Soil samples were collected in October 2015 from a tropical peatland, locally known as Quistococha (03∘50.024′ S 73∘19.235′ W), near Iquitos (Loreto, Peru). The soil geochemistry of this site has been described previously (Lawson et al., 2014; Lähteenoja et al., 2009). The samples were obtained from depths of 15–30 cm below the water table and kept strictly anoxic during transport and storage at 4 ∘C in the dark. Water saturation and organic carbon content were determined by oven drying and loss on ignition, respectively. Dissolved organic carbon (DOC) was determined by high-temperature combustion using a Shimadzu TOC-V Total Organic Carbon Analyzer (Shimadzu Scientific Instruments, Columbia, MD). Inorganic N species were quantified photometrically using an AQ2 Discrete Analyzer (Seal Analytical, Southampton, UK) and method EPA-103-A Rev.10 for ammonium (NH4+; LoD 0.004 mgNL-1, range 0.02–2.0 mgNL-1) and method EPA-127-A for nitrate (NO3-)/nitrite (NO2-; LoD 0.003 mgNL-1, range 0.012–2 mgNL-1). Hydroxylamine was measured photometrically using the iodate method (Afkhami et al., 2006).

Experiments were started within 6 weeks of soil collection. For each sterilization procedure, anoxic wet soil was exposed to the chemical sterilant 48 h prior to start of the NO2- incubation or sterilized by physical treatment and allowed to equilibrate for at least 12 h. The untreated (live) control was incubated as a slurry without any additions or treatments for 48 h prior to start of the NO2- incubation. Anoxic vials filled with wet soil were irradiated with a 60Co source for 7 d, yielding a final radiation dose of 4 Mrad (40 kGy). The irradiated soil was then prepared for incubation in an anoxic glove box (0.5 % H2 in N2) with disinfected surfaces and sterilized materials to prevent contamination. For autoclaved samples, soil was prepared for incubation in closed vials and autoclaved at 121 ∘C and 1.1 atm for 90 min. The CHCl3-treated samples were fumigated for 48 h under a 100 % N2 atmosphere. Because volatilized CHCl3 corrodes electron capture detectors used for N2O detection (see below), CHCl3 was removed by flushing the vials with N2 for 5–7 min immediately before the start of incubations.

In contrast to the physical sterilization treatments, soil samples were continuously exposed to the chemical inhibitors throughout their incubation. Sodium azide (NaN3, Eastman Organic Chemicals), zinc chloride (ZnCl2, Fisher Scientific), and mercuric chloride (HgCl2, 99.5 %, Acros Organics) were added from anoxic stock solutions to final concentrations of 150, 87.5, and 3.7 mM, respectively. The Hg concentration was the minimum needed to eliminate microbial heterotrophic growth based on visual inspection of soil extract on agar plates exposed to 0.5 to 92.1 mgL-1, which includes concentrations demonstrated to be effective previously (Tuominen et al., 1994).

After the initial physical or chemical treatment, triplicate incubations were diluted 1:10 in 20 mL of autoclaved 18.2 MΩcm water in 60 mL glass serum vials. All microcosms were prepared in an anaerobic glove box (0.5 % H2 in N2) prior to incubation. Triplicate soil slurries were amended from an anoxic, sterile stock solution to a final concentration of 300 µMNO2- (6 µmol in 20 mL) and sealed with thick butyl rubber stoppers. A parallel set of samples was amended with 300 µMNO3- to evaluate denitrification potential with CO2 measurements. Control incubations received an equivalent volume of autoclaved 18.2 MΩcm water without NOx-. Soil microcosms were incubated in the dark at a constant temperature of 25 ∘C. NO2- was quantified in all soil treatments using the Griess assay (Promega, Kit G2930). pH measurements were taken with an Orion 3 Star meter (Thermo Scientific) before and after sterilization, and at the end of the experiment after 70–76 h of incubation.

To quantify N2O and CO2 production, 200 µL of headspace gas was sampled with a gas-tight syringe (VICI Precision Sampling) and injected onto a gas chromatograph (GC, SRI Instruments) equipped with both an electron-capture detector (ECD) and a flame-ionization detector (FID). Two continuous HayeSep-D columns were kept at 90 ∘C (oven temperature); N2 (UHP grade 99.999 %, Praxair Inc.) was used as carrier gas, and for FID combustion H2 was supplied by a H2 generator (GCGS-7890, Parker Balston). For CO2 measurements, a methanizer (which reduces CO2 to the detectable CH4 via a Ni catalyst at 355 ∘C) was run in line before the FID. The ECD current was 250 mV and the ECD cell was kept at 350 ∘C. The N2O and CO2 measurements were calibrated using customized standard mixtures (Scott Specialty Gases, accuracy ±5 %) over a range of 1–400 and 5–5000 ppmv, respectively. Gas accumulation in the incubation vials was monitored over time. Gas concentrations were corrected using Henry's law and the dimensionless concentration constants kHcc(N2O)=0.6112 and kHcc(CO2)=0.8313 (Stumm and Morgan, 2012) to account for gas partitioning into the aqueous phase at 25 ∘C.

To assess the efficacy of sterilants or inhibitors visually, the bacterial viability kit LIVE/DEAD BacLight L7012 (Molecular Probes, Invitrogen) containing SYTO9 and propidium iodide dyes was used to stain and distinguish dead and living cells on the basis of intact cell walls. The green (live) and red (dead) signals were counted at 60× magnification from 10 squares of 0.01 mm2 randomly distributed in the center of a 5 µL Neubauer chamber, using an Olympus BX-61 microscope with the FITC/Cy5 filter set. Photographs were taken with an Olympus DP-70 camera attached to the microscope. Particles were counted with ImageJ software version 1.50i (Abràmoff et al., 2004).

Dissolved Fe species were extracted from peat soil incubations following the protocol of Veverica et al. (2016). The method is based on an ionic liquid extraction using bis-2-ethylhexyl phosphoric acid (Pepper et al., 2010), which was shown to be more suitable for extraction of Fe from humic-rich matrices than the traditional ferrozine or phenanthroline methods. Briefly, 2.5 mL of soil slurry was filtered (0.2 µm nylon filter; Celltreat Scientific Products) and mixed with 7.5 mL of HCl (0.67 N) in an extraction vial in a 0.5 % H2 in N2 glove box. The O2 concentration in the glove box was continuously monitored and remained <10 ppm. To separate Fe3+ from Fe2+, 10 mL of 0.1 M bis-2-ethylhexyl phosphate (95 %, Alfa Aesar) in n-heptane (99.5 %, Acros Organics) was added to the acidified sample. Next, the organo-aqueous emulsion was shaken at 250 rpm in closed extraction vials for 2 h. The bis-2-ethylhexyl phosphate chelates Fe3+ more effectively than it chelates Fe2+. The Fe2+-containing aqueous phase was sampled into a 3-fold HCl-washed HDPE vial (Nalgene) in the glove box. The Fe3+ fraction chelated in the organic phase was then back-extracted into an aqueous phase by the addition of 10 mL 4 N HCl and shaking at 250 rpm in closed extraction vials for 20 min. Fe3+ and Fe2+ fractions were quantified separately in acidified aqueous solution by inductively coupled plasma–optical emission spectrometry (ICP-OES; Thermo iCAP6300 at the Goldwater Environmental Laboratory at Arizona State University). The ICP-OES pump rate for the Ar carrier was set to 50 rpm, and Fe2395 and Fe2599 lines were used for Fe quantification. Iron concentrations were determined from a calibration curve (0.01–10 mgL-1) by diluting a standard solution (100 mgL-1, VHG Labs, product no. SM75B-500) in 0.02 N HNO3.

Three-dimensional fluorescence analysis was performed on a Horiba Jobin-Yvon Fluoromax 4 spectrofluorometer. Excitation–emission matrices (EEMs) were generated by obtaining emission spectra (λEm=300–550 nm, at a step size of 2 nm) at excitation wavelengths from 240 to 450 nm at a 10 nm step size. All EEMs were blank-corrected and normalized daily to the Raman peak of ultrapure water (deionized, carbon-free,18.2 MΩcm; Barnstead^™ NanoPure). The samples were taken at the same time as those for Fe analysis. Prior to analysis, soil slurries were filtered using a solvent-rinsed Whatman Glass Microfiber Grade F GF/F filters (nominal pore size 0.7 µm) to obtain ∼10 mL filtrate. Samples were diluted with ultrapure water if their UV absorbance exceeded 0.3 so that inner-filter corrections could be made (Stedmon et al., 2003). We calculated total fluorescence as the matrix sum of all signals in the EEM. Fluorescence indices were used to characterize various classes of fluorophores in the dissolved organic matter (DOM) pool. Fluorescence index (FI) was calculated as the sum of the intensity signal in the emission spectra from 470 to 520 nm collected at an excitation wavelength of 370 nm (Cory and McKnight, 2005). Humification index (HIX) was determined from the peak area under the emission spectrum from 435 to 480 nm, divided by the area from 300 to 445 nm, both collected at an excitation wavelength of 254 nm (Ohno, 2002). The “freshness” was determined to be β/α, the ratio of emission intensity at 380 nm to the emission intensity maximum between 420 and 435 nm, both collected at an excitation wavelength of 310 nm (Wilson and Xenopoulos, 2009).

All basic statistical tests were performed with JMP Pro software (Version 13.1.0, SAS Institute Inc., Cary, NC, USA).

The tropical peat soil used for the incubation experiments had 5.5–5.8 pH, 92.2 % water content, 307±5 mgTOCg-1 dry weight, and 3.8±0.9 g total Fekg-1 soil. The extractable iron fraction partitioned as 54±3 µM extractable Fe3+ and 213±16 µM extractable Fe2+. The native soil pore water had 13.2±1.2 mgL-1 DOC, 436±79 µgNL-1NH4+, 9.7±1.3 µgNL-1NO3-, and 3.9±0.2 µgNL-1NO2-. Hydroxylamine was below detection in all cases (<3 µM). Soil pH dropped from 5.5 to 5.8 in untreated soil to 3.6, 4.8, 5.0, 5.2, and 5.4 after treatment with Hg, Zn, γ irradiation, autoclaving, and CHCl3, respectively. Only N3 treatment increased soil pH (to 6.4).

Live and dead dyes were used to assess microbial viability by means of membrane integrity, where a “dead” signal indicates disrupted or broken cell membranes (Stiefel et al., 2015). The majority (74 %) of cells in the live incubation displayed the “live” signal (Fig. 1). The CHCl3 and γ-irradiated treatments were most effective at reducing the number of viable cells (∼15 % intact membranes after sterilization). Chemical inhibitors (Hg, Zn, and N3) were less effective at killing cells (∼30 % intact membranes after sterilization). Autoclaved samples did not fluoresce, likely due to cell lysis during steam pressurization.

Biological denitrification activity was measured over 3 d in live and sterilized soils based on the difference in CO2 production with and without added NO3-. An efficient sterilization treatment would show no changes in CO2 beyond those due to equilibration between the gas phase and aqueous phase. Nitrate stimulated CO2 production in live soil (ANOVA, p<0.05) and not in the γ-irradiated, Zn, Hg, N3, or autoclaved incubations (Fig. 2), indicating that residual cells in the sterilized treatments were not capable of denitrification.

In general, sterilization increased extractable Fe2+ and Fe3+ relative to live controls (Fig. 3). This trend was particularly pronounced in Zn treatments, which had 9× higher extractable Fe2+ (1915±26 µM) and 1.6× higher extractable Fe3+ (87±3 µM) than live controls. The Hg treatment showed the second-largest increases. In the presence of NO2-, extractable Fe2+ decreased and extractable Fe3+ increased in live, Zn, and CHCl3-fumigated treatments, as expected if Fe2+ was oxidized by NO2- during chemodenitrification. However, autoclaving, γ irradiation, and N3 lowered Fe3+ concentrations, suggesting the influence of unknown concomitant reactions. For instance, autoclaving (largest drop in Fe3+) already showed lower Fe3+ concentrations after sterilization. Production of Fe3+-reduction artifacts in treatments could lead to Fe3+ depletion and, hence, mask increase in Fe3+ due to chemodenitrification. NO2- addition resulted in near-complete depletion of extractable Fe2+ in live, CHCl3-fumigated, and γ-irradiated soils. Changes in Fe speciation with other sterilants were more moderate. Minimal changes were observed for other metals in soil samples (e.g., Mn, Al, Cu, and Zn; data not shown).

Fluorescence analysis of soil extracts using excitation–emission matrices (EEMs) was used to evaluate changes in DOM-containing aromatic moieties or conjugated double bonds (Stedmon et al., 2003; Fig. 4). The N3 treatment was excluded from this analysis due to an interference with N3 absorbance that prevented inner-filter corrections from being made. The EEM signals showed the greatest change in the “humic” region (λEx<240–270 nm, and λEm=460–500 nm; Fellman et al., 2010), especially in Zn and Hg treatments, which significantly increased the FI from 1.20 (in live soil baseline, prior to NO2- incubation) to 1.49 (Table 1). Zn and Hg may elicit direct fluorescence quenching through the formation of Zn and Hg metal complexes (McKnight et al., 2001) or possibly due to indirect quenching by more highly dissolved Fe2+. Signal strength in the humic region was enhanced by NO2- addition in the live, CHCl3-fumigated, and γ-irradiated treatments. All five sterilization treatments had lower aromaticity (HIX) than live controls (Table 1). Autoclaved samples had 10-fold higher total fluorescence compared to live soils, suggesting that autoclaving degraded insoluble humics into more soluble and less condensed OM.

In the first 48 h, NO2- consumption rates were the highest in live soil (5.2 µMh-1), closely followed by irradiated samples (4.5 µMh-1, Fig. 5). The major chemodenitrification pathway for N2O formation was likely NO2- reduction by Fe2+, resulting in consumption of ∼1.5 µmolFe2+ and accumulation of ∼1.1 µmolFe3+ in the live control (Fig. 3). After 48 h, NO2- depletion continued to completion in the live control but slowed in all treatments other than the metal additions. After 72 h of incubation, 3 %–16 % of NO2--N was converted to N2O-N across treatments. N2O production rates were assessed by linear regression of data points over the whole duration of the experiment. Higher rates were observed in live, Zn2+, and N3 treatments (0.5–0.7 nmolN2Og-1h-1, r2>0.95) than in γ-irradiated, CHCl3-fumigated, autoclaved, and Hg treatments (0.1–0.2 nmolN2Og-1h-1, r2>0.9). Production rates within treatments showing high or low rates were not significantly different (Student's t, p>0.05), although comparisons across treatments with high or low rates were statistically different (Student's t, p<0.05). Thus, we identified a higher and lower group of sterilant-dependent N2O production rates from the same soil samples. The live control showed logarithmic N2O accumulation, while the sterilized treatments had linear accumulation over time; the latter is expected in abiotic accumulation (Fig. 5).

Similar NO2- consumption rates between live and irradiated treatments imply that NO2- depletion was dominated by abiotic processes over the first 48 h. In general, abiotic reactions tend to be linear processes, whereas microbially mediated reactions can be affected by enhanced expression of genes or cell reproduction in a nonlinear fashion (Duggleby, 1995). Linearity is more reflected in the N2O curve than in the NO2- curve. The difference in linearity of N2O production in sterilized vs. live treatments (Fig. 5) suggests that biological denitrification did not occur in sterilized soils.

Compared to our study, incubations of artificial media with 200 µMNO2-, 0.5–8.1 mMFe2+, and pH 7–8 had similar rates of Fe2+ depletion but 10× higher rates of NO2- reduction, and higher (∼10 %–50 %) N2O yields (Buchwald et al., 2016; Jones et al., 2015). In our peat incubations, reactive OM likely trapped NO2- in the soil matrix via OM-bound nitrosation reactions (Thorn and Mikita, 2000; Thorn et al., 2010) and the lower pH likely promoted conversion of NO2- to NO (Kappelmeyer et al., 2003; Porter, 1969) or N2 (Stevenson et al., 1970). Studies in low-pH northern temperate peat soils have shown the primary product of abiotic NO2- reduction was NO and not N2O (McKenney et al., 1990).

Azide and Zn exhibited enhanced NO2- conversion to N2O, at rates at least 2 to 5 times as high as those measured for the other sterilants (Fig. 5), likely due to higher pH and Fe availability, respectively. In the N3 treatments, elevated N2O production could be explained by the reaction of protonated NO2- with N3 in a pH-dependent manner (Stedman, 1959) and other changes in soil solution that originated from the increase in pH. Nitrite reaction with N3 has been characterized in marine and freshwater solutions reaching its maximum at pH 4.5 and proceeding slowly yet significantly (20 % conversion in 1 h) at pH >5 (McIlvin and Altabet, 2005), as in our slurries. Moreover, N3's self-fluorescence impeded OM measurements, making N3 an incompatible sterilizing agent for chemodenitrification studies. Zn increased Fe availability and may have increased NO2- affinity for reactive OM groups; both effects would lead to an abiotic increase in N2O production (Clark, 1962; McCalley and Sparks, 2009; Parton et al., 2007). Zinc treatment lowered the soil pH, which may have promoted cation displacement and stability of dissolved Fe2+ (Hutchins et al., 2007), thus enhancing N2O production. Several studies have used Zn treatments as a valuable agent for field applications (Babbin et al., 2015; Ostrom et al., 2016). Zn is less hazardous to humans than some of the other sterilants. We propose that the use of Zn could provide useful information about abiotic in situ rates as long as Zn-induced chemodenitrification is accounted for. A correction could be applied if a complementary laboratory assessment (using the more efficient γ irradiation) were used to develop an ecosystem-specific correction factor.

Divalent Hg2+ can be abiotically methylated by fulvic acid-type substances (Rogers, 1977). The reaction oxidizes OM and can diminish its reducing power, as indicated by decreased reactivity of humic acid with NO2- (Gu et al., 2011; Zheng et al., 2011), thus interfering with the abiotic assay. Given the pH effect of the Hg treatment, we cannot rule out that decomposition of nitrous acid (HNO2) contributed to NO2- consumption (Fig. 5, Park and Lee, 1988). Another potential factor associated with the Hg treatments is metal sorption. At low pH (3.6), 98 % of Hg was sorbed to humic acids, whereas only 29 % of Zn was sorbed at pH ∼4.8 (Kerndorff and Schnitzer, 1980). Full sorption capacity of peat is presumably reached in seconds (Bunzl et al., 1976), and the differing sorption behavior of Hg and Zn may play a role in the reaction potential of NO2- with OM. It has been demonstrated that Hg introduced into peat soil leads to sorption of Hg ions to various functional groups, including phenols (Drexel et al., 2002; Xia et al., 1998). Hence, it is plausible that Hg sorbed to functional groups subject to electrophilic attack by NO+ (e.g., nitrosophenol, Scheme 1) may hamper nitrosation, and therefore protect OM from reacting with NO2-. This could lead to a selective suppression of the OM-dependent N2O production pathway.

Chloroform fumigation resulted in potential N2O production rates within the lower production range treatments with minor differences in Fe speciation and DOM fluorescence. However, unlike the other sterilized samples, CHCl3-fumigated samples showed enhanced CO2 production stimulated by NO3- addition. Removal of CHCl3 from our samples before substrate addition could have provided an opportunity for a few surviving heterotrophs to regrow and use the easily degradable organic material derived from dead cells. Indeed, chloroform can lyse cells, providing substrates for growth to CHCl3-resistant microorganisms (Zelles et al., 1997). Continued exoenzyme activity has been also described as a CO2 source; however, this would not include denitrification enzymes, since no enzymes involved in the denitrification pathway are exoenzymes (Blankinship et al., 2014; Jenkinson and Powlson, 1976a). Chlorination of natural OM may prompt formation of quinones (Criquet et al., 2015), which are intermediates in the OM-based abiotic N2O production (Thorn and Mikita, 2000); indeed, regions of the EEMs corresponding to hydroquinones (Cory and McKnight, 2005) appear to be slightly higher in CHCl3 treatments. The benzene derivative produced during nitrosophenol reaction with NO2- leads to reduced π electron delocalization (Scheme 1). Because excitation of π electrons produces fluorescence, reactions with NO2- might be expected to reduce OM fluorescence. However, the experiment duration is important, and if microbial cells indeed reproduce after the treatment, short experimental periods (e.g., hours or days) or reapplication of CHCl3 might keep down the numbers of any potential denitrifiers improving the use of this method.

Autoclaved peat soil revealed abiotic N2O production rates close to the average of the lower production range group, along with ICP-OES and fluorescence spectroscopy results that showed significant changes in Fe speciation and DOM composition. EEMs demonstrate lower values for the HIX in autoclaved peats (Table 1), consistent with fluorescence data from a study that demonstrated a decrease in the aromaticity and polycondensation of soil extracts from autoclaved soil (Berns et al., 2008). Autoclaving likely caused degradation and solubilization of insoluble humic components. The direct effects of autoclaving are very much dependent on the heat and pressure stability of the indigenous soil constituents, but the substantial soil structural changes likely introduce chemical artifacts that are absent in the native live soil.

The fewest chemical artifacts were observed in γ-irradiated samples. Soil that had been exposed to γ rays showed the lowest N2O production rates, approximately one-fifth of those observed in live samples. Irradiation also caused only very small changes in Fe speciation relative to live controls and yielded EEMs that were remarkably similar to those obtained from live soil extracts. Our measurements of sterility and respiratory activity indicated the lowest potential for biological activity and hence the least amount of interference for the time period tested. We therefore confirmed γ irradiation to be a preferred method for sterilizing soil (Trevors, 1996) and for assessing abiotic N2O production potential. In practice, the long preparation time needed to reach a sufficient dose (dependent on radiation source; see Methods) was compensated for by the lack of chemical artifacts during the experiment and the reduced number of hazardous waste products. Limited accessibility to irradiation facilities and the absence of a field portable option remain the main challenges to wide distribution of this approach.