Linkage between N2O emission and functional gene abundance in an intensively managed calcareous fluvo-aquic soil

The linkage between N2O emissions and the abundance of nitrifier and denitrifier genes is unclear in the intensively managed calcareous fluvo-aquic soils of the North China Plain. We investigated the abundance of bacterial amoA for nitrification and narG, nirS, nirK, and nosZ for denitrification by in situ soil sampling to determine how the abundance of these genes changes instantly during N fertilization events and is related to high N2O emission peaks. We also investigated how long-term incorporated straw and/or manure affect(s) the abundance of these genes based on a seven-year field experiment. The overall results demonstrate that the long-term application of urea-based fertilizer and/or manure significantly enhanced the number of bacterial amoA gene copies leading to high N2O emission peaks after N fertilizer applications. These peaks contributed greatly to the annual N2O emissions in the crop rotation. A significant correlation between annual N2O emissions and narG, nirS, and nirK gene numbers indicates that the abundance of these genes is related to N2O emission under conditions for denitrification, thus partly contributing to the annual N2O emissions. These findings will help to draw up appropriate measures for mitigation of N2O emissions in this ‘hotspot’ region.

practice is a very intensive double-cropping cereal system with irrigated winter wheat and rain-fed summer maize rotations characterized by the application of large amounts of fertilizer N and irrigation with large amounts of groundwater to obtain relatively high yields 16 . These practices lead to substantial total N 2 O emissions in this region which has become a 'hotspot' of national N 2 O emissions with global significance 17 . Our laboratory and field experiments on these intensively managed low-carbon calcareous soils over numerous years have found that high peaks of N 2 O emissions (mainly occurred within the first week after fertilization) were always induced by the application of large amounts of NH 4 + -or urea-based fertilizers to the soils, with no strong emissions during other periods; even the high nitrate concentration in the soils contribute only a small proportion of the annual N 2 O emissions 18 . Moreover, a total of 67-88% of the N 2 O is emitted during the summer maize season in the annual N 2 O emissions [18][19][20][21][22][23][24][25][26][27] . The N 2 O emission factors are generally lower than the IPCC (2006) 28 default value (1%), with 0.10-0.59% on an annual basis, 0.08-0.21% in the winter wheat season, and 0.44-0.59% in the summer maize season 16,18,20,22,26,27 . We speculated that high ammonia oxidation activity linked to denitrification by heterotrophic denitrifiers or by nitrifiers are the major processes causing the observed instant high N 2 O peaks 18,24,27,29 . In the present study we quantify the relevant functional genes of the soil microbial community to further elucidate the biological mechanism of N 2 O emissions in this important region.
The community composition of both nitrifiers and denitrifiers is controlled by distal and proximal factors 30 . Distal factors are those factors that control the composition and diversity of nitrifying and denitrifying communities over the long term, and the proximal factors are those factors that affect instantaneous nitrification and denitrification rates. The ammonia oxidizing bacteria (AOB) community could be shaped by distal factors such as soil pH, ammonia and oxygen availability through which proximal factors such as application of NH 4 + -or urea-based fertilizer and/or manure and temporal soil temperature and moisture affect the nitrification rate [31][32][33][34] . Denitrifier community composition is structured over the long term by distal controls including the temperature and moisture conditions, substrate availability, competition and disturbances. The rate and kinetics of denitrification at any particular moment are controlled by proximal factors such as oxygen, carbon and nitrate availability 35 .
It is still not clear how the abundance of the bacterial amoA gene for nitrification and functional genes for denitrification respond instantly to N fertilization, how they relate to high N 2 O emission peaks, or how long-term incorporated straw and/or manure affect these linkages in the intensively managed low-carbon calcareous soils of the NCP. Our hypotheses are as follows. Firstly, under temporally high NH 4 + concentrations the ammonia oxidizing bacteria will actively grow and accumulate high concentrations of NO 2 − , leading to anoxic conditions which, in turn, induce denitrification by heterotrophic denitrifiers, and the above chain processes cause the high N 2 O emission peaks after N fertilization. Secondly, long-term incorporated straw and/or manure will increase soil organic carbon (SOC) and total nitrogen (TN) and lead to increased abundance of denitrification functional genes as a distal control and the denitrifiers act under depleted oxygen conditions when rainfall or irrigation occur, thus partly contributing to annual N 2 O emissions.
We have therefore quantified the functional genes in soils that were sampled in situ three times and have performed an Illumina-based 16S rRNA gene sequencing analysis in the 2012-2013 winter wheat-summer maize rotation based on a long-term field experiment that began in 2006 at the Shangzhuang Research Station in suburban Beijing. N 2 O emissions and the concentrations of ammonium, nitrite and nitrate were determined at the same time. The field experiment had been running for seven years at the time of soil sampling. We were therefore able to effectively quantify the relatively long-term effects of different fertilization regimes including chemical fertilizer N combined with straw and/or manures on the total bacterial copy numbers (16S rDNA) and the populations of ammonia-oxidizing and heterotrophic denitrifying bacteria.

Results
Changes in the soil chemical and biological properties after 7 years. The SOC in N opt , CN opt and CM increased by 15.8, 18.5, and 67.1%, and the TN correspondingly increased by 10.1, 15.2 and 52.5%, respectively, compared with the N 0 treatment. This indicates that all N fertilization treatments tended to increase SOC and TN compared to the N 0 control but the effect was only statistically significant for the manure treatment ( Table 1). The soil C/N ratio and NH 4 + concentration did not change significantly, and the ammonium concentration was very low, approximately 2 mg N kg −1 , among the four treatments.
The nitrate concentration in the N opt , CN opt and CM treatments was 38.5, 36.5 and 26.2 mg N kg −1 , respectively, approximately 1.5 to 2.6 times higher than in the N 0 control. The soil potential nitrification rate in N opt , CN opt and CM was 3.39, 3.90 and 3.80 times significantly higher than in the N 0 control, respectively. The soil potential denitrification rate in N opt , CN opt and CM was 1.09, 1.80 and 3.12 times higher compared with the N 0 control; CN opt and CM were significantly higher than in the N opt and N 0 treatments, respectively; and only the CM treatment was significantly higher than the other three treatments. The soil pH measured on the sampling date of 16 th April was approximately 7.57-7.75 and was not significantly different among the four treatments (Table S3). However, it was significantly lower in the three fertilization treatments compared to the N 0 control when measured on the sampling dates of 9 th August and 14 th August, indicating that urea hydrolysis and nitrification would reduce the soil pH shortly in the calcareous soil studied 36 .
The above changes in the C and N status among treatments are likely the most important distal drivers of bacterial community composition, the abundance of nitrifier and denitrifier genes and related to instant generation of N 2 O by proximal drivers. Nitrous oxide emission. N 2 O fluxes were very low on 16 th April and there were no significant differences among the four treatments ( Fig. 1a) during this zero-N fertilization period; there were even negative fluxes in the N 0 and N opt treatments which may be explained by the low soil temperature (9.6 °C) and moisture content (approximately 11%) in April (Table S3) 20,23 . The fluxes appeared to be higher on 9 th August than on 16 th April in all treatments but were not significantly different (although CM treatment was the highest) due to the rising soil temperature (24-26 °C) and moisture content (approximately 16%) in August, and the higher concentrations of nitrite and nitrate in the soil matrix (Table S3; Fig. 2a,b,c). Surprisingly, there was still 3.2-10.4 mg N kg −1 of nitrite in the fertilized soils on 9 th August, which is unusual in the soil studied (further explanation in the Discussion section). Although the SOC and TN status changed among the different treatments after 7 years, the N 2 O fluxes did not respond significantly during the periods of zero fertilization in the studied soil.
However, the N 2 O fluxes increased greatly within 7 days during the 10 th leaf fertilization of summer maize on the third soil sampling date of 14 th August (Fig. 1a). The fluxes in the N fertilization treatments (N opt , CN opt , CM) were all significantly higher than in the N 0 control, increasing to 7.0, 4.3 and 6.1-times, respectively, but were not statistically significant among the fertilization treatments, possibly due to the high spatial variation of measured N 2 O fluxes within the field plots. High concentrations of NH 4 + , NO 2 − and NO 3 − were detected in all of the fertilization treatments and were significantly higher in the CM treatment. The high NH 4 + concentration in the CM treatment might result mainly from the high N mineralization due to the significantly higher SOC and TN contents in this treatment 37,38 (Table 1). A significantly high NH 4 + concentration in the CM treatment also induced high NO 2 − and NO 3 − concentrations in the soil matrix as a consequence of the typical nitrification chain. The above results indicate that the high N 2 O emission was mainly derived from the processes of urea hydrolysis to NH 4 + and then oxidation to NO 2 − and finally to NO 3 − . These processes were little affected by the distal  drivers of the background C and N status but rather by N fertilization events. We will further explain this critical, instant and strong emission period of high N 2 O peaks by bacterial amoA abundance in the next section. The annual N 2 O emissions were in the sequence CM > CN opt > N opt > N 0 and were significantly higher in the CN opt and CM treatments than in the N opt and N 0 treatments (Fig. 1b). These emissions were increasing to 2.6, 4.3 and 6.2 times in the N opt , CN opt , and CM treatments compared to the N 0 control. This increase corresponded well to the order of the total N applied rate ( Table 2) and followed the order of the SOC and TN status among treatments (Table 1). We further calculated the emission factor to investigate the N 2 O emissions per unit of N input in order to determine whether the manure and/or straw N interacted with chemical N to stimulate N 2 O emissions per unit of N input 39 .
The emission factors also increased in the sequence CM > CN opt > N opt , but there were no significant differences among these treatments, possibly also due to the high spatial variation of measured N 2 O fluxes in the field plots ( Fig. 1c). In our study the annual N 2 O emission factor was 0.20-0.40%, much lower than the IPCC default value of 1% 28 but was in line with our previous studies 18,20,23-25 . Abundance of the 16S rRNA gene and nitrification and denitrification genes. The order of the 16S rRNA gene copy number was CM > CN opt > N opt > N 0 on all three sampling dates (Fig. 3), a similar order to SOC, but only significantly higher in the CM treatment compared with the other three treatments. The 16S rRNA gene copy numbers in the CM treatment were 1.6-1.9-, 1.7-2.1-and 1.3-2.2-times larger than in the CN opt , N opt , and N 0 treatments on the three sampling dates, indicating that soil with long-term incorporated manure harbored the largest 16S rRNA gene number. Surprisingly, the 16S rRNA gene copies in all four treatments were higher on the 16 th April soil sampling date than on 9 th August or 14 th August, possibly due to the different crops because the first sampling date was winter wheat and the last two sampling dates were summer maize (further explanation in the Discussion section). The soil 16S rRNA gene copy numbers in the CN opt treatment appeared to be 27.2, 16.7 and 27.7% higher than in the N opt treatment on the three sampling dates, indicating that straw return increased the soil 16S rRNA gene copy number but this was not statistically significant.
The gene copy numbers of bacterial amoA in all three fertilization treatments were significantly higher than in the N 0 control but were not significantly different among fertilization treatments on all three sampling dates (Fig. 4), reflecting the legacy of historical long-term application of urea-based fertilizer and/or manure as distal drivers to regulate nitrification. These numbers were 1.4e + 7 to 1.9e + 7 and 1.4e + 7 to 1.6e + 7 on 16 th April and 9 th August, respectively. On the third day after the 10 th leaf fertilization, the amoA gene number of N opt , CN opt and CM increased 14.7, 124.5 and 107.6% compared to that before fertilization (on 9 th August), respectively, indicating that the 10 th leaf fertilization of maize enhanced the amoA gene copy number as proximal factor and led to production of high N 2 O emission peaks in this short period (Fig. 1a). This result further confirms our previous conclusion that ammonia oxidation is an engine to generate nitrous oxide in this intensively managed calcareous Fluvo-aquic soil 24 .    The abundance of the denitrification genes encoding nitrate reductase (narG), nitrite reductases (nirS and nirK) and N 2 O reductase (nosZ) followed a similar sequence, namely CM > CN opt > N opt > N 0 , on 16 th April and 9 th August, but in most cases these genes were significantly higher only in the CM treatment compared with the other three treatments (Fig. 5a,b,c,d), which seems to follow the trend of SOC and TN. The third sampling date on 14 th August, after fertilization, likely disturbed this trend, and none of the genes were significantly different among fertilization treatments. These results indicate that the abundance of denitrification genes was regulated by both distal drivers, such as SOC and TN, and proximal drivers, such as N fertilization. The gene copy numbers of narG, nirS and nirK in different treatments on 16 th April and 9 th August corresponded well with the daily N 2 O emissions during these two zero fertilization periods and with the annual N 2 O emissions (Fig. 1a,b). The gene copy numbers of narG, nirS and nirK in different treatments on 14 th August seem also to partly correspond with high N 2 O emission after fertilization. Although the nosZ gene number was significantly higher in the CM treatment in most cases, the N 2 O emission factor in this treatment was still higher than that of the other treatments, likely due to the effect of the nosZ gene being partly offset by increased effects of narG, nirS and nirK genes.
Sequencing analysis. The PCoA score plots based on the Bray-Curtis distance reveal that fertilization with straw and/or manure changed the structure of the soil microbiota (Fig. 6a) and there were statistically significant differences between the N 0 control and the fertilization treatments (Fig. 6b), i.e., N opt , CN opt and CM were separate  from N 0 . Among the fertilization treatments, N opt was separated from the CN opt treatment in the PCoA score plots, but this separation was not statistically significant; CM was significantly separated from the N opt and CN opt treatments (Fig. 6a,b). The bacterial community on 14 th August after fertilization was also separated from those on the first two sampling dates. These results further indicate that the soil microbial community was influenced by both distal drivers such as changes in the SOC and TN, and proximal drivers, such as N fertilization application.

Correlations between N 2 O emission and some soil chemical and biological parameters.
The annual N 2 O emissions were significantly correlated with SOC and TN, potential denitrification rate (PDNR), amoA gene number, denitrification functional gene numbers (narG, nirS, nirK, nosZ) and 16S rRNA gene number by Spearman's rank correlation analysis (Table 3) but were not significantly correlated with soil potential nitrification rate (PNR), likely due to all of the fertilization treatments having a high soil PNR (Table 1). Thus, soil PNR is not a limiting factor for controlling N 2 O production and emission. All of the other soil properties and abundance of functional genes were significantly correlated with each other except for amoA with SOC and TN and PNR with TN, which corresponds well with the above results showing that the amoA gene number did not respond to distal control (SOC and TN) but responded strongly to instant fertilization as a proximal control (Fig. 4).

Discussion
We investigated the bacterial amoA for nitrification in situ during both the zero-fertilization period and a fertilization event using a long-term field experiment with different fertilization regimes in an intensively managed low-carbon calcareous soil on the NCP. These results show that the long-term application of urea-based fertilizer and/or manure induced and enhanced bacterial amoA gene copies, which was related to a strong nitrification process, and consumed O 2 in the soil matrix then triggered denitrification as shown by our previous study 24 . The N 2 O emission induced by ammonia oxidation is also likely dependent on the heterotrophic respiration rate because    incorporation of straw and/or manure aggravated the emissions 23 . This suggests that the ammonia oxidizing bacteria will actively grow under temporally high NH 4 + concentrations, leading to microoxic or anoxic conditions, which in turn induce denitrification by heterotrophic denitrifiers or by nitrifiers, and the above chain processes lead to the high N 2 O emission peaks after N fertilization. The triggering of denitrification was further demonstrated by enhanced nirS and nirK genes in the current study. Previous studies in the same cropping rotation system showed that these peaks occurring instantly after N fertilization contributed up to 30-70% of the annual N 2 O emissions and were the key periods for controlling N 2 O emissions on the NCP 18,20,22,25,27,40 . This process may be slowed down using nitrification inhibitors and reducing field N 2 O emissions by up to 77% using DMPP 18 and 55% using the liquid nitrification inhibitor Piadin 22 . These emissions may be reduced by as much as 80-99% using DMPP or DCD in laboratory experiments 24,29 in calcareous fluvo-aquic soils from the NCP.
Our results provide molecular microbial evidence to illustrate that ammonia oxidation is an engine as the start for generating N 2 O in the study soil, which is why nitrification inhibitors are good for reducing N 2 O emissions 18,22,24,29 in this soil. The results of the bacterial amoA gene agree with other related studies under similar soil and climatic conditions 33,37,41,42 .
It has been reported that AOAs are ubiquitous in soils but do not respond to NH 4 + oxidization or N 2 O production in intensively managed agricultural soils 32,43 . Currently, the soils where AOAs have a significant impact on NH 4 + oxidization are mainly acidic 44 . One study conducted in our study region also revealed that Nitrosospira-like AOBs were dominant over AOAs in oxidizing NH 4 . Therefore, the contribution of AOAs to N 2 O production is small in our calcareous soil and this is why we did not quantify the amoA gene of archaea in our study.
Few studies have investigated the abundance of denitrification genes and their linkage with N 2 O production in our study soil, especially considering the distal and proximal control. We quantified the critical functional genes narG, nirS, nirK, and nosZ for denitrification and associated their abundances with annual N 2 O emissions. The significant correlations between annual N 2 O emissions and narG, nirS, and nirK gene numbers show that these genes abundances were related to production of N 2 O under some favorable conditions for denitrification; for example, rainfall or irrigation induce a low oxygen concentration in soil microsites [18][19][20][21][22][23][24][25][26][27] , which may be further supported by the significant correlation between annual N 2 O emissions and SOC and between narG, nirS, and nirK gene abundances and SOC. These findings further explain our previous results showing that the small pulses occurred when rainfall or irrigation events occurred in the same cropping system, especially during the warm and moist summer [18][19][20][21][22][23][24][25] . Therefore, the increased abundance of denitrification genes by long-term incorporated manure can produce N 2 O in our study soil and also partly contribute to the annual N 2 O emissions, which were enhanced by increased SOC and TN. A similar study was conducted using soils sampled in Broadbalk Wheat and the "Broadbalk Wilderness" long-term experiments 45 but using the incubation technique labeled with KNO 3 solution to investigate the influence of different long-term fertilization and cultivation treatments on denitrifier communities and to produce N 2 O. The results show that bacteria containing nirK were most likely responsible for the increased denitrification potential associated with high SOC and TN, which generally agrees with our findings.
Why was there nitrite accumulation on 9 th August (Fig. 2b) when the soil ammonium concentration was quite low (Fig. 2a)? We suspected that it might be due to the reduction of nitrate in more anaerobic microsites together with high levels of dissoluble organic carbon induced by fast root metabolism during the period of strong summer maize growth at higher soil temperatures (approximately 24 °C) and water content (approximately 16%) 46,47 . Thus, this nitrite accumulation could contribute to the annual N 2 O emissions. This hypothesis is supported by a recent study that demonstrated the significant contribution of nitrite to N 2 O emissions in maize-cropped soil 29,48 .
The explanation for the lower 16S rRNA gene copy number on the two sampling dates of maize (9 th and 14 th August) than on the first sampling date of wheat (16 th April) (Fig. 3) was likely due to competition for resources between crop roots and soil microbes 49 . Summer maize roots might exhaust their resources during fast growth and suppress the growth of microbes, which may result in a lower 16S rRNA gene copy number in the summer maize season. Few studies have reported this phenomenon and it merits further study.
Our study highlights the linkage of instant high N 2 O emission peaks with the abundance of the bacterial amoA gene for nitrification; annual N 2 O emissions and a small N 2 O pulse after rainfall or irrigation with the abundance of denitrification genes, providing insight into the mechanism of N 2 O production and the factors controlled by distal and proximal drivers in this intensively managed calcareous fluvo-aquic soil (Fig. 7). It is critical to suppress the growth of bacteria containing the amoA gene instantly after urea-based fertilization to mitigate N 2 O emission in these strong nitrification soils. We also need to be concerned about the enhanced abundances of denitrification functional genes under favorable conditions.

Methods
Site description and soil sampling. The soils on the NCP are derived from alluvial loess transported by the Yellow River and its tributaries and most of the soils are calcareous Fluvisols or Cambisols (FAO Soil Classification System) with a silt texture and relatively uniform profile characterized by high mineralization, high nitrification and low denitrification rates 17,38 . The calcareous soils have a pH of 7.5-8.5 and an organic matter content of approximately 10-15 g kg −1 . Annual cumulative mean temperature for days with temperatures above 10 °C is 4000-5000 °C and the annual precipitation is 500-700 mm with 60-70% of the rainfall occurring during the summer (July-September) 23 . The long-term field experiment began in October 2006 and is located at Shangzhuang Research Station (39°48′ N, 116°28′ E) of China Agricultural University in suburban Beijing. The expression "long-term" in this paper refers to fertilization treatments compared to the no fertilizer control after a period of seven years.
The cropping system was a winter wheat-summer maize rotation. The top 20 cm of the soil profile, sampled at the beginning of the field experiment, had 28% clay, 32% silt, and 40% sand contents and a pH of 8.1 (soil: water ratio 1:2.5). The bulk density was 1.31 g cm −3 and the nutrient contents were as follows: SOC 7.1 g kg −1 , TN 0.8 g kg −1 , NO 3 -N 24.5 mg kg −1 , NH 4 -N 1.20 mg kg −1 , Olsen P 7.8 mg kg −1 and available K 76.2 mg kg −1 . The soil is a typical calcareous fluvo-aquic soil widely distributed across the NCP.
Eight treatments were set up as described in a previous paper 50 . Our current study sampled from four treatments, i.e., N 0 , N opt , CN opt and CM, in the 2012-2013 winter wheat-summer maize rotation based on the significant difference of some soil chemical and biological properties at 0-20 cm soil depth sampled before the sowing of winter wheat in 2012 (Table 1). N 0 is no N application, wheat and maize straw removed; N opt and CN opt are chemical fertilizer N application according to the improved N min (NO 3 − -N + NH 4 + -N) test, wheat and maize straw removed or returned, respectively; CM is cattle manure supplementary applied N based on the N balance calculation, wheat and maize straw returned. The design is a completely randomized block with three replicates and each plot area is 64 m 2 (8 × 8 m). Winter wheat was sown at the beginning of October and harvested in the middle of the following June and summer maize was subsequently sown and harvested at the end of September. The chemical N fertilizer used was urea. Straw of both maize and wheat was chopped mechanically into 5-8 cm lengths. Chemical fertilizers and chopped maize straw were incorporated into the soil with tillage at the beginning of October before wheat was sown, and the wheat straw was mulched on the soil surface after the wheat harvest.
In the 2012-2013 winter wheat-summer maize rotation basal fertilizer was applied and winter wheat was sown on 2 October 2012 (Table 2), and winter irrigation (60 mm) was carried out on 17 November 2012. We collected the first soil samples (0-20 cm) at the jointing stage of winter wheat on 16 April 2013 to investigate the effects of long-term treatments on the soil bacterial community composition and the abundance of functional genes. After this sampling, jointing fertilizer was applied on 21 April 2013, followed by 60 mm of irrigation; irrigation (60 mm) was carried out on 14 th May 2013 at the heading stage of winter wheat. Winter wheat was harvested on 21 st June 2013, summer maize was sown immediately, and the 4 th leaf fertilizer of maize was applied on 20 th July 2013. The second and third soil samples (0-20 cm) were collected before one day (9 th August) and after three days (14 th August), respectively, when the 10 th leaf fertilization of maize was performed on 11 th August 2013 to investigate the effects of N fertilization on N 2 O emissions, the corresponding bacterial community composition and the abundance of functional genes.
Five soil cores were taken from each plot and were mixed to form one composite sample; the samples were stored in an ice box before returning to the laboratory within half a day. The samples were divided into two parts. One part of the fresh soil was used to measure the chemical properties, including ammonium, nitrite, nitrate, water contents and pH, immediately after arriving at the laboratory. The remainder of each sample was stored at − 80 °C for subsequent DNA extraction and downstream analysis, including the quantitative real-time PCR (Q-PCR) of the 16 S rRNA gene, functional genes of amoA gene of bacterial, nitrate reductase gene (narG), nitrite reductase genes (nirS and nirK), and N 2 O reductase gene (nosZ), by high-throughput sequencing of 16S rRNA based on an Illumina platform analysis (Illumina Inc., San Diego, CA). N 2 O emission measurements. N 2 O emissions were measured using the closed static chamber method as detailed in Huang et al. 23 . They were measured on days of first two soil sampling dates, i.e., on 16 th April and 9 th August. For third soil sampling day on 14 th August, daily measurements were carried out for 7 days after the 10 th leaf fertilization on 11 th August in order to cover the entire N 2 O peaking period during this N fertilization event. During the whole crop rotation, daily measurements were also carried out for 7 days after each fertilization event and 5 days after each rainfall or irrigation event; for the remaining periods the emissions were measured twice per week and once per week when the soil was frozen 23 .
In the current study N 2 O emissions were calculated daily as μ g N 2 O-N m −2 d −1 on 16 th April and 9 th August (Fig. 1a), but the N 2 O emissions on 14 th August were calculated as the average emissions over 7 days (peak lasting period) after N fertilization in order to reflect the total N 2 O emissions induced by N fertilization. The annual N 2 O emissions (Fig. 1b) were calculated from the sum of measurement days and zero-measurement days (estimated by linear interpolation) 23 . The annual N 2 O emission factor was calculated by subtracting the annual N 2 O emissions of the control treatment from those of the fertilization treatment and then divided by the total N application rate, including straw N and manure N (Fig. 1c), according to IPCC methodology.
Determination of soil properties. Mineral N was extracted using 1 M KCl solution at a soil:water ratio of 1:5 (w/v) and determined using a continuous flow analyzer (AA3, Seal Analytical, Norderstedt, Germany). Soil nitrite was measured as described by Stevens and Laughlin 51 , the soil nitrification and denitrification potential were determined according to Hart et al. 52 and Tiedje et al. 53 respectively. Details can be found in the Supplementary Information (SI).
Soil molecular analysis. DNA extraction. DNA was extracted from frozen soil using a method based on the CTAB (hexadecyl trimethyl ammonium bromide) method 54 with some modification, details are described in the SI.
Illumina-based 16S rRNA gene sequencing. The gene-specific sequences that were used targeted the 16S rRNA gene V3 and V4 regions. The primers that were used for the Amplicon PCR are listed in Table S2. The target fragment was approximately 550 bp. More details can be found in the SI. The 16S rRNA gene sequence information in this study has been submitted to the NCBI Sequence Read Archive (SRA) database under accession number SRP083579.
Statistical analysis. The data were computed using Microsoft 2010, and the results are reported as the means (± standard error), and the figures were created using SigmaPlot v. 12.5 and Visio 2013. Differences in the soil properties and functional genes were analyzed by one-way analysis of variance and compared by Duncan's multiple range test at the 5% level using IBM SPSS Statistics 20. We calculated the functional gene numbers considering the effects of treatment and neglected sampling time to assess the significance of correlations between N 2 O emissions and related soil parameters. A non-parametric analysis was used to generate a Spearman's rank correlation matrix using IBM SPSS Statistics 20. The method of dealing with sequencing data is described in the SI.