Introduction

Sustainable agriculture for an increasing global population requires a balance between optimizing fertilizer inputs and minimizing adverse outcomes whilst increasing food production. Emissions of the greenhouse gas nitrous oxide (N2O) from agricultural soils due to bacterial denitrification make a significant contribution to global warming1. When soils are wet and O2 availability is limited but there is sufficient organic matter, a common condition in temperate agricultural soils, nitrate (NO3−) can act as an alternative electron acceptor and is reduced to N2O by many different groups of facultatively anaerobic bacteria2,3. This occurs with different types of fertilizer, whether organic, biological (e.g. legumes), urea or inorganic ammonia/nitrate compounds4. Some bacteria can further reduce N2O to N2 which is not environmentally harmful but nevertheless reduces nitrogen use efficiency, wasting N fertilizer inputs. In a survey of genome-sequenced prokaryotes, 7% contained denitrification genes exemplified by the nitrite reductase gene nirK, the alternate gene nirS or the gene encoding nitrous oxide reductase, nosZ5. However, not all prokaryotes with a nitrite reductase gene carry nosZ and in others the gene is not active. Consequently, for these organisms, the final denitrification step is the release of N2O. Also, there are also a substantial number of bacteria that contain nosZ without nirK or nirS and are assumed to sequester and reduce N2O, emitting N26. Many soil fungi can also denitrify: their contribution is minor when compared to that of bacteria in agricultural systems and arable soil but increases post-harvest and after incorporation of organic residues7. Fungi form only a small proportion of the soil microbiome under growing crops, and fungal denitrifiers contain nirK but not nosZ thus do not reduce any N2O they generate to N28. N2O generated by other processes including nitrification is at relatively low levels compared to the activity of denitrifying bacteria in anoxic, fertilized agricultural soils9,10.

The alternate nitrous oxide reductase genes have been known for many years but the significance of carrying either nirK (encoding a Cu-containing enzyme) or nirS (encoding a cytochrome cd1) remains unclear11. The two forms can occur within different individual species of the same genus but (with rare exceptions) they are mutually exclusive12. Recent stable isotope probing experiments to identify denitrification genes in soil indicated that nirK was the most abundant and active, with greater sequence diversity than previously recognised13. The nosZ genes are also a diverse group and have been divided into two clades: nosZI in the Proteobacteria; nosZII in the Gemmatimodates, Bacteroidetes and Choroflexi14.

Whilst there have been many surveys of N2O fluxes from agricultural soils, some using intact or repacked soil cores under lab conditions, others made in the field, few have specifically investigated the role of crop plants or measured N2 emissions. Most are concerned with soil organic carbon (SOC), nitrogen and N-fertilizer forms and rates of application, soil moisture and temperature1,15,16. There is considerable information on the roles of these drivers but the effects of interaction between plant and soil factors including the microbiome on denitrification is less clear. A compilation of field data concluded that the presence of plants increased N2O emissions and also the ratio of N2:N2O17. Experiments in air-free systems with Argon or Helium replacing N2 have indicated that arable soils with young plants evolve more N2O and N2 than controls with no plants7,18. However, plants were also implicated in reduced soil moisture and less N fertilizer remaining in soil, leading to lower denitrification19. A review of 26 separate studies investigating cover crops in the field showed that 40% resulted in lower and 60% in higher emissions compared to control soils without cover crops20. Presence or absence of legumes, high soil N, rainfall and crop residues were all implicated in these differences. However, a more recent meta-analysis of 129 publications indicated that the presence of cover crops significantly reduced N2O emissions compared to bare soil21. In an experiment using soil cores with artificial root exudates, no N2O was detected in the control soil in contrast to that receiving exudates22 whereas a field experiment comparing bare soil (chemical fallow) with non-legume cover crops found no significant difference in N2O emissions23. These contradictory results concerning N2O and N2 emissions from agricultural soil in the presence or absence of plants reinforce the need for a deeper understanding of the processes involved.

In this paper, we report on a glasshouse experiment designed to measure production of N2O and N2 in soils taken from plots with contrasting long-term treatments which were anticipated to alter microbial communities with potential consequences for the relative abundance of denitrification genes. Furthermore, we aimed to examine the impact of growing plants and of applying KNO3 fertilizer on gas production. The soils were taken from the long-term Broadbalk Winter Wheat experiment at Rothamsted Research (UK), where different fertilizers and N-rates have been applied, resulting in soils with distinctly different edaphic properties. Using conditions designed to promote denitrification, soils from different long-term treatments, in pots with or without wheat plants, were treated with 15N-labelled fertilizer, control soils receiving no fertilizer and sealed in chambers with ambient air. The plants had been grown to the stem elongation stage where root exudation is at a maximum24 and the roots had colonised the pots so all soil was “rhizosphere”, for comparison with unplanted controls. The concentrations of N2 and N2O in these sealed chambers were measured daily over 4 days to provide a snapshot of denitrification activity in the rhizosphere and the control bulk soil; soil DNA and mRNA were extracted at this point to assess the abundance and activity of denitrification genes. From this we show the relative importance of antecedent soil treatments that influence edaphic factors including the soil microbiome, crop plants and N-fertilizer in promoting denitrification.

Material and methods

Soils

In the Broadbalk Experiment, running since 1843 at Rothamsted Research (UK), different fertilizers and N-rates have been applied consistently to winter wheat. In 1882, one section (previously referred to as “wilderness”) was taken out of cultivation and left to regenerate to woodland25,26. We chose plots that received farmyard manure (FYM) at 35 t ha−1, no N fertilizer (N0) or 288 kg N ha−1 annually (N6), together with the woodland soil (Wood). Table 1 shows the soil properties and annual fertilizer applications of the four arable treatments and the plot converted to woodland.

Table 1 Fertilizer applications and soil properties for Broadbalk plots.
Full size table

Broadbalk is not fully replicated since it was set up before the advent of modern statistical design, hence each plot to be sampled was subdivided into three equally sized pseudoreplicates, each 9.3 m × 6 m in the arable plots and 9.3 m × 4 m in the FYM plot. The woodland area (80 m × 15 m) was also subdivided into equally sized pseudoreplicates. Ten soil cores (top 20 cm) were collected and pooled from each area in May 2014, generating three replicate soil samples for each of the four plots. Soil was sieved to 4 mm and stored at 4 °C for 14 days before transferring to pots; weighed subsamples were dried overnight at 80 °C to estimate the soil dry weight (dw).

Plants and pots

A total of 48 pots were set up: three soil replicates from each of the 4 plots, with four treatments (+ /− wheat; + /− N fertilizer). Each Ø 10 cm pot contained 350 ml vol. soil; 24 pots were planted with 4 pre-germinated wheat seeds (Triticum aestivum cv Cadenza) per pot, grown 6 weeks and given foliar feed if required (diagnosed by slight yellowing of leaves) and watered as normal with tap water. At 6 weeks post germination, wheat root exudation is at a maximum24; previous experiments showed wheat roots to fill the pots at this stage. The 24 unplanted pots were set up and kept in glasshouse alongside those with wheat in a randomised block design and were watered at the same time as the plants to maintain soil moisture. Soil bulk and particle density had been measured previously in the source soils27 and was used to calculate the pore space. This enabled subsequent manipulation of the water-filled pore space (wfps).

Chambers

When plants reached stem elongation but pre-flowering (c. 6 weeks post germination), the soils were adjusted to 95–100% wfps to create the anaerobic conditions conducive to denitrification, by adding sterile deionised water (sd H2O), or a solution of K15NO3 (99% atom enrichment) in sd H2O as to give equivalent to 100 kg N ha−1 (i.e. 40 µg N g−1 dw soil) in each pot. Immediately after adding water or fertiliser, a 2 L polycarbonate bottle with the base removed was placed over each pot with a SubaSeal at the top, as shown in Supplementary Fig. S1, creating closed chambers containing air for the 4 day sampling period.

Gas sampling and measurement of N2 and N2O

The first samples (day 0) were taken immediately after sealing the chambers following the addition of K15NO3 fertilizer or water. Subsequently, the chambers were sampled at 24 h intervals for the next 3 days (day 1, day 2, day 3, day 4), with two ambient air samples taken at each sampling time. A single 22.5 ml air sample was taken at each timepoint for N2O analysis; for 15N isotope ratio sampling, a second 12 ml air sample was taken from each treatment fertilized with K15NO3 to measure 15N2-N and 15N atom% in N2O. The total (chamber plus pore space) volume and the soil dry weight in each pot was used to calculate N2O-N and 15N2-N g−1 dw soil for each replicate.

The N2O was measured using gas chromatography with an electron capture detector and an automated sample injection system; a TG2 trace gas analyser (Europa Scientific, now Sercon, Crewe, UK) interfaced to a Sercon 20–22 isotope ratio mass spectrometer (IRMS) was used to measure 15N enrichment of N2. Solutions of 6 and 30 atom% ammonium sulphate ((NH4)2SO4) were prepared and used to generate 6 and 30 atom% N2O28 and used as reference and quality control standards. Atmospheric air, with natural abundance 15N (0.3663 atom%) was used as the reference for N2. The N2 concentration in each sample was calculated from the ratios of the intensity of ion beams at mass to charge ratios 28, 29 and 30, using the equations of Stevens and Laughlin29,30.

DNA and RNA extraction and amplicon sequencing

After 4 days, the chambers were removed. Soil samples were processed within 5 min to preserve the integrity of the RNA and DNA. In pots with wheat, the roots had spread throughout the soil and soil adhered to the roots at sampling. This soil was shaken off and mixed, similarly for unplanted pots, soil was mixed, a subsample taken, sieved to 2 mm and frozen in liquid N2 for subsequent DNA and RNA extraction using the RNA PowerSoil® isolation kit and RNA PowerSoil® DNA Elution Accessory Kit (MO BIO Laboratories, Inc) following manufacturer’s instructions31. RNA samples were DNAse treated to remove DNA contamination using the DNase Max Kit (Qiagen, Manchester, UK), following the manufacturer’s protocol. Direct PCRs and gel electrophoresis were carried out on DNAse treated RNA to confirm all contaminating DNA had been removed. The quantity and quality of extracted DNA and DNAse-treated RNA were analysed by fluorometer Qubit® 2.0 dsDNA and RNA BR Assay Kits and Nanodrop microvolume spectrophotometer (Thermo Fisher Scientific).

The bacterial and archaeal diversity was determined from the 12 samples taken at the time of field sampling by amplicon sequencing of 16S rRNA genes using the primers 515F/806R, sequenced on Illumina’s MiSeq platform and analysed to phylum (sub-phylum for Proteobacteria) level and to OTU (97% sequence identity) using the QIIME 1.8 pipeline. The method was described in detail previously32. The amplicon sequence data for this study have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB36852 (https://www.ebi.ac.uk/ena/browser/view/PRJEB36852).

Quantification of bacterial and denitrification gene abundance

The primers used to detect the 16S rRNA genes, the alternative genes for nitrite reductase (nirK and nirS) and the nitrous oxide reductase gene nosZ clades I and II are shown in supplementary Table S1. Quantitative-PCR (qPCR) amplifications were performed in 10 µl volumes containing 5 µl QuantiFast SYBR Green PCR Master Mix for DNA and QuantiFast SYBR Green RT-PCR Master Mix for RNA (Qiagen, Manchester, UK), 0.1 µl of each primer (1 µM), 0.1 µl of QuantiFast RT Mix for RT-qPCR, 2 µl of template DNA at 5 ng µl−1 or 2–4 µl of RNA at 10 ng µl−1 and nuclease-free water (Severn Biotech, Kidderminster, UK) up to 10 µl, using a CFX384 Touch™ Real-Time PCR Detection System (Bio-Rad, Hemel Hempstead, UK). The standards for each molecular target were obtained using a tenfold serial dilution of PCR products amplified from an environmental reference DNA and purified by gel extraction using the Wizard® SV Gel and PCR Clean Up System (Promega, Southampton, UK) following the manufacturer’s instruction then quantified by fluorometer Qubit® 2.0 dsDNA BR Assay Kit (Thermo Fisher Scientific). Standard curve template DNA and the negative/positive controls were amplified in triplicate. Amplification conditions for all qPCR assays consisted of an initial denaturation at 95 °C for 5 min followed by 40 (two step) cycles; 95 °C for 10 s and 60 °C for 30 s. The RT-qPCR program had an initial reverse-transcription step at 50 °C for 10 min. Each amplification was followed by melt curve analysis (60 °C to 95 °C, with incremental readings every 0.5 °C) to confirm the specificity of each assay. Efficiency of amplification for each primer set was > 82% with r2 ≥ 0.996 (Supplementary Table S2). Results are expressed as gene copies g−1 dw soil.

Soil properties

At the end of the pot trial, samples from each pot were oven-dried as described above, to measure moisture and estimate the final wfps. Extracts (1:5 soil dw:2 M KCl, shaken for 2 h at 300 rpm, 20 °C) were analysed for nitrite (NO2−), nitrate (NO3−) and ammonia (NH4+) using a Skalar colorimetric continuous flow Analyzer. Results are given as μg g−1 dw soil.

Statistical analyses

GenStat 17th Edition (VSN International Ltd, Hemel Hempstead, UK) was used to perform ANOVA with soil origin, sampling date (i.e. time of incubation), presence or absence of plant and addition or not of fertilizer as factors, to compare values obtained from soil and gas analyses and from qPCR estimations of gene and transcript copy numbers. To check that each set of measured values met the assumptions of ANOVA and were normally distributed, residuals were plotted. If they did not show normal distribution, data was log-transformed and again checked for normal distribution of residuals. Where ANOVA results were significantly different (P < 0.05), means were further tested using Tukey’s post-hoc method in the GenStat multiple comparison menu with 95% confidence. Where appropriate, standard errors of difference of means (s.e.d.) are indicated. When only two treatments were compared, Student’s t-test in the Excel data package was used. The statistics package PAST v. 3.1633 was used to analyse the relative abundance of 16S rRNA amplicons using: SIMPER to determine the percentage contribution of each phylum to each treatment; non-metric multidimensional scaling (NMDS) analysis at the OTUs level based on the Bray–Curtis similarity index; PERMANOVA to assess the significance of the NMDS plot and Spearman’s rank correlations of gene abundance, soil properties and gas emissions. Unless otherwise indicated, statistically significant differences are assumed to occur at P ≤ 0.05 and are referred to as “significant” throughout the text; results with no significant differences are referred to as NSD.

Results

Soils and their microbial communities

The soil properties shown in Table 1 indicate variation in soil texture across the Broadbalk field, with less clay in the N0 and FYM plots, situated on the north side of Broadbalk field compared to N6 and woodland towards the south side. The soil pH ranged from 7.1 to 8.2, lowest in the mineral-nitrogen fertilized soil N6 and highest in the N0 soil that received no N fertilizer. The bulk density of woodland soil is much lower and the % SOC much higher compared to other soils; the FYM soil has lower bulk density and higher % SOC than the other arable soils. The ratio of SOC:total N was approximately 10:1 in the arable soils and 13:1 in the woodland soil.

The community structure of bacteria and archaea revealed by 16S rRNA amplicon sequencing of metagenomic DNA extracted from the soil samples, at collection from the field, shows significant differences, and distinct separation on a NMDS plot (Fig. 1). Of the 14 phyla (sub-phyla for the Proteobacteria) comprising > 0.1% of the community in at least one of the soils, only the δ-Proteobacteria did not show significantly different (P ≤ 0.05) mean abundance in at least one soil, according to ANOVA (Fig. 1). For example, the woodland soil has more α-Proteobacteria and Verrucomicrobia but fewer Thaumarchaeota (archaea) and β-Proteobacteria than the other treatments. Both the FYM and woodland soil have more γ-Proteobacteria and fewer Gemmatimonadetes; the FYM soil has more Firmicutes than the other soils (Fig. 1).

Figure 1
figure 1

Relative mean abundance of prokaryotic phyla/sub-phyla in soils of origin on collection from the field. Phyla with at least 0.1% of the total community present in at least one soil treatment are included. Proteobacteria sub-phyla: a = alpha, b = beta, d = delta, g = gamma; s.e.d. for each group is shown; letters indicate mean significantly different means within each group (P = 0.05, according to Tukey’s post-hoc test on ANOVA). Insert top right shows NMDS plot of OTU for prokaryotic communities – PERMANOVA F = 9.477, P (same) = 0.0001.

Full size image

16S rRNA and denitrification gene abundance

At the end of the experiment, DNA was extracted and amplified from all samples but sufficient RNA for further analysis was obtained only from the FYM and woodland soils which contained more organic matter and larger microbial communities. ANOVA comparing the abundance for each set of genes and transcripts measured using qPCR showed that the soil of origin had a significant influence in all cases (Table 2). However, other factors (presence/absence of wheat and addition or not of N-fertilizer) and interactions between them were not significant, except for nosZI which was significantly influenced by the plant. Bacterial abundance indicated by 16S rRNA gene copy number was 2 × 109 g−1 soil in the N0 and N6 soils and significantly higher in the FYM and woodland soil, 5 × 109 and 7 × 109 copies g−1, respectively (Fig. 2). This pattern of relative abundance was seen for nirK (7 × 108–4 × 109 copies), nosZI (5 × 107–2 × 108 copies) and nosZII (4 × 106–1 × 107 copies g−1 soil). The exception was nirS where N0, N6 and woodland soil had similar gene abundance (1 × 107 copies g−1 soil) and FYM significantly more with 4 × 107 copies g−1soil (Fig. 2). The ratio nirK:nirS gene copies in the woodland soil was 300:1, significantly more than the mean of 55:1 in the arable soils (F3,32 = 102.63, P < 0.001). Woodland also had a significantly higher ratio of nir:nosZ genes, 20:1 compared to 13:1 in the arable soils (F3,32 = 10.97, P < 0.001). This was influenced only by the origin of the soil, not the plant or fertilizer treatment (supplementary Fig. S2, Supplementary Table S3).

Table 2 ANOVA for soil edaphic factors and gene abundances.
Full size table
Figure 2
figure 2

Gene abundance from qPCR at the end of the experiment, pooling all treatments for each soil of origin (n = 12); letters denote significantly different values within each set of genes (P = 0.05) according to Tukey’s post-hoc test in ANOVA; s.e.d. = standard errors of difference of means; note that 16S and nirK are plotted as 10–9, the other genes as 10–6 copies g−1 dw soil.

Full size image

The mean number of 16S rRNA transcripts overall in the woodland soil, 5.3 × 108 copies g−1, was significantly higher (t22 = 2.23, P = 0.02) than the FYM soil with 1.8 × 108 copies g−1. The same pattern was seen with nirK transcripts: 2.4 × 105 copies in the woodland; 1.2 × 105 copies in the FYM soil (t22 = 3.75, P < 0.001). There was no significant difference between the two soils for nosZI transcripts which were much less abundant, mean 6 × 103 copies g−1 soil.

Soil properties at the end of the experiment

The concentration of soil NO2−-N at the end of the experiment was below the limit of detection in most samples and is not included. The NH4+-N followed the same trend as the % N and bacterial abundance, significantly higher in the woodland soil (Fig. 3a). According to ANOVA, it was influenced by the presence of plants but not K15NO3-fertilizer additions (Table 2). This was confirmed using t-tests: the mean NH4+-N concentration for all soils with plants was 2.6 μg g−1 soil, significantly higher (t46 = 2.6, P = 0.007) than 1.6 μg g−1 for bare soils. The NH4+-N is around tenfold less than the NO3−-N in unfertilized soils, indicating nitrifier activity in the aerobic soils prior to setting up the chambers whereby soil pore saturation to create anaerobic conditions is predicted to reduce nitrification rates.

Figure 3
figure 3

Soil properties at the end of the experiment. (a) concentration of NH4+-N; (b) NO3–-N; (c) % wfps; means for soils with all treatments (n = 3); different letters denote significantly different values according to Tukey’s post-hoc test in ANOVA (P = 0.05), s.e.d. = standard errors of difference of means for all samples.

Full size image

ANOVA showed the NO3–-N concentration to vary significantly between soils and to be influenced by both K15NO3 fertilizer additions and the presence of plants (Table 2). The NO3–-N concentration remaining in soils receiving no K15NO3 was significantly higher for woodland and N6 than for the N0 and FYM soils (Fig. 3b) and the overall mean in bare soil was 16.6 μg g−1, significantly higher (t22 = 3.0, P = 0.006) than in the presence of plants (0.36 μg g−1 soil). Where K15NO3 was applied, differences due to soil of origin was not significant (Fig. 3b) but the mean value for unplanted soils was 49.9 μg g−1, significantly higher (t22 = 6.0, P > 0.001) than where plants were present (5.4 μg g−1).

The % water-filled pore space (wfps), set at an estimated 95% at the start of the experiment, had fallen to 60–80% in most soils by the end, and to 40% for the woodland soils with wheat (Fig. 3c, Table 2). Water had drained from the pot into the tray and had also been redistributed around the sides of the chambers as condensation; plants but not K15NO3 fertilizer addition had a significant influence (Table 2). The overall mean wfps in all bare soil soils 72.6% was significantly higher (t46 = 3.3, P < 0.001) than 61.0% for all planted soils.

Gas production

Gas measurements made immediately after adding K15NO3 fertilizer or water and sealing the chambers (day 0) were similar to ambient values and were not included in subsequent analyses (e.g. mean N2O-N from 48 chambers was 0.27 ppm, s.e.d. 0.0035; ambient N2O in 10 glasshouse air samples was 0.28 ppm, s.e.d. 0.0048). Subsequent samples were taken at 24 h intervals (day 1–4) until the experiment concluded, the chambers were dismantled, and the soil was sampled. ANOVA indicated that the presence/absence of plants and K15NO3 had a significant effect on N2O-N but not the sampling date either alone or in combination with the other factors; in contrast, CO2 levels were additionally influenced by sampling date (Table 3). For this reason, each day was treated as a repeat sampling for N2O (Fig. 4a). In unfertilized soil, mean N2O-N from unplanted soil was 5.2 ng g−1, significantly higher (t22 = 3.0, P = 0.003) than with plants (1.1 ng g−1 soil). However, where K15NO3 was applied, the mean N2O-N was NSD in bare and planted soil. Over all treatments, the N2O-N measurements were highly variable with NSD between most means and overall differences due to the soil of origin were also NSD (Table 3.) The exception was significantly higher N2O-N in woodland compared to FYM soil where K15NO3 was applied and no plants were growing (Fig. 4a).

Table 3 ANOVA for gaseous losses from all soils and treatments.
Full size table
Figure 4
figure 4

Mean gas production over 4 days. (a) N2O-N, all treatments (48 pots); (b–e) K15NO3-fertilized treatments only (24 pots). (b) 15N atom % measured in N2O; (c) N2O-N indicating %15N (upper s.e.d. relates to N2O-N; lower bar relates to 15N atom %), (d) N2-N, (e) N2O-N as % (N2-N + N2O-N). Different letters denote significantly different values according to Tukey’s post-hoc test in ANOVA (P = 0.05), s.e.d. = standard errors of difference of means for all samples.

Full size image

The mean CO2 measured was 1 µg g−1 dw soil for all soils whether plants were growing or not, with or without K15NO3 application. The exception was bare woodland soil, with a mean of 25 mg g−1 dw soil (Supplementary Fig. S3). This indicates a similar rate of production and consumption by soil and plants for soils taken from the long term arable treatments, with only the woodland soil with high SOC and microbial biomass producing significantly more CO2 than the system could consume, peaking at 48 h (Supplementary Fig. S4).

Measurement of 15N was only possible in the K15NO3-fertilized plots. The proportion present in N2O, indicated by the 15N atom% (Fig. 4b), varied significantly between soils of origin. The sampling date and presence/absence of plants did not have a significant influence according to ANOVA (Table 4) but the mean value for all times and samples in bare soil, 11.7% was significantly (t94 = 2.0, P = 0.02) less than the mean value where plants were present (16.8%). The data was used together with N2O-N measurements to show that the proportion of 15N in N2O-N was significantly less from the woodland soil when wheat was present (Fig. 4c). This could be because the more open pore structure of the woodland soil (demonstrated by the drop in wfps at the end of the experiment) resulted in better root growth and proportionally greater uptake of the 15N-labelled fertilizer by the wheat.

Table 4 ANOVA for gas measurements in K15NO3-fertilized soils only.
Full size table

The N2-N measured in the K15NO3-fertilized plots showed that the date of sampling as well as soil of origin and plant had significant influences (Table 4). The presence of plants appeared to increase N2 production over time (supplementary Fig. 5), in contrast with total N2O-N production which showed no significant changes. The overall mean for all soils and times with plants was 533 ng g−1 N2-N, significantly more (t47 = 2.8, P = 0.004) than without plants (239 ng g−1 N2-N). In the absence of plants there was NSD between the mean N2-N in the different soils but when plants were present, N6 produced significantly more N2 than the FYM and woodland soils (Fig. 4d). The total gaseous N (15N2-N + N2O-N) was dominated by N2-N with an overall mean of 390 ng g−1 soil compared to 4 ng g−1 for N2O-N. However, the relative abundance showed a significantly higher % N2O-N in bare N6 and woodland soils compared to those with wheat (Fig. 4e) and the presence of plants had an overall significant effect according to ANOVA (Table 4).

Relating edaphic and microbiological factors with gas production

To investigate factors influencing gaseous emissions, Spearman’s rank correlation was derived for gas, soil and microbial parameters for all 48 pots where N2O was measured (supplementary Table S4A) and the 24 pots where K15NO3 fertilizer was added and 15N2-N and 15N atom% in N2O-N was measured (Supplementary Table S4B). Mean values of gas concentrations from all four sampling times were used. Overall, N2O-N was moderately correlated to wfps and strongly correlated to NO3–-N, factors both known to support denitrification. However, in the fertilized soils the NO3–-N relationship was not apparent, presumably because the relatively high NO3–-N was not limiting. The N2O-N from fertilized treatments was positively correlated moderately with N2-N and highly with the 15N atom% in N2O-N. N2O-N as a % of (N2-N + N2O-N) and 15N atom% showed moderate negative correlation with NH4+-N (Supplementary Table S4b). There was no correlation between either N2O-N or N2-N and the total bacterial community indicated by 16S rRNA gene numbers, or with any of the denitrification genes although there was a moderate negative correlation between N2O-N and the nir:nos ratio in the subset of K15NO3-fertilized soils. Abundance of all the denitrification genes was strongly correlated with 16S rRNA indicating the relationship between SOC and microbial abundance, except for nirS. This relationship was supported further by a strong positive correlation of the 16S rRNA and denitrification genes with NH4+-N, derived from the mineralization of soil organic matter. There was no significant correlation between nirS and nirK although both correlated with nosZI and nosZII abundance. The ratio nirK:nirS showed weak negative correlation with the soil wfps, strong positive correlation with NH4+-N and moderate correlation to CO2 (positive) and, in the 15N-fertilized pots, 15N atom% (negative). The nir:nos ratio showed similar trends and additionally showed moderate negative correlation to N2O-N and, in the 15N-fertilized fertilized pots, to N2O-N as a % of (N2-N + N2O-N). Neither ratio showed significant correlation with N2-N. There were no statistically significant correlations between nosZI and nirK mRNA abundance and gaseous emissions.

The CO2 emissions (prominent only in woodland soil without plants) correlated with NO3−-N; 16S rRNA, nirK and nosZI gene abundance; and with N2-N in the K15NO3 fertilized subset.

Discussion

The role of microbial diversity and abundance

Microbial diversity was influenced by previous long-term treatment of the different plots, resulting in the distinctly different microbiome composition revealed by amplicon sequencing. These contrasting treatments: arable management with N-fertilizer (N6); no N (N0); organic FYM; and the woodland soil left untilled and unamended, provide very different scenarios and each appears to favour divergent combinations of phyla and subphyla as shown in Fig. 1. The overall size of the microbial biomass (also related to SOC) indicated by 16S rRNA gene abundance was related to the abundance of the denitrification genes nirK and nosZI, which did not appear to be increased by the regular application of N fertilizer. The denitrification genes are relatively common across different phyla, and it is likely that although the community composition varied, gene abundance was not limiting in these experiments. In contrast to the production of CO2, N2 and N2O did not show statistically significant correlation to 16S rRNA gene abundance, nor to any of the denitrification genes although there was a moderate negative correlation between nir:nos ratio and N2O-N, N2O%, and 15N atom%. This indicated that as the proportion of nitrous oxide reductase genes (nosZI + nosZII) increased compared to nitrite reductase genes (nirK + nirS), there was proportionally less N2O in the chambers as it was converted to N2.

The biogeography of denitrifying bacteria and their genes, and differences in their relative abundances22,34, is one reason for the lack of consensus between different studies which attempt to define the most important genes. Edaphic factors such as pH are known to influence soil microorganisms: changing soil pH in different plots caused significant differences in denitrification gene abundance35. The relative abundance of nirK and nirS reported elsewhere appears to depend on soil properties and management: relatively more copies of nirS were reported in unfertilized bare fallow compared to nitrate or FYM-fertilized soil, where nirK was more abundant36. Our results, from different soils and treatments, are not directly comparable with these but are consistent with previous measurements in Broadbalk soil25. The relatively higher abundance of nosZI compared to nosZII is consistent with reports that nosZI occurs in Proteobacteria14 which together form the most abundant phyla in all Broadbalk soils. Recently, changes in the diversity and abundance of nirK and nosZ genes expressed in soil microcosms have been related to N2 and N2O emissions37. In our experiments, we could measure transcription of nirK and nosZI only in FYM and woodland soil and there were twice the number of copies of nosZI per nirK in FYM compared to woodland soil which could explain the lower N2O and N2O:N2 measured in the FYM compared to woodland soil in the absence of plants although correlations were not significant. When plants are present, differences in the uptake of the added 15N fertilizer that limit the amount left in soil as a denitrification substrate may change this relationship (see below). We were surprised that 16SrRNA gene abundance was not found to be significantly lower in bare soil compared to planted pots containing rhizosphere soil. Bacterial cell numbers are higher in the rhizosphere than bulk soil38 but the qPCR assays in our study were probably insufficiently sensitive to detect this, compared to the much larger differences due to the source of the soil.

The importance of plants in gas production

In water-saturated conditions conducive to denitrification, both nitrate fertilizer and plants had a major influence on N2O production. This agrees with the findings of a recent meta-analysis where fertilizer quantity and the presence of crops were the most important drivers influencing N2O emissions21. The relative abundance of the various genes involved in NO2− and N2O reduction did not appear to drive the gasses measured in our experiments, indicating that denitrification was not constrained by abundance of nirK, nirS, nosZI and nosZII in the arable soils although differences in relative abundances (nirK:nirS; nir:nosZ) might explain the significantly greater N2O production in bare, fertilized woodland compared to FYM soil. The relatively low N2O-N measured in woodland soils where wheat was grown could reflect the drop to 40% wfps, conditions less conducive to denitrification. The proportion of 15N in N2O-N in bare woodland soil was lower than in the arable soils, and significantly lower where wheat was growing, indicating that some of the “extra” N2O in woodland soil arose from NO3−resulting from nitrification of NH4+ derived from the larger total N pool. The NH4+-N is likely to be a product of SOM mineralisation rather than a legacy of previous fertilization regimes; the results indicate that plants can stimulate the soil microbiome to increase this mineralisation. Nitrifying bacteria, archaea and fungi in soil can oxidize NH4 to NO2− and subsequently NO3−, especially in aerobic conditions.

In our experiments, the most N2O was produced when fertilizer was added to bare soil, and least in unfertilized soil with wheat, and the concentration measured in the chambers did not change significantly over the sampling period. In contrast, N2 increased over time, with more produced when plants were present. The 15N atom% in N2O was also higher in soil with plants. The increase in N2 indicates that N2O was being actively reduced during the experiment, that production and consumption of N2O was in equilibrium, and full denitrification from NO3– to N2 was more efficient in the presence of plants despite their competition for NO3–-N. Most microbes are more active in the presence of plants, stimulated by the energy-rich root exudates, and this may explain both increased 15N atom% in N2O and increased reduction of N2O to N2. In addition, this high degree of conversion of N2O to N2 may arise from the closed experimental system causing denitrifying microorganisms to remain in contact with N2O for longer than in open soil in the field, where N2O is likely to be rapidly lost to the atmosphere. In contrast to other experiments measuring the effect of small plants in enclosed chambers7,18, we used mature plants resulting in all soil being in contact with roots, effectively rhizosphere soil, for comparison to bulk soil in unplanted pots. This would have enhanced the plant effect, explaining the much higher conversion of N2O to N2 that we observed together with the reduction in soil NO3−-N which was rapidly assimilated by plants. The abundance of nosZ is reported to be influenced by resource availability, with relatively fewer copies detected in nutrient-poor environments and an increased proportion present in rhizosphere communities34. The relevance of other differences such as the presence of more copies of nirS in FYM soil, are unclear since they did not result in significant differences in gaseous emissions, although it should be acknowledged that the gene abundance was measured only at the end of the experiment and could have varied over the preceding 4 days.

Soil factors driving denitrification

Emissions of N2O and N2 are known to fluctuate and much greater replication is needed to demonstrate which factors apart from NO3− and wfps are driving denitrification. The closed design of our experiment with daily gas sampling did not allow measurement of gas fluxes but result indicate that the peak concentration of N2O was reached by 24 h and did not increase significantly after this time because it was being reduced to N2, which in contrast, increased each day in most treatments. Our observations that most N2O produced in soil is further reduced to N2 by active denitrifying bacteria is consistent with many reports. However, less N2 than N2O was measured in planted soil fertilized with KNO3 in a sealed system with He7 although the converse was observed in an earlier experiment with Ar18; fertilizer type and water saturation also influenced denitrification in these studies. Our system, in contrast, was designed to have an enhanced rhizosphere effect and growth of plants in a normal atmosphere. The maximum amount of K15NO3-N converted to N2 over the 4 days of our experiment was 4%, in the N6 soil, where 1.6 µg 15N2-N g−1 dw soil was detected after 4 days, derived from 40 µg K15NO3-N g−1 dw soil added when the chambers were sealed. With plants, ~ 5 µg NO3−-N g−1 soil remained after 4 days, suggesting that much more was taken up by the wheat than was available for denitrification. In other studies, N2O-N comprised 7% of (N2O-N + N2-N) emitted from cores of wet agricultural soil39 and 25% in saturated soils amended with artificial root exudates40. There are several reports that the ratio of N2O:N2 increases with the concentration of NO3− in soil7,29,41 but the concentration of K15NO3 in our experiments, 40 μg N g dw soil−1 equivalent to 100 kg N ha−1, was relatively modest (and decreased by 85% over 4 days in the presence of plants) compared with many other studies in vitro. For example, the in vitro study40 with model root exudate applied 100 μg N g dw soil−1 and application rates cited in field surveys were 160–300 kg N ha−1 year−129 and 200–500 kg N ha−1 year−141, respectively. In our pots, NO3− was more evenly distributed throughout the soil than in field applications, so avoiding localised high concentration activity hotspots. High concentrations of NO3− are reported to inhibit nitrous oxide reductase and hence the reduction of N2O to N242,43. In soil microcosms amended with with 50 μg NO3–-N g dw soil−1, N2O comprised > 1% (N2O + N2) after 48 h whereas adding 500 μg NO3–-N g dw soil−1 diminished N2 emissions by 38–90%42. Another study44 found fungal denitrification dominated initially, after high levels of straw and nitrate (200 mg NO3–-N g−1) were added to soil but when NO3–-N fell to 40 mg g−1 soil, N2 evolution dominated.

The constraints on measuring de novo N2 production in the presence of 78% N2 in air, make it difficult to design experiments without adding isotopically labelled N fertilizer and plants reportedly do not flourish when N2 in air is replaced by Ar or He. The acetylene inhibition method previously used to prevent conversion of N2O to N2 has many disadvantages45. Our experiment was designed to determine the effect of well-developed wheat plants at peak root exudation on denitrification, and on the bacterial genes involved, following saturation of the soil, rather than following the development of the plant and denitrification rates over time. We were constrained by the need to sample gas from the chambers manually: future experiments using repeat robotic gas sampling would offer great advantages. Over the short 4 days enclosure, we observed that plants decreased soil NO3− and moisture levels and increased CO2, all of which may influence effect denitrification and could be investigated further.

It would be interesting to attribute the denitrification genes that we detected to the organisms from which they originated. Further studies to sequence amplicons may reveal whether different groups are present and (with improved mRNA extraction) active in the different soils, coupled with estimates of gene abundance over time (within 24 h and over days), and improved monitoring of gaseous emissions may help to explain the high variability observed. Nevertheless, our work reported here confirms that N-substrate (nitrate) availability is the main driver for N2O production in water-saturated soils and furthermore, it shows that the presence of plants promotes further reduction of N2O to N2, which is the major denitrification product. The relative abundance of the various genes implicated in denitrification did not play a major role in these experiments although long-term pre-treatment of soils had generated significant differences in the composition of their soil microbiomes.