Edaphic factors and plants influence denitrification in soils from a long-term arable experiment

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).

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 (F_3,32 = 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 (F_3,32 = 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).

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.

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).

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.

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.

Source: Ecology - nature.com