Evidence for fungal and chemodenitrification based N2O flux from nitrogen impacted coastal sediments

Although increasing atmospheric nitrous oxide (N2O) has been linked to nitrogen loading, predicting emissions remains difficult, in part due to challenges in disentangling diverse N2O production pathways. As coastal ecosystems are especially impacted by elevated nitrogen, we investigated controls on N2O production mechanisms in intertidal sediments using novel isotopic approaches and microsensors in flow-through incubations. Here we show that during incubations with elevated nitrate, increased N2O fluxes are not mediated by direct bacterial activity, but instead are largely catalysed by fungal denitrification and/or abiotic reactions (e.g., chemodenitrification). Results of these incubations shed new light on nitrogen cycling complexity and possible factors underlying variability of N2O fluxes, driven in part by fungal respiration and/or iron redox cycling. As both processes exhibit N2O yields typically far greater than direct bacterial production, these results emphasize their possibly substantial, yet widely overlooked, role in N2O fluxes, especially in redox-dynamic sediments of coastal ecosystems.

N itrogen (N) loading from anthropogenic activities profoundly impacts ecosystems worldwide, with loading to coastal zones among the largest challenges facing humanity, as nearly half the global population lives within 100 km of the coast. Coastal sediments are known hotspots of biogeochemical transformations and recognized as effective agents for removing excess nitrogen 1,2 . However, biological removal of reactive nitrogen may also occur at the expense of increased production of nitrous oxide (N 2 O), a potent climatically active trace gas. Despite being the largest ozone depleting substance currently emitted to the atmosphere 3 , N 2 O remains unregulated by the international community and large uncertainties exist concerning N 2 O budgets (4100%), particularly for heterogeneous environments such as coasts 4,5 . Redox-dynamic estuarine and coastal sediments routinely experience high N loading and low dissolved oxygen (O 2 ), conditions that are strongly linked to elevated N 2 O and underlie their estimated 10% contribution to the global N 2 O flux 4,6-9 . Thus, understanding their role in both nitrogen removal and N 2 O production is important for improving predictions of long-term impacts of human activity across globally relevant scales.
Many studies have focused on the relative contribution of bacterial denitrification (bDNF) or nitrification (oxidation of ammonia (NH 3 ) to nitrite (NO 2 À ) or 'AMO'), as controlling processes underlying N 2 O emissions ( Fig. 1). While yields of N 2 O from both AMO and bDNF are low (o1% in terms of total moles of N converted), their magnitude and ubiquity across ecosystems translates into major atmospheric fluxes. Increasingly, however, the potential for other N 2 O production processes has become apparent, including production of N 2 O by fungi and/or abiotic reactions coupled to redox cycling of metals such as iron [10][11][12] . In particular, the organic-rich and redox-dynamic regimes of estuarine and coastal sediments may promote both fungal activity and rapid redox cycling of iron. To examine controls and mechanisms of N 2 O production in coastal sediments (Fig. 1), we incubated natural sediment cores under flow-through conditions, manipulating both dissolved O 2 and nitrate in the overlying water (using conditions typifying anthropogenically impacted ecosystems), while monitoring both porewater N 2 O profiles and stable isotopic fluxes of ammonium, nitrate, nitrite and N 2 O. Given the complexity of processes involved, we also leveraged the use of a less-traditional isotope system ( 17 O, described below) to provide even broader perspective for disentangling operative N 2 O cycling mechanisms.
The steady-state emission flux of N 2 O (F N2O ) is governed by six possible production fluxes (F) ( Fig. 1; bacterial denitrification (bDNF), fungal denitrification (fDNF), chemodenitrification (cDNF; specifically the abiotic reduction of NO 2 À to N 2 O by Fe(II)), ammonia oxidation by bacteria (bAMO) or archaea (aAMO) and nitrifier-denitrification (nDNF)), as well as respiratory consumption by denitrifying bacteria (N 2 O RED ) such that: Stable isotopes of N 2 O have been widely used for studying its production and consumption, including both oxygen ( 18 O/ 16 O) and bulk nitrogen ( 15 N/ 14 N) (d ¼ (((R sample /R standard ) À 1) Â 1,000) and R ¼ 15 N/ 14 N or 18 O/ 16 O) [13][14][15][16][17][18][19] . In addition, the unique intra-molecular distribution of 15 N within N 2 O molecules has emerged as a powerful tool for constraining N 2 O cycling, as differences in 15 N content between the central 'a' and outer 'b' atoms of the N 2 O molecule ('site preference' or SP N2O , where SP N2O ¼ d 15 N a À d 15 N b ) have been shown to reflect formation pathways 13,14,16,18 . Numerous studies have measured the steady-state d 15 N offset between precursor molecules (NO 3 À , NO 2 À and NH 4 þ ) and N 2 O ('Dd 15 N' ¼ d 15 N source À d 15 N N2Obulk ), as well as SP N2O values towards characterizing signature compositions for the processes in equation 1. Some compositional overlap notwithstanding, the isotopic separation of many of these endmembers has been used to distinguish their relative contribution to N 2 O production, especially nitrification and denitrification 14,19,20 . Notably, the respiratory reduction of N 2 O by denitrifying bacteria can increase the d 15 N bulk of the remaining N 2 O (decreasing Dd 15 N values), as well as SP N2O values (with a distinctive relationship between isotope effects imparted on the d 15 N of the bulk N 2 O and its site preference, 15 e bulk and SP e, respectively 21,22 ) modifying mixing relationships. Nevertheless, measured values of Dd 15 N and SP N2O are quantitative integrators of the proportion of each process (which may then be modified by N 2 O reduction; see Methods), serving as independently responsive tracers for constraining production mechanisms. To further interrogate N 2 O dynamics, we also used a natural atmospherically derived NO 3 À having an unusual triple oxygen isotopic composition (containing excess 17 O) that provides novel 'isotope space' for further resolving co-occurring processes. By comparing levels of 17 O-excess within discrete N pools (see Methods), we are able to independently quantify the proportion of N 2 O deriving from NO 3 À (or NO 2 À ) and thereby, in concert with the more conventional isotopic measurements (Dd 15 N, SP), uniquely explain variations in N 2 O producing pathways. Below we summarize our results, combining microprofile perspectives with the use of these mass and isotopic fluxes to constrain N 2 O production mechanisms in coastal sediments of the Wadden Sea under a variety of incubation conditions ( Supplementary Fig. 1 Specifically, we find that the isotopic composition of increased N 2 O fluxes resulting from elevated nitrate loading in our incubations requires substantial contribution by processes not regularly considered in coastal ecosystems, namely fungal and/or chemodenitrification. We suggest that variations in the contribution of these processes to N 2 O fluxes from coastal and other ecosystems may help to explain the notorious variability that is frequently encountered in studies of N 2 O dynamics.  (Fig. 3), a common characteristic of organic-rich sediments 26   Triple oxygen isotopes as a tool for constraining nitrogen cycling.

Microsensor
Our multi-pool D 17 O measurements enable disentangling of processes that are otherwise overlapping (in SP N2O values, for example), providing a complementary perspective to the N isotope analyses. First, these analyses revealed that nitrification played a relatively small role in NO 3 À production. As noted, amended nitrate had a high D 17 O NO3 value ( þ 18.5%), which when combined with preexisting nitrate in the supply seawater (D 17  À precursor under high NO 3 À incubations and thus indicates that the increased production occurred via reductive pathways (Fig. 4) and not by a/bAMO. Together with the elevated SP N2O and only small changes in Dd 15 N, this suggests fungal and/or chemodenitrification as possible contributors (Fig. 4). Both fDNF and cDNF are dependent on supply of NO 2 À and typically exhibit yields far greater than bacterial N 2 O production (that is, the relative amount of N 2 O emitted per mole of NO 3 À or NO 2 À reduced or NH 3 oxidized). Thus, only small levels of these processes would be required to contribute relatively large amounts of N 2 O-setting the stage for a potentially important role for these biogeochemical processes in regulating N 2 O fluxes wherever they occur.

Discussion
Diversity and abundance of fungi in oxygen-depleted coastal sediments is generally thought to represent a small fraction of their soil-hosted counterparts [30][31][32] . Nevertheless, their ecological role remains unclear-with recent studies challenging the perspective that fungi are only ecologically significant under aerobic conditions [31][32][33] . Adapted for organic-rich environments often depleted in O 2 , many fungi have a range of cellular adaptations to life under suboxic conditions 32,34-36 , including the ability to couple denitrification 37 directly to mitochondrial    respiration 38 -a metabolic capacity that has been documented in a variety of environments including coastal sediments 34,39 . Given this respiratory flexibility, fDNF is poised to be especially important under hypoxic conditions and wetland environments, where access to O 2 in overlying water is juxtaposed with anoxic, carbon-rich conditions 40 . The most characteristic feature of the fungal-denitrifying system is a P450 cytochrome operating as a nitric oxide reductase (P450nor) 41 giving rise to the characteristically high SP of B35-37% (refs 42,43), the biochemical nature of which was recently interrogated in the purified enzyme 44 . Although assessment of fungal metabolic activity was beyond our scope, sequencing of the fungal ITS region revealed the presence of fungi across all incubations and study sites ( Supplementary Fig. 3 38 , yields from fDNF are also typically 1-2 orders of magnitude greater than for bDNF (generally o0.1%) and N 2 O production appears to be physiologically widespread among fungi 37 . Indeed the importance of fungi in contributing to N 2 O production is well-recognized across a range of terrestrial ecosystems 45 . While their overall role in the reductive elimination of reactive N may be small relative to that of bacterial denitrification, these high yields mean that even small levels of fDNF could have disproportionately large impacts on N 2 O release, serving as an important, yet under-recognized source to the atmosphere. Although biological N 2 O production has received much attention, abiotic production of N 2 O is also widely documented, typically via reactions involving intermediates such as NH 2 OH and NO 2 À -though its environmental role remains unclear (Zhu-Barker et al. 11 , and references therein). Specifically, reduced iron (Fe(II)), especially mineral or surface-bound Fe(II), is an effective catalyst of NO 2 À reduction under a range of conditions 46 , and the presence of mineral surfaces and elevated levels of Fe(II) has also been shown to increase N 2 O yield 47,48 . Although data are limited, SP N2O from chemodenitrification is generally 410% and recent evidence suggests that elevated reaction rates, promoted by high levels of Fe(II), may also increase SP N2O (up to 26%) ( 47-51 ). The production of reactive Fe(II) as the result of direct or indirect microbial activity is a ubiquitous feature of marine sediments. Our sites contained between 67 and 1344 mM HCl-extractable Fe(II) g À 1 wet sediment ( Supplementary Fig. 4). However, prediction of reaction kinetics between Fe(II) and NO 2 À in these porewater environments is complex, particularly given the range of binding environments of Fe(II), which largely controls its reactivity 52 yet also one yielding an elevated site preference such as chemodenitrification and/or fungal denitrification.  with the porewater Fe(II) levels suggests that cDNF may also have contributed to N 2 O production. Interestingly, however, despite its lower Fe(II), the sandy site (SD) exhibited similar overall N 2 O isotope dynamics to the other two more Fe-rich sites ( Table 2), suggesting that perhaps the increased response of N 2 O production to NO 3 À loading was perhaps not as tightly linked to Fe(II) content.
On the basis of the isotope systematics described, we use an isotope mass balance (based on equations 1, 2, 5 and 6 and defined endmember compositions (Supplementary Table 3)) to estimate relative contribution of operative N 2 O producing mechanisms (see Methods). While fDNF and cDNF are not mutually exclusive, we consider them separately to more robustly evaluate their potential contribution. Previous studies appear to demonstrate a strong relative dominance of ammonia oxidizing bacterial abundance compared to archaea in organic-rich coastal sediments 54,55 . An assumed numerical dominance of bacterial ammonia oxidizers notwithstanding, pure culture studies of archaeal ammonia oxidizers typically produce N 2 O reflecting a isotopic compositional mixture of both the AMO and nDNF pathways [56][57][58] , as has been more directly characterized in bacterial ammonia oxidizers 59 . Ongoing studies of N 2 O production mechanisms in ammonia oxidizing archaea will undoubtedly provide more insight on their unique biochemical nature. Differences in biochemistry aside, however, given the apparent similarity in isotopic composition of N 2 O deriving from bAMO and aAMO (especially a high SP N2O value, Figs 3 and 4), here we opt to combine bacterial and archaeal AMO for consideration in our mass balance analysis-setting endmember values to those previously determined for bAMO, as these have been studied in far more detail 59 Table 6). In contrast, ammonia-oxidation contributed on average only 3-12% (via NH 2 OH decomposition) and 8-17% (via nDNF) for the base case. Consideration of cDNF (in lieu of fDNF) as the endmember  Table 4) were considered equal to zero for this figure, with other values weighted accordingly to sum to 100%. The reader is also referred to Supplementary Table 4 for associated error in mass balance estimates calculated by Monte Carlo simulation, as well as Supplementary Tables 5 and 6 for sensitivity analyses of alternative endmember compositions.
having both a high SP N2O and a NO 2 À precursor required an even higher proportion of this process to satisfy mass balance (Supplementary Table 4). However, two cores in this case exhibited isotopic compositions violating mass balance (those with highest SP N2O ), evidently requiring at least some contribution of fDNF (having a higher endmember SP N2O ) over cDNF. Although the LN and LOLN treatments did not involve the D 17 O approach, the statistically higher SP N2O values under elevated nitrate (relative to low nitrate; Fig. 3, Supplementary Table 2) point to a shift in N 2 O production mechanisms in response to NO 3 À , which must have included increased contribution by fDNF and/or cDNF. Ultimately, while the precise contribution of N 2 O pathways varies depending on prescribed endmember compositions, all scenarios indicated substantial contribution by these non-traditional N 2 O production pathways.
Increased N 2 O emissions from coastal systems receiving elevated NO 3 À are well documented 4,8,9 and the 'central role' of NO 2 À in relation to N 2 O has been emphasized by others 6 . For example, large increases in N 2 O from sediments amended with NO 2 À (relative to NO 3 À ) was previously interpreted as evidence for 'obligate nitrite-denitrifying bacteria' that reduce NO 2 À to N 2 O (ref. 6). Similarly, based on SP N2O it was concluded that N 2 O production in estuarine sediments was controlled by an as yet 'unidentified process' 60 having an isotopic composition consistent with more recent studies of fungal and chemodenitrification. On the basis of our results, we suggest that these previously 'missing' and/or 'unidentified' pathways likely represent non-traditional pathways including denitrification catalysed either by fungi or reactions involving Fe(II).
To the degree that our sediment incubations reflect processes ongoing under natural conditions, elevated NO 3 À loading to coastal sediments appears to increase N 2 O fluxes largely through reactions involving a NO 2 À intermediate, yet also exhibiting elevated SP values. This combination of characteristics pinpoints an increased involvement of processes not regularly considered in coastal ecosystems-namely fungal and chemodenitrification. We suggest that both may represent important, yet under-appreciated sources regulating N 2 O fluxes from redox-dynamic, organic-rich environments and warrant further examination. Studies are frequently challenged by the dynamic nature of N 2 O fluxes, which are often episodic and difficult to link to specific factors or processes (for example, refs 23,25). Although our study was conducted at steadystate (enabling our assessment of fDNF and cDNF), we posit that the commonly observed patchy and dynamic nature of N 2 O fluxes may stem from a complex network of differential contribution by direct and indirect, biological and abiotic processes, including the metabolic activity of fungi and biogeochemical redox cycling of iron. In particular, compared to bacterial denitrification and/or ammonia oxidation, their especially high yields poise these processes to be important, yet under-recognized, contributors to N 2 O dynamics in many systems.

Methods
Study site and experimental setup. Twenty-four sediment cores were collected in August of 2013, from three intertidal sites near Königshafen on the island of Sylt in the North Sea, Germany. Sites were B100 m apart and chosen based on qualitative differences in sediment grain size and location characteristics. The 'Schlickwatt (MD)' and 'Mischwatt (MX)' sites were located inside a small lagoon, while the 'Sandwatt (SD)' site was more openly exposed to wind and waves ( Supplementary Fig. 1). Thirty intact push cores (30 cm length, 10 cm OD, 1/8 00 wall thickness) were taken using polycarbonate core liners having vertical lines of silicone sealed holes (ø 3 mm) at 1-cm intervals to allow porewater collection using Rhizon samplers. Cores were retrieved leaving B10 cm of overlying water and sealed with double o-ring Delrin caps to minimize gas exchange during transport, and brought immediately back to the laboratory. In addition to the cores used for the incubations, two additional cores were used from each site for immediate microsensor profiling (O 2 , N 2 O) and pore-water extraction ('field cores'). The remaining cores were prepared in parallel for incubations. On completion of the incubations, microsensor profiling of O 2 and N 2 O was conducted immediately followed by extraction of porewaters.
Incubations. The gas-tight sealed sediment cores were incubated in the dark at in situ temperatures (19°C) in a temperature-controlled room at the Alfred Wegener Institute-Waddensee Field Station. Throughout the incubations the overlying water of the cores was continuously supplied with filtered seawater from large carboys, which were refilled as needed. The o-ring sealed core tops contained inlet/outlet fittings for continual delievery of fresh seawater through gas impermeable PEEK tubing (1/8 00 OD). Peristaltic pumps were used to regulate flow rates at 1.8 ± 0.06 ml min À 1 (measured gravimetrically at each sampling point) for B8 days. The inflow line was placed near the sediment-water interface to minimized stratification. For experimental manipulations, four different inflow seawater compositions were used: 'Low nitrate' (air sparged; B20 mM; LN), 'Low oxygen, low nitrate' (sparged with N 2 to 30-35% O 2 saturation; B20 mM; LOLN), 'High nitrate' (amended with NaNO 3 to B120 mM (above background nitrate); HN) and 'low oxygen, high nitrate' (combined treatments; LOHN).
Sample collection. Samples of each sediment core effluent were taken twice per day. For dissolved ions, effluent was directed into HDPE bottles and allowed to fill for B60 min before subsampling, filtering (0.2 mm) and freezing ( À 20°C). Separate 20 ml aliquots were taken for measurement of dissolved inorganic nitrogen concentrations (nitrate, nitrite and ammonium) and stable isotopic composition. Concentrations of nitrite and ammonium were made immediately (see below), while nitrate concentrations were measured later in the Wankel lab at WHOI. Samples for dissolved N 2 O were directed through gas impermeable PEEK tubing directly into pre-evacuated Tedlar gas sampling bags followed by gentle transfer into 160 ml serum bottles using a 1 4 00 OD silicone tubing, filling from the bottom to minimize turbulence and gas exchange. Sample water was allowed to overflow the bottle volume for at least two volumes before crimp-sealing with grey butyl septa and preserving with 100 ml of a saturated HgCl 2 solution.
Porewater sampling. Pore water samples were collected from sediment cores in 1-cm depth intervals using Rhizons 61 , which were inserted into intact sediment cores through silicon-filled ports in the walls of the core tubes. Samples of 5-10 ml volume were taken starting at the sediment-water interface down to 16 cm depth and frozen immediately for later analysis. Parallel cores were sectioned in 1-cm intervals for the analyses of iron. HCl extractable Fe(II) and the amorphous, poorly crystalline fraction of the Fe(III) minerals were measured by procedures described in ref. 63, with the modifications as in ref. 64.
Concentration and flux measurements of N bearing species. Concentrations of NO 3 À þ NO 2 À were measured by chemiluminescence after reduction in a hot acidic vanadyl sulfate solution on a NOx analyser 65 . Concentrations of NO 2 À were quantified by using the Griess-Ilosvay method followed by measuring absorption 540 nm, and NO 3 À was quantified by difference 66 . Concentrations of NH 4 þ were measured by fluorescence using the OPA method 62 . Concentrations of N 2 O were made using the integrated peak area of the m/z 44 beam on the IRMS (see below), standardizing to analyses of known amounts of N 2 O (injected into N 2 sparged seawater in 160 ml serum bottles) and normalizing to sample volume (158 ml).   74 . Parallel conversions of internal nitrite standards (WILIS 10, 11 and 20) were conducted to assess potential changes in reagents with time. Internal nitrite standards were also used correct for any variations due to peak size linearity and instrumental drift, with a typical reproducibility for both d 15  þ concentrations, the DON flux was generally of the same magnitude as the NH 4 þ flux (not shown). For use in the mass balance calculations (for estimation of the bAMO endmember Dd 15 N value), the d 15 N of the TRN pool was assumed to be a reasonable proxy for the d 15 N of the NH 4 þ pool. In general, this assumption had only a very small impact on the apportionment N 2 O sources by mass balance (o1%).
Nitrous oxide. For dissolved N 2 O, samples were extracted from the 160 ml serum bottles using a purge and trap approach 78 . Liquid samples were quantitatively transferred from the sample bottle into a purging flask using a 20 psi He stream, followed by He-sparging (B45 min) and cryogenic trapping using the same system described above for nitrate and nitrite derived N 2 O. Isotopic composition of the dissolved N 2 O was measured by direct comparison against the N 2 O reference tank, as no isotopic reference materials were available at the time of the analyses (USGS 51 and USGS 52 have since been publicly released: http:// isotopes.usgs.gov/lab/referencematerials.html). The composition of this tank (d 15 Isotope mass balance approach. Isotope mass balance calculations were made for estimating the relative contribution of N 2 O production pathways in the sediment incubations (denitrification by bacteria (bDNF), by fungi (fDNF), or by chemodenitrification (cDNF), as well as combined production by ammonia oxidizing bacteria and archaea via NH 2 OH decomposition (AMO) or nitrifier denitrification (nDNF)). By combining four independent mass balance expressions (equations 5, 6, 9 and 11 below) we can solve for the contribution of four independent N 2 O production processes (here we describe consideration of fDNF (case 1), with cDNF being considered separately (case 2)). Equation 1 can be expressed in terms of the fractional contribution (f) of each production process to the total flux of N 2 O: , incorporating the D 17 O measurements, is used but neglecting cDNF for this case: As the isotopic composition of N 2 O from bAMO and aAMO are very similar in the context of the isotope space evaluated here [56][57][58] , we choose to combine these terms into a single term (AMO), having the composition of bAMO (Supplementary  Table 3). Together with the fact that we will treat cDNF separately, equation 7 thus simplifies to: Similar to equation 7, the fractional contribution of each process to the measured SP N2O of the effluent can be expressed as: where f denotes the fractional contribution of a given process having a particular SP value, and where f(N 2 O RED ) is equal to 1 À (F N2O /(F bDNF þ F fDNF þ F cDNF þ F bAMO þ F aAMO þ F nDNF )) and SP e N2ORED is the kinetic isotope effect on SP for N 2 O reduction of À 6% (refs 21,22). As in equation 9, consideration of four processes simplifies equation 10 to: By combining equations 5, 6, 9 and 11-we can uniquely solve for the fractional contribution of four processes (bDNF, fDNF, AMO and nDNF) to the total observed N 2 O fluxes of the core incubations ( Fig. 5 Error propagation and sensitivity analysis. Error estimates for these mass balance calculations (Supplementary Tables 4,5 and 6) were calculated using a Monte Carlo error propagation approach in R (with 10,000 simulations), in which randomized Gaussian distributions of values were generated, as defined by their mean and s.d. given in Supplementary Table 3. This approach takes into account both the error associated with measurement of steady-state isotopic compositions (Supplementary Table 2; which implicitly incorporates both analytical error during instrument measurement, as well as natural variability during operation of the incubations) as well as error associated with the definition of endmembers (Supplementary Table 3).
Beyond these estimates of error, we also evaluated the sensitivity of specific endmember values to the calculated mass balance estimates-focusing on variations for endmember values having the least amount of certainty. First, the endmember values of Dd 15 N for bDNF and fDNF (and cDNF) are prescribed to be low as an inferred consequence of diffusion-limited expression of intrinsic (enzyme level) isotope effects (Supplementary Table 3). For nDNF, we chose to use a Dd 15 N value established by a study of nitrifiers under varying oxygen tension (Frame and Casciotti 59 ), as nitrifiers will be growing at the sharp oxic/anoxic interface in our sediment core incubations. Whether their supply of NO 2 À (as a substrate for nDNF) could be considered to be limited by diffusion is perhaps a matter of debate. However, we reasoned that nitrifiers will most generally be denitrifying the product NO 2 À that they themselves are producing (for detoxification) and therefore would not be limited by diffusion of NO 2 À from anoxic depths below. Further, the positive flux of NO 2 À out of the sediments-also indicates that the diffusive supply of NO 2 À should not have been limited (regardless of the source of the NO 2 À ). Nevertheless, we discretely examined the impact that these assumptions make by decreasing the endmember Dd 15 N value for nDNF from 56.9 to 28% and then to 14%. Under these scenarios (assuming 10% N 2 O reduction, for example)-the average relative contribution of nDNF increases from 8 to 14 and 26%, respectively, though mostly at the expense of bDNF, which decreases on average from 45 to 38 and 25%, respectively. In comparison, these scenarios actually increase estimated contribution of fDNF from 36 to 43 and 56%, respectively (Supplementary Table 5).
Endmember SP values have been generally well established for bDNF, nDNF and bAMO through culture studies under a variety of growth conditions. SP for fungal DNF is admittedly less well studied, however, several studies have shown that SP values are universally elevated (often clustering B35-37%). Recent studies 37,42 observe that most N 2 O producing fungal cultures yielded SP values 430%. Finally, even the purified N 2 O producing fungal enzyme (p450nor) has been shown to exhibit elevated SP values, albeit at somewhat lower values (15-29%). Notably, our choice of þ 37% for the fDNF endmember is conservative in estimation of the relative contribution of fDNF. For example, decreasing this value to the mean reported by Maeda et al. 37 of þ 30.3 ± 4.8% (which notably also contained some questionably low values), results in an average of a 6% increase in the contribution of fDNF to N 2 O production, with the average contribution of 36% shifting up to an average of 41-42% (Supplementary Table 6).
Finally, we evaluated our assumption that the extracellular NO 2 À pool could be disregarded as a reactant source for nDNF (that in situ, nDNF only occurred from NO 2 À supplied via bAMO), by setting the prescribed D 17 O-N 2 O/D 17 O-NO 2 À value for nDNF to a value of '1' instead of '0'-representing the most extreme case. Indeed, under this scenario-the average estimated contribution of bNDF and nDNF actually do not change by 4B1%, while bAMO is increased (from 12 to 19%) and fDNF is decreased (from 36 to 28%) (Supplementary Table 6).
Fungal genetic sequencing. Sediment samples for fungal sequence analysis were collected and stored frozen at À 80°C. Genomic DNA from marine sediment samples was extracted using a bead beating protocol according to the manufacturer's instruction (Mo Bio, Carlsbad, CA). ITS region sequences were amplified using the fungal ITS primer pair F (ITS5): GGAAGTAAAAGTCGTAAC AAGG and R (ITS4): TCCTCCGCTTATTGATATGC generating fragments of B600 bps in length. PCR products were cloned using Zero Blunt TOPO PCR Cloning (Thermo Fisher, Carlsbad, CA). After a ligate buffer exchange the plasmid was transferred into TOP10 electrocompetent cells. Cells were plated and grown on LB agar containing kanamycin. Single colonies were recovered from each plate and amplified using M13F&R primers. The products were sequenced by Sanger method (EtonBio, San Diego, CA). Sequences were analysed using BLAST. Taxonomy was assigned for fungal sequences (Supplementary Figure S3) by comparison against untrimmed ITS in the UNITE database (01/08/2015 version), using QIIME v1.91. Sequences were assigned only if the database match had a similarity of at least 90% and maximum e-value of 0.001.
Data availability. The data sets generated during this study are available by request from corresponding author.