Effect of the nitrification inhibitor 3,4-dimethylpyrazole phosphate (DMPP) on N-turnover, the N2O reductase-gene nosZ and N2O:N2 partitioning from agricultural soils

Nitrification inhibitors (NIs) have been shown to reduce emissions of the greenhouse gas nitrous oxide (N2O) from agricultural soils. However, their N2O reduction efficacy varies widely across different agro-ecosystems, and underlying mechanisms remain poorly understood. To investigate effects of the NI 3,4-dimethylpyrazole-phosphate (DMPP) on N-turnover from a pasture and a horticultural soil, we combined the quantification of N2 and N2O emissions with 15N tracing analysis and the quantification of the N2O-reductase gene (nosZ) in a soil microcosm study. Nitrogen fertilization suppressed nosZ abundance in both soils, showing that high nitrate availability and the preferential reduction of nitrate over N2O is responsible for large pulses of N2O after the fertilization of agricultural soils. DMPP attenuated this effect only in the horticultural soil, reducing nitrification while increasing nosZ abundance. DMPP reduced N2O emissions from the horticultural soil by >50% but did not affect overall N2 + N2O losses, demonstrating the shift in the N2O:N2 ratio towards N2 as a key mechanism of N2O mitigation by NIs. Under non-limiting NO3− availability, the efficacy of NIs to mitigate N2O emissions therefore depends on their ability to reduce the suppression of the N2O reductase by high NO3− concentrations in the soil, enabling complete denitrification to N2.


Results
Physical and chemical properties for the two soils used in this experiment are shown in Table 1. The contrasting soils, a horticultural and a pasture soil, are henceforth referred to as sandy clay loam (sandy CL) and loam, according to their texture from 0-10 cm.
Nitrogen transformations and soil microbial parameters. Gross N transformation rates were quantified with a 15 N tracing model (Fig. 1) and differed markedly between soils when N-fertilizer was applied without the NI DMPP, referred to as the fertilizer only treatment ( Table 2). Gross mineralization rates (M tot ) in the loam exceeded those in the sandy CL by a factor of 39. In the loam, M tot was dominated by the mineralization of labile N (M Nlab ), while the mineralization of recalcitrant organic N (M rec ) dominated in the sandy CL. Gross nitrification (Nit tot ) was higher in the loam with 18.7 ± 0.03 μg N g −1 soil day −1 compared to 5.8 ± 0.03 μg N g −1 soil day −1 in the sandy CL. Autotrophic nitrification (O NH4 ) was the main pathway of NO 3 − production in both soils, as heterotrophic nitrification of organic N (O Nrec ) accounted for only 7% of Nit tot in the sandy CL, and was negligible for Nit tot in the loam. Immobilization of NH 4 + (I NH4tot ) and NO 3 − (I NO3 ) was higher in the sandy CL compared to the loam, and was dominated by I NO3 . In the sandy CL, only minor amounts of NO 3 − were recycled in the NH 4 + pool via dissimilatory NO 3 − reduction to NH 4 + (DNRA, referred to as D NO3 in the 15 N tracing model), while D NO3 contributed with more than 2 μg N g −1 soil day −1 to NH 4 + production in the loam. Microbial C (C mic ) and N (N mic ) as indicators for the size of the soil microbial biomass (SMB) were higher in the loam, exceeding C mic and N mic in the sandy CL by a factor of 5 and 7, respectively (Table 3).

Emissions of N 2 O and N 2 .
The dominant N 2 O production pathway in both soils was denitrification, accounting for more than 90% of the N 2 O produced (Fig. 2). Over 48 hours, 0.24 ± 0.03 and 1.46 ± 0.38 μg N 2 O -N g −1 soil were emitted from the sandy CL and the loam, respectively. Both N 2 O emissions via denitrification (N 2 O d ) and nitrification (N 2 O n ) were higher from the loam, exceeding those from the sandy CL by a factor of >8 (Fig. 2). Over the two day incubation period, 0.47 ± 0.09 μg N 2 -N g −1 soil and 0.87 ± 0.11 μg N 2 -N g −1 soil were emitted as N 2 from the sandy CL and the loam, respectively. The main product of denitrification (N 2 O d + N 2 ) from the sandy CL was N 2 , with N 2 O d accounting for 36% of total denitrification losses. Denitrification losses from the loam however were dominated by N 2 O d , accounting for 75% of total denitrification. There was no indication for hybrid production of N 2 O or N 2 .
The response of the N 2 O reductase gene nosZ to fertilization and the use of DMPP. The abundance of nosZ prior to fertilization differed markedly between soils (Fig. 2). Copy numbers of nosZ in the loam exceeded those in the sandy CL by a factor of 6. After fertilization and the increase in soil moisture from 50% to 75% water-filled pore space (WFPS), nosZ copy numbers decreased in both soils, with a reduction by 77% and 32% for the sandy CL and the loam, respectively. DMPP did not affect nosZ abundance in the loam. DMPP however increased nosZ copy numbers by 227% compared to the fertilizer only treatment in the sandy CL.
Effect of DMPP on N 2 O and N 2 emissions. DMPP significantly reduced N 2 O emission from the sandy CL but had no effect on N 2 O emissions from the loam (Table 4). DMPP reduced N 2 O d from the sandy CL by 46% (P < 0.05), but did not affect N 2 O n (Fig. 2). There was no effect of DMPP on N 2 emissions from the two soils. In the sandy CL, DMPP shifted the product ratio of denitrification (N 2 O d /(N 2 O d + N 2 )) to N 2 , decreasing the percentage of denitrification emitted as N 2 O d from 36% to 20%.

Discussion
The fertilization and irrigation of agricultural soils triggers a cascade of N transformations associated with pulses of N 2 O and N 2 emissions. These short-term events are critical to understand the effects of NIs on N 2 O production and consumption in agricultural soils. Linking N turnover to emissions of N 2 O and N 2 and the abundance of the N 2 O reductase gene nosZ in a short-term incubation demonstrated (a) that increasing NO 3 − availability after fertilization suppressed nosZ abundance, (b) that nosZ abundance, nitrification and N 2 + N 2 O emissions remained largely unaffected by DMPP in the loam and (c) that DMPP decreased nitrification and increased nosZ abundance in the sandy CL, shifting the N 2 :N 2 O ratio towards N 2 . Our findings highlight the short-term effect of DMPP as highly soil specific, and show that reduced nitrification by DMPP can limit the suppression of the N 2 O reductase by high NO 3 − concentrations in the soil, enabling complete denitrification to N 2 . Nitrogen transformation rates identified the loam as the more active soil regarding N turnover compared to the sandy CL (Table 2). Gross mineralization rates (M tot ) of more than 8 μg N g −1 soil day −1 together with a low immobilization of mineral N (I NH4tot and I NO3 ) denote high mineral N availability due to the rapid mineralization of organic N. This is further supported by the dominant contribution of the labile organic N pool to mineralization (M Nlab ), representing the microbial biomass and low molecular organic N compounds with a fast turnover. The high nitrification rates in the loam (>18 μg N g −1 soil day −1 ) denote rapid conversion of mineralized N to NO 3 − and show the dominant role of NH 4 + oxidation for N-turnover in this soil. Gross mineralization was markedly lower in the sandy CL with M tot at only 0.21 μg N g −1 soil day −1 and dominated by the mineralization of recalcitrant organic N, indicating limited and slower supply of mineral N via mineralization. Mineralization accounted for only 4% of nitrified N in the sandy CL, as compared to 45% in the loam, implying a rapid depletion of the NH 4 + pool in the sandy CL. The observed differences between soils are consistent with microbial C and N contents (Table 3), indicating a larger soil microbial biomass in the loam and reflect the impact of perennial versus short term/annual and tilled versus undisturbed plant-systems on soil organic matter and microbial activity: Intensive tillage and irrigation in horticultural systems lead to loss of soil organic C 21 , while an extensive root system under permanent pasture is likely to promote microbial activity through constant inputs of C and N. These findings establish the differences in magnitude and relative importance of N transformations and microbial activity between the two contrasting soils.
The main source of N 2 O in both soils was denitrification, accounting for more than 90% of N 2 O produced ( Fig. 2), which is in line with previous results from both field 11 and laboratory studies 10,12 . The ability of soils to act as an N 2 O sink, i.e. the trait to reduce N 2 O to N 2 has been linked to the abundance of nosZ, used as proxy for microorganisms involved in the reduction of N 2 O. In the study presented here, we compared nosZ abundance with direct measurements of N 2 and N 2 O, evaluating the influence of DMPP on of microorganisms reducing N 2 O. The abundance of nosZ prior to fertilizer addition was higher in the loam, which is consistent with the reported positive correlation of nosZ copy numbers with soil organic C 22 . The synthesis of the N 2 O reductase is promoted by anoxic conditions 23 , and the increase in soil moisture together with the addition of fertilizer should have increased nosZ abundance. However, nosZ abundance decreased in both soils in the fertilizer only treatment (Fig. 2), indicating that increased NO 3 − availability due to fertilization and nitrification promoted the reduction of NO 3 − rather than N 2 O, shifting the N 2 O d /(N 2 O d + N 2 ) ratio towards N 2 O. The magnitude and N 2 O:N 2 partitioning of denitrification losses is consistent with the nitrification rates in both soils and as such shows the N 2 O d /(N 2 O d + N 2 ) ratio as a function of soil intrinsic N -turnover. Cumulative N 2 O d losses of >2 μg N g −1 soil and 75% of denitrification (N 2 O d + N 2 ) emitted as N 2 O from the loam show increased substrate availability for denitrification and simultaneous suppression of nosZ abundance by high NO 3 − availability (Fig. 2). In turn, lower denitrification losses with only 36% emitted as N 2 O d reflect slower N turnover in the sandy CL. These findings suggest that the suppression of the N 2 O reductase and increased N substrate availability are responsible for the large pulses of N 2 O from agricultural soils observed after fertilization and irrigation. Our results denote an increased risk of N 2 O loss from highly productive agricultural soils 19 , where increased mineralization of soil organic N due to fertilization, i.e., priming is likely to amplify the preferential reduction of NO 3 − , and as such the production of N 2 O via denitrification. DMPP reduced N 2 O emissions from the sandy CL by more than 54% (Table 4). This is reflected in DMPP's effect on autotrophic nitrification (O NH4 ) showing a reduction of 63% in the sandy CL ( Table 2). The minor reduction of O NH4 by DMPP had however no effect on N 2 O emissions from the loam. In both soils, N 2 O derived from nitrification mediated pathways accounted for less than 15% of overall N 2 O, showing no response to the DMPP treatment. For the sandy CL, this suggests that DMPP primarily affected N 2 O production pathways indirectly, that is by reducing NO 3 − availability for denitrification, demonstrated by the reduction of N 2 O derived from denitrification by 46%. DMPP increased nosZ abundance in the sandy CL by a factor >2 compared to the fertilizer only treatment (Fig. 2). In the absence of direct N 2 measurements, this effect has been interpreted as a shift of denitrification losses towards N 2 24 . Experimental evidence linking increased nosZ abundance with DMPP to N 2 and N 2 O emissions 25 is based on the acetylene inhibition method, which has been shown to lead to an irreproducible underestimation of denitrification rates 9 . Furthermore, acetylene itself is a potent NI, questioning the use of this method when investigating the effects of NIs on the magnitude and the   (Fig. 2). These results link the increase of nosZ abundance in response to DMPP in the sandy CL to a shift in the N 2 O d :N 2 ratio towards N 2 , based on direct measurements of N 2 and N 2 O d using the 15 N gas flux method. In contrast to previous incubation studies investigating N 2 O:N 2 partitioning in response to DMPP 26,27 , emissions of N 2 O and N 2 were quantified after incubation under atmospheric O 2 conditions and without adding an easily available C source to stimulate denitrification, as these conditions would have altered short-term N dynamics in response to DMPP. Importantly, the shift towards N 2 was not observed for the loam, where DMPP had a negligible effect on nitrification. Our findings indicate that the reduction of nitrification by DMPP in the sandy CL reduced the suppression of the N 2 O reductase after fertilization, enabling complete denitrification to N 2 . Emissions of N 2 O produced via nitrification mediated pathways were not affected by DMPP in this soil, showing the reduction of N 2 O emissions by DMPP as an indirect effect limiting NO 3 − availability for denitrification. The spatial coverage of nitrifying microsites by the inhibitor is critical for efficient inhibition of nitrification. Limited diffusion of DMPP may explain the he observed inefficacy of DMPP to reduce autotrophic nitrification in the loam, which is consistent with reports from other pasture soils 15 . The amount of DMPP applied with N fertilizer is small, and the initial sorption to organic matter and uneven distribution of DMPP may hinder its short-term effectiveness to reduce nitrification in specific micro sites. Sorption of DMPP is likely to be more pronounced in the loam as a pasture soil with higher organic matter content as compared to the sandy CL owing to the positive correlation of DMPP sorption with organic C 28,29 . The high microbial activity in the loam also infers a larger number of microsites with nitrifying activity compared to the sandy CL, suggesting the spatial separation of DMPP from nitrifiers may be responsible for the short-term inefficacy of DMPP to reduce autotrophic nitrification in the loam. This theory is further supported by a study where DMPP did not affect the initial pulse of N 2 O after fertilization and irrigation from the loam, but reduced denitrification losses after that initial period 11 . This shows a delayed effect of DMPP in this soil, demanding further research on how diffusion in the soil matrix, sorption and distribution affects DMPPs efficacy to reduce autotrophic nitrification.
DMPP also affected non-targeted N transformation in the sandy CL: Mineralization and immobilization turnover was stimulated by DMPP, demonstrated by the five-fold increase of total mineralization (M nrec + M nlab ) and the simultaneous increase of NH 4 + immobilization (I NH4rec + I NH4lab ) by a factor > 2 ( Table 2). Increased mineralization/immobilization turnover has been reported after the application of DMPP 15 and dicyandiamide (DCD) 14 and can be attributed to higher NH 4 + availability, stimulating microbial immobilization of NH 4 (I NH4lab ) and mineralization of labile N org (M Nlab ) to NH 4 + . This effect may further prime the mineralization of recalcitrant N (M Nrec ) in response to DMPP 30 . Interestingly, DMPP increased DOC availability in both soils, confirming previous results from a wheat-maize cropping system 31 (  Table 3. Soil mineral N concentrations 30 minutes and 48 hours after N fertilizer application with and without the nitrification inhibitor DMPP; and dissolved organic C and soil microbial C and N prior and 48 hours after fertilizer application with and without DMPP in a horticulture and a pasture soil. Letters denote significant differences between treatments within a soil. *denote significant differences (P < 0.05) between soils within a treatment. (2020) 10:2399 | https://doi.org/10.1038/s41598-020-59249-z www.nature.com/scientificreports www.nature.com/scientificreports/ organic matter induced by DMPP contributed to higher DOC availability, but no such effect was observed for M Nrec in the loam. Based on the data available, it remains unclear what caused the increase in DOC in response to DMPP. This increase has however important implications for N-turnover, in particular for the sandy CL as soil with limited labile C availability. DMPP increased DNRA by a factor >5 in the sandy CL, suggesting labile C promoted NO 3 − consumption via DNRA 10,23 . DNRA competes with denitrification for available NO 3 − , but the magnitude of DNRA in the sandy CL was insignificant regarding NO 3 − availability for denitrification. More importantly, labile C affects denitrification 32 , by supplying a reductant for denitrifiers, or through the stimulation of heterotrophic soil respiration, decreasing soil O 2 levels and thus promoting denitrification. Furthermore, readily decomposable C can decrease the N 2 O d /(N 2 O d + N 2 ) ratio of denitrification 23 . The increase in DOC observed in this study demonstrates an important non-targeted effect of DMPP, which can alter both rate and N 2 O:N 2 partitioning of denitrification losses and therefore warrants further research. www.nature.com/scientificreports www.nature.com/scientificreports/ Nitrification activity during pre-incubation increased NO 3 − levels in both soils. In the loam, NO 3 − levels were above those measured at the respective field site, which is also reflected in higher N 2 O d /(N 2 O d + N 2 ) ratios 11 . This phenomenon often occurs in incubation studies, where the absence of plant uptake, pre-incubation 33,34 , and the addition of glucose 26 increases NO 3 − levels in the soil. It is therefore important to consider N substrate availability when interpreting the effects of NIs on rate and N 2 O:N 2 partitioning of denitrification losses. The mineral N levels in both soils indicate no N substrate limitation for denitrification regardless of the treatment. Under these conditions, DMPP had no effect on overall denitrification losses in both soils. The minor reduction of nitrification by DMPP in the loam did not reduce NO 3 − availability to a degree that limited preferential reduction of NO 3 − . The high initial NO 3 − values in the loam are also likely to have overwritten a significant reduction of nitrification. The reduction of N 2 O emissions, together with the increase of nosZ abundance in the sandy CL suggests however that DMPP lowered NO 3 − availability below a soil specific treshold 35 , limiting the preferential reduction of NO 3 − over N 2 O. The results from the sandy CL confirm the proposed mechanism of N 2 O reduction via a shift in the N 2 :N 2 O ratio 26 , and show that DMPPs inhibitory effect on nitrification can limit the suppression of the N 2 O reductase, promoting complete denitrification to N 2 .
The demonstrated link between nosZ and directly measured N 2 O and N 2 emissions suggests that DMPP promotes the abundance of nosZ carrying denitrifiers. Including a comprehensive assessment of abundance and activity of nitrifying and denitrifying microbial communities in future research could further help to understand mechanisms of N 2 O mitigation by DMPP. Our study shows N dynamics in response to DMPP on a soil microcosm scale. This approach does not account for plant-microbe interactions and plant N uptake under field conditions but enables to isolate effects of NIs on key N transformations, with practical implications for the use of NIs in different agricultural soils. The relative magnitude of N 2 O emissions reflects cumulative losses observed from the same soils in the field, demonstrating a larger N 2 O mitigation potential for the pasture soil as compared to the horticultural soil. The short term inefficacy of DMPP to reduce nitrification in the pasture soils demands therefore improved strategies regarding rate and application of NIs. In soils with high organic matter content, and high soil intrinsic N turnover, repeated applications of DMPP, increasing the rate of DMPP, and/or the application of DMMP prior to fertilization may increase DMPPs efficacy, limiting the effect of N fertilizer priming on N 2 O emissions. Decreased nosZ abundance after fertilization and irrigation indicates suppression of the N 2 O reductase by increased NO 3 − availability, identifying NO 3 − availability as the control for the reduction of NO 3 − vs. N 2 O, which determines the magnitude of N 2 O losses. These findings apply to conditions of non-limiting NO 3 − availability for overall denitrification, which can be found in agricultural soils after N fertilization and irrigation when plant N uptake is limited. Under these conditions, the efficacy of NIs to mitigate N 2 O emissions depends on their ability to limit the suppression of the N 2 O reductase by high NO 3 − concentrations in the soil, enabling complete denitrification to N 2 .

Material and Methods
Soils and site. Soil samples (0-10 cm) were collected randomly (n = 4) from a vegetable cropping site (Gatton, Qld) 20 and an intensively managed dairy pasture (Gympie, Qld) 11 in subtropical Australia, referred to according to their texture in the first 10 cm as sandy clay loam (sandy CL) and loam, respectively. Site characteristics including physical and chemical soil properties are shown in Table 1. Soil samples were bulked, air dried and sieved to <4 mm and stored in a cold room at 4 °C. Soil microcosms. Before treatment application, the soils were incubated in bulk for 4 days at a gravimetric water content of 30%. The experimental design consisted of two soils and two treatments: ammonium nitrate (NH 4 NO 3 ) and NH 4 NO 3 with DMPP (DMPP), each with four different 15 N label combinations and four replicates. The NH 4 NO 3 was applied with either (a) the NH 4 + ( 15 NH 4 NO 3 − ) or (b) the NO 3 − (NH 4 15 NO 3 − ) labeled at 10 atom %. NH 4 15 NO 3 − at 60 atom % (c) was used to quantify N 2 emissions 36 , while non-labeled NH 4 NO 3 (d) was used for the quantification of the SMB, DOC, and nosZ abundance. For the incubation, soil microcosms were established in centrifuge tubes (50 ml) using the equivalent of 8 g oven dry soil at a soil bulk density of 1 g cm −3 . NH 4 NO 3 equivalent to 35 µg N g −1 soil was applied in solution (1 ml) with 0.6% DMPP (w/w) added for the DMPP treatment. Additional water was applied to achieve the water-filled pore space (WFPS) of 75%. Water and fertilizer solutions were applied dropwise on two layers of 4 g of soil to ensure homogenous 15 N labeling. After fertilization, centrifuge tubes were closed with Suba-seals (Sigma Aldrich) and were kept closed in an incubator at a constant temperature of 25 °C between gas sampling events. Additional soil microcosms (a and b, n = 4) were established for destructive sampling 30 minutes after fertilizer application.  www.nature.com/scientificreports www.nature.com/scientificreports/ Soil analysis. Soil mineral N. All soil mineral N extractions were conducted in the centrifuge tubes to avoid subsampling errors using 40 ml 2 M KCl (1:5 w/v ratio). Four soil microcosms per soil were extracted before fertilizer application to determine initial conditions. Soil microcosms a and b were extracted with 40 ml 2 M KCl, 30 minutes (t = 0) and 48 h (t = 2 days) after N fertilizer application. The centrifuge tubes were shaken with a horizontal shaker (150 rpm) for one hour, and extracts were filtered through Whatman no. 42 filter paper. After sample dilution, concentrations of NH 4 + and NO 3 − were determined using colorimetric methods, NH 4 + with a modified indophenol reaction 37 and NO 3 − with the VCL3/Griess assay 38 . The 15 N enrichments of the NH 4 + and NO 3 − pool were determined for soil microcosms a and b by the diffusion method 39 .
Quantitative PCR analyses. For qPCR analysis, subsamples of 0.25 g of soil were taken prior to fertilizer application, and 48 h after (t = 2 days) from soil microcosms d and extracted immediately for total DNA using the Soil microbial biomass. Microbial C (C mic ) and N (N mic ) were quantified before and two days after fertilizer application using the chloroform fumigation-extraction 41 . Two aliquots of 3.5 g soil were sampled from each soil microcosm (d) with one aliquot subsequently fumigated with chloroform for 24 h. Fumigated and non-fumigated samples were extracted with 2 M KCl (1:10 w/v) and stored frozen until further analyses. Samples were acidified to remove inorganic C and analyzed for total N and organic C with an automated TOC/TN analyzer (TOC-V CPHE200V) linked with a TN-unit (TNM-1 220 V, Shimadzu Corporation, Kyoto, Japan). C mic and N mic were calculated as the difference in N and C between fumigated and non-fumigated samples without using a correction factor 42 . Dissolved organic C (DOC) was quantified as the amount of total C in the extracts of the non-fumigated samples.
Gas sampling and analysis. Air samples (n = 4) were taken daily before closing the centrifuge tubes to quantify ambient N 2 O concentrations. Specific background samples were taken above the respective soil microcosms treated with NH 4 15 NO 3 at 60 atom % (c) for 15 N 2 analysis before closing the tubes. The entire headspace atmosphere was sampled 24 and 48 h after closure using a gas-tight syringe from soil microcosms a, b and c. After the 24 h gas sampling, the Suba-seals were removed for 10 minutes, allowing the headspace atmosphere to equilibrate 10 . Gas samples were transferred into pre-evacuated 12 ml exetainer tubes with a double wadded Teflon/silicon septa cap (Labco Ltd, Buckinghamshire, UK) and stored until N 2 O and CO 2 analysis by gas chromatography (Shimadzu GC-2014). Gas samples from soil microcosms c were also analyzed for the isotopologues of N 2 ( 15 N 14 N, 15  Fluxes of N 2 , N 2 O and CO 2 . The triple labelling approach generates gas samples from three 15 N fertilizer treatments with four replicates: a,b and c. Cumulative N 2 O and CO 2 fluxes given in Table 4, were calculated based on gas samples from 15 N fertilizer treatments a,b and c. Fluxes of N 2 and N 2 O d , as well as denitrification losses (N 2 + N 2 O d ), were calculated based on the gas samples from treatment c. Calculating cumulative N 2 O fluxes based on 15 N fertilizer treatments a, b and c or c alone did not result in significant differences. The reduction of N 2 O by DMPP in the sandy CL was significant regardless of the calculation chosen.
The flux rates of N 2 O and CO 2 were calculated from the slope of the linear increase in gas concentration during the closure period and were corrected for temperature and air pressure 20 . The 15 N enrichment of the NO 3 − pool undergoing denitrification (a p ) and the fraction of N 2 and N 2 O emitted from this pool (f p ) were calculated following the equations given by Spott, et al. 43 detailed in the supplementary material. The headspace concentrations of N 2 O and N 2 were multiplied by the respective f p values giving N 2 and N 2 O produced via denitrification (referred to as N 2 and N 2 O d ), with their respective fluxes expressed in g N 2 or N 2 O d -N emitted g −1 soil day −1 . Potential hybrid formation of N 2 and N 2 O was found to be irrelevant 30 . The precision of the IRMS for N 2 based on the standard deviation of atmospheric air samples (n = 18) at 95% confidence interval was 4.4 × 10 −7 and 6.0 × 10 −7 for 29 R and 30 R, respectively. The corresponding method detection limit ranged from 0.005 µg N 2 -N g −1 soil with a p assumed at 50 atom % to 0.014 µg N 2 -N g −1 soil with a p assumed at 20 atom %.
Gross N transformations. Gross N transformations were quantified using a 15 N tracing model 44 (Fig. 1), which uses a Markov Chain Monte Carlo method optimizing the kinetic parameters for the various N transformations by minimizing the misfit between modeled and observed NH 4 + and NO 3 − concentrations and their respective 15 N enrichments (soil microcosms a and b). The model considers five N pools including the NH 4 + and NO 3 − pool, a labile (N lab ) and a recalcitrant (N rec ) organic N pool, and a pool for NH 4 + adsorbed to cation exchange sites (NH 4 + ads ). These pools are defined by 10 simultaneous occurring gross N transformations calculated by zero-, first-order or Michaelis-Menten kinetics ( Calculations and statistical analysis. The optimization routine used for the 15 N tracing model gives a probability density function for each model parameter, which is used to calculated average values and standard errors of the mean. Average gross N transformation rates are obtained by integrating these values over the incubation period. Differences between N-transformations were assessed testing whether the 95% confidence intervals overlap 45 . The Benjamini Horchberg (BH) adjustment 46 was performed to assess the effect of the different fertilization strategies on microbial C and N, DOC and nosZ gene abundance for each soil type. Analysis of variance was performed to assess differences in cumulative emissions of N 2 , N 2 O, total denitrification (N 2 + N 2 O) and CO 2 between soils within treatments and within soils between fertilization strategies. All values unless otherwise stated are given as mean ± standard error of the mean.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary  Information files).