration of 3 mM KNO 3 , 1.5 mM succinate, 1 mM glucose, and 3 mM acetate. To determine the potential denitrification activity, acetylene was added to reach 0.1 atm partial pressure followed by 30 min incubation at 25 • C and agitation (175 rpm). Gas samples were taken every 30 min for 150 min (Pell et al., 1996). The N 2 O concentrations were determined using a gas chromatograph (Trace GC Ultra, Thermo Scientific) equipped with an EC-detector.

Nucleic Acid Extraction and Abundance of Bacterial Communities
DNA extraction for all samples was performed in accordance with ISO 11063 (Petric et al., 2011). 0.25 g of soil was homogenized with a 1 ml homogenization buffer for 30 s at 1600 rpm in a mini-bead beater cell disruptor (Mikro-Dismembrator S; B Braun Biotech International), followed by centrifugation at 14000 × g for 1 min to eliminate soil and cell debris. For protein precipitation, supernatant was incubated on ice for 10 min with 1/10 volume of 3M sodium acetate and centrifuged (14000 × g, 5 min, 4 • C). The DNA was precipitated by adding one volume of cold isopropanol (−20 • C) over 24 h. The mix was then centrifuged for 30 min at 14,000 g (4 • C), the resulting pellet was washed with 70% ethanol, and the DNA was resuspended with 100 μL of TE buffer (pH 8). The DNA was purified in two steps: first using polyvinylpolypyrrolidone (PVPP) microbiospin columns (Bio-Rad, CA, USA), and then a Sepharose 4G column (Sigma-Aldrich, United Kingdom). The DNA quality was checked by electrophoresis on agarose gel and quantified by spectrofluorometer using the Quant-iT PicoGreen R dsDNA Assay Kit (Invitrogen, Cergy-Pontoise, France) following the manufacturer's instructions.
The abundance of denitrifiers was assessed by real-time quantitative PCR (qPCR) by targeting N 2 O-producers, nirK and nirS Kandeler et al., 2006) and N 2 O-reducers nosZI (Henry et al., 2006;Jones et al., 2013) and nosZII (Table S3). Abundance of total bacteria was assed using 16S rRNA primers (Muyzer et al., 1993) as previously described . qPCR Reactions were carried out in a StepOnePlus Real time PCR System (Life Technologies, Carlsbad, CA, USA). The abundance was based on the increasing fluorescence intensity of the SYBR Green dye during amplification. The qPCR assay was carried out in a 15 μl reaction volume containing 1 ng of DNA, 7.5 μl of SYBRgreen PCR Master Mix (Absolute qPCR SYBR GreenRox, Thermo, Courtaboeuf, France), 1 μM of each primer, 250 ng of T4 gene 32 (QBiogene, Illkrich, France). Before assessing the abundance of the bacterial communities, an inhibition test was performed by mixing DNA extracts with a known amount of control plasmid DNA and no inhibition was detected. Three independent quantitative qPCR assays were performed for each gene. Controls and no-template controls giving null or negligible values were run for each quantitative qPCR assay. The qPCR efficiencies for the various genes ranged between 70 and 96%.

Phylogenetic Diversity of N 2 O-reducers
A diversity analysis of nosZI and nosZII was performed by 454 pyrosequencing as previously described in Jones et al. (2014). Briefly, the DNA was prepared using a two-step PCR procedure (Berry et al., 2012). In the first step, 20 PCR cycles were performed with primers nosZI and nosZII (Table S4)  In the second PCR 15 or 18 cycles PCR were performed using the forward primers preceded by 10 bp-long barcodes, the sequencing key and the forward sequencing adapter; the reverse primers being only preceded by the sequencing key and the reverse sequencing adapter as described in Jones et al. (2014) (Table S4).Because there were only very small amounts of products for nosZII after the first PCR, the second PCR for this gene was extended to 18 cycles. The product of 3 independent second PCR was then gel extracted and purified using the QIAEX II kit (Qiagen; France). Pyrosequencing was performed by Genoscreen (Lille, France) on a Roche's 454 FLX Genome Sequencer according to manufacturer's instructions.

Sequence Processing
The QIIME pipeline (Caporaso et al., 2010a) was used for quality trimming of raw 454 pyrosequencing data (QIIME version 1.8.0). The minimum and maximum sequence lengths were 230 and 410 bp respectively. Sequences with an average score below 25 using a sliding window of 50 bp were discarded. After quality checking, 123,130 sequences were found for nosZI and 121,500 sequences for nosZII. Sequences were then processed using the "pick_otus.py" script within QIIME, and the "usearch" option (Edgar, 2010) with reference-based and de novo chimera checking, and clustering of sequences at 97% similarity. Raw sequences were deposited at the NCBI under the accession number SRP058080. The process of raw sequence submission was greatly simplified by using the make.sra command of Mothur software (Schloss et al., 2009).

nosZ phylogeny
Reference sequences for nosZ were downloaded from all 4135 draft and completed microbial genome nucleotide sequences available in the National Center for Biology Information (NCBI) . These reference sequences were used as templates for aligning 454 reads with PYNAST (Caporaso et al., 2010b). Phylogenetic trees for nosZI and nosZII were constructed with fastree (Price et al., 2010) and ITOL was used to visualize and manipulate of the trees (Letunic and Bork, 2007).

Statistical Analysis
Statistical analysis and graphics were produced using the R statistical software, R version 3.0.3, (R Core Team, 2013) and the agricolae (Mendiburu, 2014) and vegan (Oksanen et al., 2015) packages. The effect of agricultural practices and cropping systems was determined by analysis of variance and post hoc Tukey HSD test. Collinearity between explaining variables within each group (soil properties, microbial community abundances, and denitrifiers diversity) was checked, and one of each pair of collinear variables was kept for subsequent analyses. Non-metric MultiDimensional Scaling (NMDS) of the Unifrac distance matrices (unweighted and weighted) was used to describe community structure. Ordinations with the lowest stress values were used. The soil properties, community abundances and diversity were plotted onto the ordination map as vectors. Permutation tests (n = 10000) were used to test the significance of vector fits and only significant ones were depicted (P < 0.05). Vector and surface fitting of variables within ordinations were performed using the envfit and ordisurf functions in the vegan package respectively. ANalysis Of SIMilarity (ANOSIM) was used to test for significant differences in community structure between cropping systems (permutations = 1999, P < 0.05).
Significant explanatory variables of [rN 2 O/r(N 2 O+N 2 )], potential N 2 O and PDA were chosen by linear regression and model selection (backward) and by minimizing the Akaike Information Criterion (AIC). The statistical significance was assessed by 1000 permutations of the reduced model. The resulting significant explanatory variables (Table S2) were used to access their contribution to explaining the variation of potential N 2 O, PDA and [rN 2 O/r(N 2 O+N 2 )], using the function varpart (Peres-Neto et al., 2006).