Rapid increases in soil pH solubilise organic matter, dramatically increase denitrification potential and strongly stimulate microorganisms from the Firmicutes phylum

Soil collection, pH adjustment and DEA assays

The Wakanui silt-loam soil used in this study was sourced from no-till plots in a long-term field trial (12 years) at Lincoln, Canterbury, New Zealand. The basic chemical characteristics of the soil were: pH 5.6; total C, 27 g kg⁻¹; total N, 2.4 g kg⁻¹; NO₃-N, 20 mg kg⁻¹ and NH₄-N, 4 mg kg⁻¹. Further details about the soil and sampling site can be found in Curtin, Peterson & Anderson (2016) and Anderson, Peterson & Curtin (2017). Soil samples were treated with four rates of KOH (base addition rates of 0, 6.0, 16.0 & 20.0 cmol_c kg⁻¹ soil). These treatments were selected based on results from previous experiments (Anderson, Peterson & Curtin, 2017) with the 6.0 cmol_c kg⁻¹ and 16.0 cmol_c kg⁻¹ treatments representing “low/moderate” and “upper limit” pH increases following animal urine deposition. The 20 cmol_c kg⁻¹ treatment represented an alkaline pH outlier, where DEA was expected to be minimal (Anderson, Peterson & Curtin, 2017).

A total of 240 soil microcosms were prepared, covering 4 KOH treatments and 5 incubation times. This provided twelve analytical replicate microcosms for each KOH rate ×  incubation time combination; where, four microcosms were designated for soil chemical analysis prior to DEA assays, two were designated for nucleic acid extraction and microbial culturing work (prior to DEA), and the remaining six microcosms were used for DEA assessments (two triplicate DEA assays, with or without acetylene) (Fig. S1).

Microcosms were prepared as described by Anderson, Peterson & Curtin (2017). Briefly, 25 g (dry weight equivalent) soil was placed in 250 mL bottles and KOH was added together with KCl to balance electrical conductivity across treatments. The final solution volume in each microcosm was adjusted to 25 mL. The bottles were evacuated (to −1 atm.) then flushed three times with N₂ (instrument grade, <0.001% O₂) over a 30 min period until O₂ was <0.03%. The microcosms were then incubated at 20 °C on an orbital shaking platform (150 rpm) for 16, 24, 32, 40 or 48 h. After each incubation the headspace was sampled for N₂O and CO₂.

The following sampling protocol was followed: for chemical analyses, 5 mL of slurry was collected from four microcosms to determine dissolved CO₂ (acidified with 2 mL of 2M HCl to dissolve any carbonates). The remaining slurry from these four microcosms was centrifuged (5min at 20,000 rpm), and the supernatants filtered (<0.45 µm) then frozen (−20 °C), pending pH, EC, DOC, DON and NO${}_3^{-}$/NH${}_4^{+}$ analyses. From a further two microcosms, 1 mL aliquots of slurry were taken for nucleic acid extraction and bacterial colony isolations.

To determine DEA, 5 mL of water containing 50 mg NO₃-N kg⁻¹ (Luo et al., 1996) was added to the remaining six microcosms, and the anaerobic atmosphere was regenerated by evacuation and flushing three times with N₂. To three of the microcosms a volume of scrubbed (Hyman & Arp, 1987) acetylene (final ratio of 10% v/v) was added by syringe and allowed equilibrate (with shaking) for 10 min before venting these microcosms to atmospheric pressure. No external C-sources such as glucose were added. The microcosms were incubated at 20 °C on a shaking platform (150 rpm) and sampled hourly over a 4 h period to measure headspace N₂O and CO₂. Headspace gases removed were replaced with an equivalent volume of N₂.

At the end of the DEA assay, dissolved CO₂, pH, EC, DOC, DON and NO${}_3^{-}$/NH${}_4^{+}$ were measured in the remaining soil slurry samples.

Chemical analysis and gas chemistry

Analytical methods are described in Anderson, Peterson & Curtin (2017). Briefly, concentrations of N₂O and CO₂ were determined on a Shimadzu Corp. GC-17A gas chromatograph and the DEA value (i.e., potential denitrification rate) and respiration rates were calculated from the linear relationship between evolved N₂O or CO₂ and time. The extracts were analyzed for pH (ThermoScientific Orion™ AquaPro™ pH combination electrode) and electrical conductivity (Eutech Instruments PC510 conductivity meter). Dissolved organic C was determined using a Total Organic Carbon Analyzer (Shimadzu TOC-V _CSH, Shimadzu Corp, Japan). Total N was determined by persulfate oxidation, as described by Cabrera & Beare (1993), and organic N was estimated by subtracting mineral N (KCl extracted NH${}_4^{+}$ and NO${}_3^{-}$ determined using an automated colorimeter) from total N.

Microbial population profiling via next generation sequencing

A 1 mL aliquot of soil slurry was centrifuged at 14,000 rpm for 5 min. DNA from the resulting pellet was extracted with the MoBio Powersoil DNA kit (Carlsbad, CA). The V3–V4 variable regions of the bacterial 16S rRNA was amplified with the 341f and 785r primer pair (Klindworth et al., 2013). The fungal internal transcribed spacer 1 (ITS1) region was amplified with NSI1a_mod (5′-GATTGAATGGCTTAGTGAGK-3′) and 58A2R (5′-AGTCCTGCGTTCTTCATCGAT-3′), both adapted from (Martin & Rygiewicz, 2005). Primers included the Illumina adapter sequences.

PCR amplifications contained ∼10 ng DNA template, 10 nmol each primer, 1×  mastermix, and 0.5 U KAPA3G polymerase (Merck, Auckland, New Zealand), in a final volume of 20 µl. Reactions were performed in duplicate. Cycling parameters were 94 °C for 2 min; 30 cycles of 95 °C for 30 s, 50 °C for 30 s, 72 °C for 30 s. Duplicate reactions were combined and purified with AMPure XP beads (Agencourt, Beckman Coulter Life Sciences). Purified amplicons were quantified by gel electrophoresis and UV absorbance (NanoDrop ND-1000). Amplicons were 2 ×300 bp paired end sequenced on an Illumina MiSeq platform (New Zealand Genomics Limited, Auckland).

Sequence processing and statistical analysis

USEARCH v8.0.1517 (Edgar, 2013) was used to merge the paired end reads, filter chimeric sequences and cluster sequences at 97% similarity. An expected error of 1.0 was used for filtering. Singleton reads were discarded. The bacterial 16S OTUs were identified using the RDP Naïve Bayesian Classifier implemented in USEARCH against Greengenes (version 13_8) and fungal ITS OTUs against the UNITE reference dataset (Version 6, 04∕07∕2014, downloaded on 08∕07∕2014) (http://www2.dpes.gu.se/project/unite/UNITE) (Koljalg et al., 2013). Biom (OTU) tables were produced using biom-format (http://biom-format.org/) (McDonald et al., 2012) in USEARCH with rarefication performed in phyloseq.

The phyloseq (McMurdie & Holmes, 2013) and ggplot2 (Wickham, 2009) packages within R (R Core Team, 2017) were used for the analysis and visualisation of data at phyla level. OTU tables were also analysed using Primer 7 with PERMANOVA add-on (Primer-E Ltd, Plymouth, UK). Both rarefied and non-rarefied data was analysed based on the work of McMurdie and Holmes (McMurdie & Holmes, 2014). Where data was not rarefied, samples were standardised by total and no statistical inferences were made regarding differentially abundant species (OTUs), rather our conclusions were based on assessing broader scale relative changes in the microbial communities only. Relationships among microbial community profiles based on Bray–Curtis similarity matrices were graphed using unconstrained non-metric multidimensional scaling (nMDS) ordinations using 250 restarts along with cluster analysis. Relationships observed among all OTUs were then statistically tested using 2-factor permutational ANOVA within the Primer 7 package (PERMANOVA) with estimated components of variation being a standard output of the analysis. Data was untransformed unless otherwise stated, whereupon log(X + 1) transformations were applied.

To understand gradients and group structures across treatments, matrix plots of standardised data were prepared using a reduced sample set of 20 OTUs, with the OTUs retained having the greatest contribution to total counts for the individual samples compared at each pH value, or in the case of fungi, across all pH values. Using reduced sample sets allows simplification of the matrix plots by removing those organisms accounting for a negligible proportion of the total number of OTUs. Samples in the matrix plots were clustered using Bray Curtis similarity (based on all OTUs) while the 20 OTUs presented were clustered according to similarity based on an index of association across samples tested.

Isolation of bacteria, N-use characterisation

Chemical characterisation of pH amended soils prior to DEA assays

After 48 h of anaerobic incubation, soil slurries with additions of 0, 6, 16 and 20 cmol_c kg⁻¹ KOH had average pH values of 4.7, 6.7, 8.3 and 8.8 respectively. During the 48 h incubation EC reached ∼6.4 mS cm⁻¹ in all microcosms except the pH 6.7 microcosms which reached ∼5.7 mS cm⁻¹ (Table 1). DOC in the control microcosms increased from 37 to 63 mg kg⁻¹ during the 48 h incubation. In the pH 6.7, 8.3 and 8.8 microcosms DOC increased ∼11, 185 and 240-fold respectively in response to KOH addition, with the majority of that change (>75%) occurring during the first 16 h of incubation. During the 48 h incubation DON in the control microcosms increased from 2.4 to 7.4 mg kg⁻¹, while in the pH 6.7, 8.3 and 8.8 microcosms DON increased ∼9, 140 and 223-fold in response to KOH addition, respectively (Table 1). There was a strong correlation between the amounts of DOC and DON solubilised at each pH value, irrespective of incubation time (R² = 0.98) (Fig. S2).

Higher respiration (CO₂ production) was associated with DOM increases but the amount of DOC (or DON) solubilised via pH change was not a good predictor of respiration. Respiration in the control and pH 6.7 microcosms generally followed an increasing trend for the first 32 h before decreasing (Table 1). The pH 8.3 microcosms exhibited respiration rates that increased throughout the 48 h incubation period (maximum recorded rate of ∼1,600 ng CO₂-C g⁻¹ h⁻¹). Respiration in the pH 8.8 microcosms followed an opposing trend, declining slightly during the first 32 h, followed by recovery.

In control microcosms, ammonium (NH${}_4^{+}$) increased linearly over 48 h with a slope of 0.09, starting from native soil concentrations of ∼4 mg kg⁻¹. The NH${}_4^{+}$ profiles in the pH 6.7, and 8.3 microcosms were similar (approximately linear) but with greater slopes of 0.28 and 0.79 respectively (R² values of 0.99 and 0.98). In the pH 8.8 microcosms, NH${}_4^{+}$ quickly elevated to 30 mg kg⁻¹ during the first 16 h of incubation and remaining at that concentration until after 32 h, when a further increase from 30 to 50 mg kg⁻¹ occurred by 48 h (Table 1). The native soil nitrate (NO${}_3^{-}$) concentration was ∼20 mg kg⁻¹. This NO${}_3^{-}$ was almost completely reduced after 32–40 h incubation in the control, pH 6.7 and pH 8.3 microcosms (Table 1). In the pH 8.8 microcosms, NO${}_3^{-}$ remained at ∼20 mg kg⁻¹ until after 32 h incubation, dropping to ∼6 mg kg⁻¹ during the following 16 h.

The control microcosms had the highest concentrations of N₂O in the headspace, increasing from ∼1,650 to 6,800 µg kg⁻¹ between 16 and 48 h of incubation (Table 1). The pH 6.7 microcosms had ∼850 µg kg⁻¹ of N₂O in the headspace after 16 h incubation, but by 24 h this had declined to <15 µg kg⁻¹. Headspace N₂O in the pH 8.3 and 8.8 treatments was <15 µg kg ⁻¹for all time points except 24 h for the pH 8.3 treatment (125 µg N₂O kg⁻¹ soil).

DEA assays

Microcosms incubated for 16 h had the lowest DEA rates with 80, 540, 34 and 0.2 ng N₂O g⁻¹ h ⁻¹ recorded for the control, pH 6.7, 8.3 and 8.8 treatments respectively (Fig. 1A). DEA remained low in the control microcosms with a maximum DEA rate of only 121 ng N₂O g⁻¹ h⁻¹ after 24 h incubation, declining to 85 ng N₂O g⁻¹ h⁻¹ in microcosms incubated for 48 h (Fig. 1A). In contrast, after 24 h incubation, DEA rates in the pH 6.7 and 8.3 microcosms were 8 and 25-fold higher respectively with a maximum of ∼3,000 ng N₂O g⁻¹ h⁻¹ produced (Fig. 1A). DEA rates remained in the vicinity of 900 and 2,000 ng N₂O g⁻¹ h⁻¹ for the pH 6.7 and 8.3 treatments, respectively, for microcosms incubated up to 40 h. DEA rates greater than those in the control were not observed in the pH 8.8 treatment until microcosms were incubated for at least 48 h (∼65 ng N₂O g⁻¹ h⁻¹).

Replicate microcosms without acetylene added were used to assess N₂O-R activity during the DEA assays based on the N₂O/(N₂O + N₂) ratio. These showed that N₂O reduction was absent in the control microcosms while in the pH 6.7, 8.3 and 8.8 treatments between 56 and 100% of the N₂O produced was reduced to N₂ depending on incubation time (Fig. 1B). Near complete N₂O reduction was observed in the pH 8.3 treatment for microcosms incubated longer than 24 h. Depending on pH treatment and incubation time, N₂O production (and reduction) profiles during the DEA assays were mirrored by decreases in NO${}_3^{-}$ with an estimated 30 to 80% of the available NO${}_3^{-}$ reduced. In the control and pH 6.7 microcosms there were no differences in the NH${}_4^{+}$ before and after the 4-hour DEA assay period, however decreases of up to 9 mg kg⁻¹ occurred in the pH 8.3 microcosms after 32 h incubation and in the 8.8 microcosms after all incubation times (Table ST1).

T-tests comparing CO₂ respiration indicated that acetylene addition depressed microbial activity, but only in the pH 8.3 and 8.8 treatments (P < 0.05). It is acknowledged that lower soil respiration in the presence of acetylene will reflect both the absence of any CO₂ derived from metabolic N₂O reduction but possibly also general impediment of other anaerobic metabolisms. It is unlikely that acetylene would have served as dominant carbon source during the 4 h DEA incubation given the excess DOC available and time required to adapt to using acetylene (Felber et al., 2012; Groffman et al., 2006). There is evidence to suggest that DOC and DON declined during the DEA assay period, especially in the pH 8.3 and 8.8 treatments but the results were highly variable (Table ST1).

Microbial community adaptation to pH treatments during 48 h incubation

The total number of OTUs identified across all treatments and incubation times was 2,258 for fungi and 6,429 for bacteria with an average of 84,000 reads per sample. Sequences from this Targeted Locus Study (LTS) project has been deposited at DDBJ/EMBL/GenBank under the accession KCDA00000000. The version described in this paper is the first version, KCDA01000000. Fungal ITS1 sequences have been deposited at DDBJ/EMBL/GenBank under the accession numbers MH624180 –MH625694. Two factor tests using PERMANOVA [pH ×  incubation time] for all OTUs supported the nMDS observations indicating significant interactions between time and pH (P = 0.001 for both bacteria and fungi) (Fig. 2). Approximately ∼90% of the total variation was explained for the bacterial relationships whereas only ∼43% was explained for the fungi. Of the variation explained for the bacteria, up to 60% was attributed to pH, a further 15% to incubation time, and 18% was attributed to the interaction between the two factors. The corresponding values for the fungal communities were 28% of variation attributed to pH, 6.6% to incubation time and 8.9% to the interaction between the two factors (Fig. 3).

The phylum level bacterial profile from the control (pH 4.7) microcosms after 16 h incubation was made up of Acidobacteria (∼7%), Actinobacteria (∼12%), Bacteroidetes (∼7%), Proteobacteria (∼20%), Firmicutes (∼5%), Planctomycetes (∼5%) and phyla that had abundances of >5% including Verrucomicrobia, Chloroflexi, Gemmatimonadetes and Armatimonadetes. Up to 35% of OTUs could not be classified (Fig. S3). After 24 h incubation, the bacterial communities were still >85% similar before there was a relative decrease in Actinobacteria coupled with an increase in Acidobacteria and Bacteroidetes (Fig. S3). A decrease in Actinobacteria was also seen over the incubation periods for all three pH modifications along with changes in the proteobacterial populations with almost complete disappearance of OTUs from the order Rhizobiales (Fig. 4, Fig. S3).

After 16 and 24 h incubation, the bacterial profiles in the pH 6.7 microcosms were similar to those in the control microcosms with the exception of a larger proportion of Firmicutes (∼20% versus 5% in the controls) (Fig. S3). The increased representation of Firmicutes was driven by an expansion of OTUs from the orders Bacilliales (OTU_4) and Clostridiales (OTU_1, Tepidibacter sp.). These two OTUs then decreased later in the incubation, partially displaced by other Clostridium spp. (OTUs 16 and 26) (Fig. 4B). Across the 32 to 48 h incubation period there was a large relative increase in OTUs from the phylum Bacteroidetes, driven by Flavobacterium spp. (OTUs 62, 9, 80, 410, 37 and 41) (Fig. 4B).

In the pH 8.3 treatment after 16 h incubation, there was a very high proportion of Firmicutes (∼40%) and a low representation of Actinobacteria and Bacteroidetes compared with the pH 4.7 control microcosms. Across the incubation period, the dominance of Firmicutes increased further, initially driven by the same Bacilliales and Clostridiales OTUs (4 and 1), but at a much higher relative ‘abundance’ than observed in the pH 6.7 treatment. Their expansion coincided with the highest DEA rates, with the average ‘abundance’ of OTU_4 being ∼20% of total at this time and OTU_1 being ∼23% (Fig. 4C, Fig. S3). After 32 h these two dominant OTUs were partially displaced by a cohort of other OTUs from the orders Bacilliales and Clostridiales (OTUs 2, 75, 31, 25, 36, 12, 13, 26 and 109) (Fig. 3). At the genus level these OTUs represented Bacillus, Paenibacillus, Clostridium and Alkaliphilus spp. (Fig. 3). Communities in the 40 and 48 h incubations shared the highest level of similarity (Figs. 2, 4C and Fig. S3).

After 16 and 24 h incubation, the bacterial profiles from the pH 8.8 treatment shared >70% similarity with the samples from the control microcosms (Fig. 2). Over the incubation period to 32 h, there was a large expansion in the relative percentage of the Firmicutes, initially driven almost entirely by an increase in OTU_5 from the Bacilliales order (Fig. 4D). By 48 h, this OTU was partially displaced by a group of Bacilliales and Clostridiales OTUs that shared some similarity with those observed in the pH 8.3 treatment (OTUs 1, 4, 156, 36, 12, 13, and 199). These OTUs included the same genus level representatives—Bacillus, Paenibacillus, Clostridium and Alkaliphilus spp. (Fig. 4D).

Fungal communities shared >70% similarity across all samples regardless of pH and time (Fig. 3). Based on percentages alone, no discernible patterns could be observed for the fungal dataset (Fig. S4), however, some OTUs were displaced depending on pH. For example, Zygomycota OTU_3 was dominant in the control and pH 6.7 treatments but was partially supplanted by another Zygomycota OTU from the order Mortierellales (Mortierella sp.) (Fig. 5).

An identical analysis of rarefied bacterial and fungal OTUs datasets is presented in Figs. S5 to S10. A range of diversity measures have been presented in Figs. S11 and S12. An nMDS plot from a preliminary experiment showing the relationship between bacterial communities in the initial untreated (aerobic) field-moist soil and the soil microcosms incubated under anaerobic conditions with various pH treatments is presented in Fig. S13.

Isolation of bacteria and N-use characterisation

On TSB-nitrate medium, plates were quickly dominated by fast growing colonies sharing morphologies characteristic of motile or swarming bacteria. A total of 33 isolates were screened for nitrate utilisation. Seven showed minimal growth and respiration in liquid culture, 18 showed near complete utilisation of NO${}_3^{-}$ accompanied by production of both NH${}_4^{+}$ and N₂O, and five showed moderate utilisation of NO${}_3^{-}$ with low production of NH${}_4^{+}$ and N₂O. One isolate showed production of NH${}_4^{+}$ with little or no use of NO${}_3^{-}$ or production of N₂O and two isolates exhibited respiration but did not appear to utilise N. Of the 33 isolates, 22 were selected for identification by ribosomal 16S gene DNA sequencing. All were from the Firmicutes phylum, of which three were Paenibacillus spp. and one was a Brevibacillus sp. (all producing less NH${}_4^{+}$), while the remainder were Bacillus spp. (Table  ST2A).

From 1/10 diluted TSB-nitrate medium, 44 isolates were screened for nitrate utilisation. Of these, 24 showed no or minimal growth in liquid culture over 48 h incubation while 13 reduced NO${}_3^{-}$ to close to zero with four of these generating significant amounts of NH${}_4^{+}$. Another three reduced NO${}_3^{-}$, but to a lesser extent, while four isolates respired but did not appear to utilise N. Seventeen isolates were identified by 16S gene sequencing. Six belonged to the Bacillus genus, three to Achromobacter, six to Acidovorax, one to Bosea, and one to Rhodanobacter (Table ST2B).

All isolates except one of the Bacillus sp. had high NO${}_3^{-}$ utilisation compared to uninoculated controls. The Bacillus spp. produced up to 536 mg NH${}_4^{+}$ L⁻¹, while the Acidovorax spp. produced <45 mg NH${}_4^{+}$ L⁻¹. One Bacillus sp. and two Acidovorax sp. produced <100 mg N₂O L⁻¹, while the remaining Acidovorax sp. had the highest N₂O production at 2,840 mg L⁻¹ . In general, isolates that exhibited high use of NO${}_3^{-}$ coupled with production of NH${}_4^{+}$ and N₂O exhibited an average of ∼2.5-fold higher CO₂ production (respiration) compared with isolates that exhibited high NO${}_3^{-}$ use with little or no NH₄ and N₂O production. Relevant 16S sequences were submitted to NCBI with accession numbers assigned between MH211426 and MH211463, submission number SUB3915485 (Table S2). Six organisms were selected for future genome sequencing, three Bacillus spp. and three Acidovorax spp. (Table ST2C).

The effects of rapid pH change on soil chemistry and microbiology

KOH additions and resulting pH elevation caused the concentration of DOM in soil microcosms to greatly increase (150-fold) compared to the controls. Added hydroxyl ions displaced negatively-charged organic molecules into solution. Previous work has shown that monovalent cations like K⁺ (KOH) and NH${}_4^{+}$ (NH₄OH—product of urea hydrolysis) are much more effective in solubilising organic matter than divalent cations such as Ca²⁺ (Ca(OH)₂) (Curtin, Peterson & Anderson, 2016). The amount of DOM released at elevated pH in these experiments was concordant with our previous research using the same methodology (Anderson, Peterson & Curtin, 2017), as were the higher respiration rates.

Over medium to long time scales (months to years), pH is known to be a dominant environmental variable that shapes soil microbial communities (Lauber et al., 2009; Zhalnina et al., 2015). Changes in pH are also known to cause shifts in active organisms over short timescales (Brenzinger, Dörsch & Braker, 2015). Although the strongest predictor for both bacterial and fungal community change in these experiments was pH, community change may also be indirectly influenced by the effect that pH has on DOM release. Theoretically, high levels of DOM released via increased pH should benefit copiotrophs (r) over oligotrophs (K) (Fierer, Bradford & Jackson, 2007; Goldfarb et al., 2011), but at the same time elevated pH is likely to alter cellular homeostasis, regulation of nutrient availability, or other factors such as salinity, metal accessibility, or organic C characteristics (Lauber et al., 2009).

Previous analysis of the soil used in this study suggests that a mix of carbon sources are released as pH increases, of which 45% are bioavailable. These range from labile hexose and pentose sugars to more recalcitrant polyphenolic molecules (Curtin, Peterson & Anderson, 2016). The lack of proportionality between respiration rates and DOM released in this study suggests that higher amounts of bioavailable C did not lead to higher biomass, instead the microbial community and associated metabolic response has shifted toward more copiotrophic organisms. Addition of low molecular weight C compounds (glucose, citric acid, glycine) to soil has been previously observed to shift the structure of bacterial communities to more copiotrophic organisms (Eilers et al., 2010) with no strong correlations between respiration rates and community structure. Community changes and catabolic responses may be unlinked because some C-substrates are preferentially used without biomass changes (Devevre & Horwath, 2000).

Proteobacteria are abundant in high C soils (Fierer, Bradford & Jackson, 2007) with β and γ − Proteobacteria considered important soil copiotrophs (Eilers et al., 2010) in conjunction with Firmicutes and Actinobacteria (Zhalnina et al., 2015). Bacteroidetes and β-Proteobacteria are initial metabolisers of labile soil-C (Padmanabhan et al., 2003) and increases in the abundance of these organisms have been correlated with C mineralisation rates (Fierer, Bradford & Jackson, 2007). Our study is consistent with regard to expansion of Bacteriodetes (Flavobacteriales), specifically in the control and pH 6.7 microcosms (∼5.5-fold-increase in DOM with ∼2-fold increase in respired CO₂), suggesting that the microbial community does respond to higher concentrations of bioavailable C at pH values <7.

In general, we observed a decrease in α-Proteobacteria (specifically Rhizobiales), Actinobacteria and Acidobacteria OTUs at all pH and DOM values, while a few β- and γ-Proteobacteria OTUs increased. Our results suggest that the chemical changes induced by KOH addition to soil are comparable to soils where pH and DOM are elevated due to higher urea inputs. Niche differentiation occurs in soil where higher bovine density (and presumably urea inputs) induces increases in pH and total organic carbon, with Actinobacteria, α-Proteobacteria and Verrucomicrobia decreasing and Bacteriodetes increasing (Philippot et al., 2010; Philippot et al., 2009a; Philippot et al., 2009b).

In the study by Fierer, Bradford & Jackson (2007), abundance of Firmicutes could not be predicted by C-mineralisation (nor other measured soil parameters), while in the Park Grass experiment in the UK, total C and N and pH were negatively correlated with Firmicutes (Zhalnina et al., 2015). Our study differs from the literature with regard to the Firmicutes as they are the most responsive to pH and DOM increases. Our experimental conditions are quite different with the combined complexity of alkaline pH and anaerobic conditions likely playing a larger role than just DOM in defining niche differentiation and shaping microbial community structure (Banerjee et al., 2016; Husson, 2013; Pett-Ridge & Firestone, 2005).

Comparisons of OTU distributions in our study indicate that the dominant feature driving sample dissimilarities was large increases in Firmicutes from the classes Bacilli and Clostridia. Large expansion of Firmicutes, first dominated by Bacillales (up to 46%) and then followed by Clostridiales (up to 53%), (along with large decreases in Proteobacteria) have been observed in alkaline soil crusts (pH 8.5) that were rehydrated and incubated under dark anoxic conditions (Angel & Conrad, 2013). Our results are also concordant with O’Callaghan et al. (2010), who observed a 38% increase in Firmicutes, and decreases in Proteobacteria (18%), Acidobacteria (8%), Actinobacteria (5%), and Bacteriodetes (5%), in soil where pH rose to values of >8, two days after bovine urine addition.

Genus level identifications from the Bacilli and Clostridia in our study included Bacillus, Paenibacillus, Tepidibacter, Alkaliphilus and Clostridium. Cultured examples of these organisms from the literature include facultative anaerobes (and obligate anaerobes) that are highly responsive to more recalcitrant C-sources, are either alkaliphilic or alkalotolerant, and show fermentative type metabolisms (Chen, Song & Dong, 2006; Goldfarb et al., 2011; Lee et al., 2007; Slobodkin et al., 2003; Urios et al., 2004). Bacillus species and related genera can be found in a wide variety of habitats. Bacillus and Paenibacillus species can be considered as drivers of soil organic matter mineralisation, are frequently abundant in situations where C and N are not limited and are capable of degrading polymeric carbonaceous substances (Mandic-Mulec, Stefanic & Van Elsas, 2015). Although there is some evidence to suggest that Clostridium are generally acid loving (Kuhner et al., 2000), the related clostridial OTUs (Tepidibacter and Alkaliphilus) identified in this study suggest a wider range of pH tolerance (Lee et al., 2007; Slobodkin et al., 2003; Urios et al., 2004).

Relative changes in fungal populations in response to pH change were smaller than for bacteria, with ∼28% of fungal variation attributed to pH, versus ∼57% for bacteria. Only ∼6.6 percent of the variation in fungal communities could be attributed to incubation time. Fungal communities are known to be less responsive to pH than bacteria (Lauber et al., 2009; Rousk et al., 2010a) and fungal abundance has been found to be negatively correlated with pH (Rousk, Brookes & Baath, 2010b), but positively correlated with C and N additions (Banerjee et al., 2016). For example, investigations of the response of fungal (and bacterial) communities to ovine urine (where pH increased from ∼3.5 up to 6.5 and DOC increased from ∼0 up to ∼2,000 mg kg⁻1) indicates no fungal biomass change (Nunan et al., 2006; Williams, Grayston & Reid, 2000), no significant correlation between biomass and pH, NH${}_4^{+}$ or NO${}_3^{-}$, but weak correlation between biomass and DOC (Singh, Nunan & Millard, 2009). In pH 6.1 soils, Banerjee et al. (2016) noted that although fungal biomass increased, the richness, evenness, and diversity decreased within 4 days after organic matter and nutrient addition leading to ‘keystone’ fungal species being favoured. Singh, Nunan & Millard (2009) suggest that because fungi are capable of degrading complex organic carbon they are less responsive to short term changes in nutritional availability.

Soil N processes and their relationship to microbiology following pH change

Although the possibility exists that nitrifier-denitrification contributed to N₂O production during the incubation period in the control and pH 6.7 microcosms, the likelihood is low given the elevated DOM concentrations, the saturated conditions and the low concentrations of native NH${}_4^{+}$ available for oxidation (Wrage et al., 2001). Under saturated conditions (WFPS 90%), experiments by Kool et al. (2011) indicated that >92% of total N₂O is derived from ‘conventional’ denitrification of NO${}_3^{-}$ and our results are broadly concordant given that NO${}_3^{-}$ was consumed and N₂O was produced. There were however variations in the amount of time required for the original supply of native NO${}_3^{-}$ to be consumed. Native NO${}_3^{-}$ was nearly completely utilized within the first 16 to 24 h in the control, pH 6.7 and pH 8.3 microcosms, but in the pH 8.8 microcosms, NO${}_3^{-}$ did not decline until after 32 h. When additional NO${}_3^{-}$ was added to measure denitrification rates, maximum DEA occurred in microcosms that had been incubated for at least 24 h. The DEA assay measures only N₂O produced during the assay period as the anaerobic headspace was refreshed. For all treatments, DEA potential declined at incubation times greater than 24 h. This may have been due to extended periods (i.e., >24 h) of low NO${}_3^{-}$ concentrations prior to the DEA assays leading to a decline in NO${}_3^{-}$ linked translation of denitrification genes such as N₂O-R and nitrate reductase (Moreno-Vivian et al., 1999; Zumft, 1997). Alternatively, the microbial communities that developed with increasing incubation times may have expressed different denitrification phenotypes (Dorsch, Braker & Bakken, 2012; Sanford et al., 2012).

It also seemed that the relative metabolic contribution of denitrification declined over time as NO₃⁻ was utilized, giving the opportunity for other anaerobic metabolisms such as fermentation to have proportionally greater influence. For example, in the pH 8.3 treatment, the respiration rate in microcosms incubated for 24 h was 40% lower than those incubated for 48 h, yet at 24 h denitrification rates were ∼4.5-fold higher. Bacillus OTUs dominated at 24 h where the lower respiration rates and higher denitrification was observed, but Bacillus was then displaced by a consortium of clostridial species after 48 h. Given that Clostridia can be obligate or facultative anaerobes, we think that this species displacement is a response to the changing chemical conditions in the microcosms marked by nitrate depletion, elevated pH and sustained anaerobicity.

In contrast to the KOH amended microcosms, the N₂O/(N₂O + N₂) ratio in the control microcosms equalled 1, indicating that this treatment did not have active N₂O-R. Liu et al. (2010a) and Bakken et al. (2012a) have previously shown that production of functional N₂O-R depends on the post-transcriptional pH being greater than 6.1, which is consistent with our results. In an agricultural environment this raises interesting ecological questions, because urea hydrolysis happens to elevate pH for several days which would immediately alleviate any post-transcriptional interference of nos Z expression and allow rapid production of functional N₂O-R.

We observe full denitrification of NO${}_3^{-}$ to N₂ within 16 h and maximum rates after 24 h which indicates that suitable redox conditions for denitrification were established quickly in our microcosms and a corresponding rapid genetic and enzymatic response followed. The predominant electron acceptors in a weakly reducing environment are O₂, NO${}_3^{-}$ and manganese oxide (MnO₂) (Uteau et al., 2015) with the threshold between oxic and anoxic soil lying somewhere between 300 and 400 mV. These conditions develop in response to high soil moisture contents that slow down gas diffusion (e.g., post irrigation or flooding) and there are good correlations between N₂O flux and relative soil gas diffusivity (D_p/D_O) (Hansen, Clough & Elberling, 2014; Owens et al., 2017; Owens et al., 2016). Biologically, low O₂ concentrations, or restricted diffusion of oxygen would trigger rapid induction of de novo denitrification enzyme synthesis depending on pH. De novo enzyme synthesis follows a sequential order, with nitrate reductase formed within 2–3 h, nitrite reductase between 4–12 h and N₂O-R between 24 and 42 h (Dendooven & Anderson, 1994; Dendooven & Anderson, 1995; Firestone & Tiedje, 1979; Smith & Tiedje, 1979). Recent investigations have observed even earlier synthesis of N₂O-R than 24–42 h, with peaks in gene transcripts for nos Z (and presumably translation of N₂O-R) occurring within <10 h (Liu, Frostegard & Bakken, 2014).

The denitrification trait is spread over a wide taxonomic range including bacteria, archaea and some eukaryotes (Zumft, 1997). We observed large proliferations of Firmicutes in conjunction with peaks in DEA. Denitrification and/or reduction of nitrate/nitrite is common in cultured Bacillus spp. and they have been shown to be numerically important culturable members of denitrifying communities in agricultural soils (Jones et al., 2011; Verbaendert et al., 2011). The closely related Paenibacillus (pH 8.3 microcosms) are also capable of heterotrophic nitrification, dissimilatory NO${}_3^{-}$ reduction to NH${}_4^{+}$ (DNRA), and full denitrification and grow optimally in neutral to alkaline pH conditions (Behrendt et al., 2010). Like Paenibacillus, some Bacteriodetes (as observed in the pH 6.7 microcosms) have N₂O-R and have been observed to fully denitrify NO${}_3^{-}$ to N₂ (Horn et al., 2005). To date, culture independent studies have not shown Firmicutes to be numerically important in denitrification, however, PCR primers and lysis techniques may not be effective for these bacteria, thereby artificially reducing their relative contribution (Verbaendert et al., 2011).

Potential for DNRA and fermentation in the microcosms

After pH change, the microcosms had high DOC/NO₃⁻ ratios with no correlations evident between estimated DOC mineralisation and NO${}_3^{-}$ nor CO₂ respiration. There was also a mismatch between increasing NH${}_4^{+}$ relative to consumed NO${}_3^{-}$, especially in the pH 8.8 microcosms, suggesting that other anaerobic metabolisms were active aside from denitrification. DNRA is an energy yielding anaerobic process that is favoured in C-replete conditions when NO${}_3^{-}$ becomes limiting (C/NO${}_3^{-}$ ratio > 12) (Giles et al., 2012; Rutting et al., 2011). Using the DNRA stoichiometry in equation [2] presented by Lam et al. (2009), if all available native NO${}_3^{-}$ in our microcosms (∼20 mg kg⁻¹ NO${}_3^{-}$-N) was reduced via DNRA (i.e., ignoring NO${}_3^{-}$ also needed for denitrification) then approximately 26 mg kg⁻¹ NH${}_4^{+}$ could be produced, yet we observed up to 50 mg kg⁻¹. The significant amounts of additional N required to balance the N requirements in our experiments are likely to be derived from the ample supplies of DON and DOC in the microcosms that could undergo depolymerisation and ammonification (Burger & Jackson, 2004; Rutting et al., 2011; Schimel & Bennett, 2004).

Bacillus species are well known as nitrate reducers and N₂O emitters, but many strains do not produce N₂O after NO${}_3^{-}$/NO${}_2^{-}$ reduction (Verbaendert et al., 2011). DNRA is known to occur in a number of Bacillus species with varying concentrations of N₂O produced (Heylen & Keltjens, 2012; Mania et al., 2014; Nakano et al., 1998; Sun, De Vos & Heylen, 2016). Although genome information is not yet available, possible N-metabolisms for the predominantly Bacillus species isolated from the microcosms in this study include denitrification, DNRA and possibly N₂ fixation. These isolates produced an excess NH${}_4^{+}$ and N₂O compared to uninoculated controls suggesting DNRA could be a dominant metabolism. Given the concentration of NO${}_3^{-}$ available in the medium, additional N is still required to support the concentrations of NH₄⁺ and N₂O produced.

The production of NH${}_4^{+}$ and N₂O at concentrations beyond that supplied by mineral-N in both microcosms and cultures allows speculation that depolymerisation and ammonification of organic matter is active (Curtin, Peterson & Anderson, 2016), along with the possibility of co-denitrification (biological and/or chemical) where 50% of the N in N₂O (and N₂) is derived from labile (nucleophilic) organic N such as amines (Phillips et al., 2016; Selbie et al., 2015; Spott, Russow & Stange, 2011). Biogenic amines can also be derived through fermentation, with Bacillus spp. being common producers (Chang & Chang, 2012).

The other main driver of treatment differences were organisms from the order Clostridiales. Tiedje (1988) describes obligate anaerobic DNRA capable Clostridium spp. (Caskey & Tiedje, 1980; Keith, Macfarlane & Herbert, 1982). Clostridia are also well known for their fermentative metabolisms that have been exploited for over 100 years (Moon et al., 2016; Wiegel, Tanner & Rainey, 2006), which combined with the observations that there was no correlation between NO${}_3^{-}$ consumption and CO₂ production adds support to the theory that metabolisms aside from denitrification/DNRA operate in these microcosms, especially beyond 40 h when Clostridiales start displacing Bacilliales. Fermentation is also known to occur in Bacillus species, specifically the well-studied B. subtilis (Ramos et al., 2000).

Recent research has shown that fermentative organisms (Clostridiales) influence the competition between denitrifiers and DNRA bacteria through competition for fermentative C-substrates (electron donors). Higher ratios between substrates and nitrate leads to a combination of fermentation and DNRA (both fermentative and respiratory) with no denitrification. When the ratio between substrates and nitrate lowers, denitrification takes a larger role until it eventually out-competes both fermentation and DNRA (Van den Berg et al., 2017a; Van den Berg et al., 2017b). Comparing these studies to our microcosms is problematic as the C-sources in our study are so diverse. However, the biogeochemical evidence indicates that NO${}_3^{-}$ quickly declines while DOM remains high which would lead to a higher substrate/nitrate ratio and thus DNRA and fermentation taking a dominant role. This possibility is further supported by increases in NH${}_4^{+}$ and CO₂ respiration rates beyond the peaks in DEA. There is also the possibility that DNRA has a more significant role that we envisage and that DEA measurements reflect reduction of N₂O via ‘atypical nosZ’ (Giblin et al., 2013; Jones et al., 2013; Jones et al., 2011; Samad et al., 2016b; Sanford et al., 2012).

Implications for urea impacted soil

The pH changes induced by our KOH additions in this study are representative of what could be expected in the field under animal urine patches or in the vicinity of urea fertilizer prills (Clough et al., 2010; O’Callaghan et al., 2010). Increases in electrical conductivity (EC) associated with elevated pH were high at 6 mS cm⁻¹, but were not out of the ordinary when compared to other studies investigating urine additions to soil, nor is the associated release of excess SOM (Curtin, Peterson & Anderson, 2016; Haynes & Williams, 1992). Recent research documenting N₂O emissions in urea-amended saturated soils with elevated pH and declines in O₂ and redox changes, also suggest that this work has direct relevance to what would be expected under field conditions (Hansen, Clough & Elberling, 2014; Owens et al., 2017; Owens et al., 2016).

Given that there is experimental and field evidence for chemical conditions conducive to denitrification in the days following urea application, this supports the idea that N-cycling in soils ideally should not be considered a sequential process, but instead is highly dynamic, with N-processing dependent on resources available and the physicochemical environment. Our work suggests that N-resources can be quickly supplied from both organic and inorganic sources with the distinct possibility that significant N could be lost as N₂ shortly after urea deposition. It is unknown what happens to the large excess of SOM released at high pH that is not metabolised during these short-term incubations. We should be aware that pulses of fresh C into soil can lead to loss of native C (priming) with some research indicating that excess N-fertilisation (and possibly intensification) leads to soil C declines (Kuzyakov, Friedel & Stahr, 2000; Mau et al., 2015).

Although pH is a universal mechanism that selects for microbial communities, the response to pH will vary according to soil types. The microbial phenotypes expressed will also be dependent on the available soil C and N resources with multiple N-transformation pathways possible. For this particular soil, pH elevation and anoxia allowed Firmicutes bacteria to flourish and contribute to rapid processing of N resources. Investigations of the ratios between various N-metabolisms in these microcosms would require isotope labelling with more defined experiments required to understand short and long-term cascades of biological N-processing and the transient ecologies driving N-transformations as soil conditions stabilise post urea addition.