N2O formation by nitrite-induced (chemo)denitrification in coastal marine sediment

Nitrous oxide (N2O) is a potent greenhouse gas that also contributes to stratospheric ozone depletion. Besides microbial denitrification, abiotic nitrite reduction by Fe(II) (chemodenitrification) has the potential to be an important source of N2O. Here, using microcosms, we quantified N2O formation in coastal marine sediments under typical summer temperatures. Comparison between gamma-radiated and microbially-active microcosm experiments revealed that at least 15–25% of total N2O formation was caused by chemodenitrification, whereas 75–85% of total N2O was potentially produced by microbial N-transformation processes. An increase in (chemo)denitrification-based N2O formation and associated Fe(II) oxidation caused an upregulation of N2O reductase (typical nosZ) genes and a distinct community shift to potential Fe(III)-reducers (Arcobacter), Fe(II)-oxidizers (Sulfurimonas), and nitrate/nitrite-reducing microorganisms (Marinobacter). Our study suggests that chemodenitrification contributes substantially to N2O formation from marine sediments and significantly influences the N- and Fe-cycling microbial community.

Chemodenitrification could be driven by the presence of Fe(II) that is produced by heterotrophic Fe(III)-reducing microorganisms 23 , as well as by the availability of nitrite, that is produced during the reduction of nitrate by heterotrophic denitrifying bacteria 18 . The abiotic production of N 2 O via chemodenitrification has been documented in laboratory experiments via reactions involving intermediates such as hydroxylamine (NH 2 OH) and  . Hereafter, the term chemodenitrification refers to the abiotic reaction of Fe(II) and nitrite. The extent of N 2 O production via chemodenitrification versus denitrification is still poorly understood. Elevated levels of N 2 O have been observed in numerous studies examining iron-and nitrate-/nitrite-rich environments, e.g. soils 26 , hypersaline ponds and brines in Antarctica 27 , and marine coastal sediments [28][29][30] . However, these high levels of N 2 O have been solely attributed to microbial denitrification, potentially overlooking the important contribution of chemodenitrification to the overall N 2 O formation. Jones et al. 31 showed with a purely chemical laboratory setup that the co-presence of Fe 2+ and nitrite clearly stimulates chemodenitrification. They also provided an approach to distinguish between biotic versus abiotic contribution to N 2 O emission based on isotopic labelling. However, the actual role and potential of chemodenitrification in environmental systems remains unclear. The connection of the biogeochemical N and Fe cycle via chemodenitrification potentially impacts on the related microbial community. The production of Fe(III) minerals during chemodenitrification triggers heterotrophic Fe(III) reduction which supplies Fe 2+ , stimulating again chemodenitrification in the presence of nitrite, as well as microbial Fe(II) oxidation. Iron redox cycling is thus, strongly related to the biogeochemical N cycle. Based on microsensor measurements Wankel et al. 30 could show N 2 O formation at the interface of the nitrate and Fe(III) reduction zone within marine sediments. These authors hypothesized that chemodenitrification could play a major role in N 2 O production and hinted towards connections between the biogeochemical Fe and N cycle. Still, the actual contribution of chemodenitrification to N 2 O emission across the sediment-water interface remains unknown. Therefore, the focus of our study was to quantify the chemodenitrification-based N 2 O formation potential in natural marine sediments and to investigate the potential impact on the microbial community.
We hypothesize (i) that chemodenitrification plays an important role in marine sediments and that the abiotic oxidation of Fe(II) (provided by microbial Fe(III) reduction) by nitrite (formed during heterotrophic nitrate reduction) produces significant amounts of N 2 O, and (ii) that chemodenitrification may influence the N-and Fe-cycling microbial community in marine sediments.
Here we present incubation experiments with marine organic-rich sediment from the coastal area of the Baltic Sea, Norsminde Fjord, Denmark. Based on the knowledge gaps described above, the objectives of the present study were (i) to quantify N 2 O formation during chemodenitrification in microcosm studies, and (ii) to understand the consequences of chemodenitrifying conditions on the N-and Fe-cycling microbial community.
In our study, we found that up to 15-25% of total N 2 O production (range of three independent experiments) can be caused by chemodenitrification. This elevated N 2 O formation caused an increase of N 2 O reductase (nosZ) transcripts and an enrichment of potential Fe(II)-oxidizers and Fe(III)-reducers, as revealed by quantitative PCR and 16S rRNA (gene) amplicon sequencing, respectively. Our study demonstrates that chemodenitrification can contribute substantially to global N 2 O formation and significantly influences the N-and Fe-cycling microbial community in marine coastal sediments.

Results
Nitrite-induced (chemo)denitrification in coastal marine sediment. The following experiments were set up: amendment of (i) nitrate and Fe(II) (Fig. 1), (ii) nitrite and Fe(II) (Fig. 1), (iii) nitrite only (Fig. S1) to both microbially active and sterilized sediments at concentrations to quantify the maximum contribution of chemodenitrification to N 2 O formation. In addition, one setup contained only native sediment (sterile vs. microbial active) (Fig. S1). The different setups are summarized in Table S1. In microcosms containing natural sediments, nitrite and Fe(II) were continuously produced and consumed via microbial and abiotic processes, and their resulting steady-state concentrations were low. In sterilized microcosms nitrite and Fe(II) were not produced during the experiment (Fig. S1). When N substrates and Fe(II) were added to the microbially active sediment, a maximum of 2653 ± 787.0 ppm of N 2 O (nitrate addition) and 4950 ± 748.6 ppm (nitrite addition) was quantified after four days of incubation (Fig. 1). This yielded a maximum amount of 5.8 ± 1.7 ppm (with nitrate) and 10.8 ± 1.6 ppm (with nitrite) per g wet sediment per hour for the respective microcosm setup. Thus, a maximum amount of N 2 O of 0.013 ± 0.001 µmol g −1 h −1 and 0.36 ± 0.1 µg N kg −1 h −1 (nitrate addition) and 0.024 ± 0.001 µmol g −1 h −1 and 0.68 ± 0.1 µg N kg −1 h −1 (nitrite addition) was formed (for calculations see Wang et al. 26 ) (Fig. 2). The addition of Fe(II) and nitrate to sterilized sediment only showed a low production of 0.6 ± 0.03 ppm N 2 O per g wet sediment after four days of incubation, which relates to 1.3 ± 0.1 nmol g −1 h −1 . In contrast, the amendment of Fe(II) and nitrite to sterilized sediment revealed approximately three times more N 2 O production (1.9 ± 0.08 ppm N 2 O per g wet sediment per hour) which relates to 4.2 ± 0.1 nmol g −1 h −1 . Thus, a maximum of 17.4 ± 6.6% of total N 2 O (result of three independent microcosm experiments) was produced abiotically via chemodenitrification.
In nitrite-amended microcosms with microbially active sediment (without Fe(II) addition, Fig. S1), 1.87 ± 0.6 ppm N 2 O was formed per g wet sediment per hour, which relates to 4.2 ± 0.2 nmol N 2 O g −1 h −1 . For comparison, the sterile control setup that was amended with nitrite only produced 0.82 ± 0.4 ppm N 2 O per g wet sediment per hour, which relates to 2.4 ± 0.07 nmol g −1 h −1 . Almost no N 2 O was formed in native sterile and active microcosms within four days of incubation (<0.04 nmol N 2 O g −1 h −1 ) (Fig. S1).
To evaluate the effect of iron and nitrogen substrates on the formation of N 2 O, we quantified the extractable and dissolved Fe(II), as well as total Fe, nitrate, nitrite, and dissolved organic matter (DOC) over time (Fig. 1). Extractable Fe(II) was relatively constant at approximately 31.7 ± 1.5 μmol wet g −1 in the nitrate-/nitrite-and Fe(II)-amended (both microbially active and sterile setups) (Fig. 1). Whilst dissolved Fe(II) decreased in all www.nature.com/scientificreports www.nature.com/scientificreports/ setups containing both microbially active sediment and an amendment of nitrate or nitrite (from 1.6 ± 0.1 mM to 0.1 ± 0.02 mM and 0.2 ± 0.03 mM, respectively), dissolved Fe(II) decreased from 1.6 ± 0.1 to 0.8 ± 0.01 mM in sterilized sediment with nitrate addition and from 1.5 ± 0.1 to 0.3 ± 0.03 mM in sterilized sediment with nitrite addition. The total extractable Fe(II) values and Fe total values were both approx. 32.5 ± 0.3 μmol wet g −1 .
The nitrate concentration in microbially active sediment with nitrate addition decreased from 4.7 ± 0.02 mM to 2.2 ± 0.1 mM, probably due to nitrate reduction by denitrifying microorganisms (see Fig. S2; e.g. potential nitrate-reducers: Sulfurimonas and Desulfuromonadales). In the sterilized nitrate-amended sediments nitrate decreased to a much lower extent (from 4.7 ± 0.04 mM to 3.9 ± 0.04 mM). We observed a decrease of nitrite in nitrite-amended microbially active and sterilized setups (from 4.0 ± 0.04 mM to 2.7 ± 0.2 mM and to 3.5 ± 0.1 mM, respectively) and an increase of nitrite in nitrate-amended microbially active and in sterilized setups (from 0 to 0.9 ± 0.1 mM and 0.2 ± 0.01 mM, respectively). Furthermore, the DOC content in sterilized sediment increased significantly, probably as a result from the gamma-radiation treatment. In contrast, in microbially active nitrite and Fe(II)-amended microcosms, the DOC increased only slightly while DOC even decreased in microbially active setups that were amended with nitrate and Fe(II) (probably due to denitrification activity) (Figs 1 and 2).

Consequences of N 2 O formation and (chemo)denitrification for gene expression specific for the nitrogen cycle.
To evaluate the influence of (chemo)denitrification on active microbial nitrogen cycling, we followed gene and transcript copy numbers involved in the different denitrification steps (NO 2 − , NO, and N 2 O reduction) during sediment incubation (Fig. 3). This was used to determine the effect of Fe(II)-, nitrate-, and nitrite-amendment on the microbial formation and reduction of N 2 O. Independent of the amendment, the copy  www.nature.com/scientificreports www.nature.com/scientificreports/ numbers for nirK, qnorB, cnorB, and atypical nosZ (clade II nosZ) did not change in the microbially active microcosms during incubation. Only, typical nosZ (clade I nosZ) was substantially upregulated in the active sediments that were amended with nitrite and Fe(II), whereas it was below the detection limit in all other samples (Fig. 3).

Consequences of N 2 O formation by (chemo)denitrification on the Fe-and N-cycling microbial community.
In order to track the microorganisms that are involved in Fe-and N-cycling under denitrifying and chemodenitrifying conditions in our sediments, we analyzed the general microbial community in Fe(II)and nitrate-/nitrite-amended microcosms (Fig. S2). We found that the addition of Fe(II), nitrate, and nitrite to the marine sediment and the resulting N 2 O formation caused a general microbial community shift based on DNA (present community) and RNA (active community) analysis (Fig. S2). The overall trends for the changes in the present and active microorganisms during microcosm incubation were similar, i.e. similar taxonomic groups were enriched in the DNA-and RNA-based analysis. Here, we present data for the RNA-based analyses. DNA-based results are provided in the supplementary information (Fig. S4).
Psychrilyobacter is known as a psychrophilic Fe(III) reducer and was enriched in ferruginous marine sediments 38 . We identified abundant and active Marinobacter (22%) to be closely related to M. litoralis (100% 16S rRNA gene sequence similarity), to M. aquaeolei (99%; a potential neutrophilic Fe(II)-oxidizer), and to M. hydrocarbonoclasticus (99%). Relatives of M. hydrocarbonoclasticus were also identified recently based on nosZ gene sequence analysis in non-amended Norsminde Fjord sediment (Otte et al., unpublished). M. hydrocarbonoclasticus has the genetic potential for all nitrogen cycle enzymes and the potential for N 2 O formation 39 while a close relative, i.e. M. aquaeolei, even has the potential for Fe(II) oxidation 40 .
Evidence for the importance of chemodenitrification on N 2 O formation. To distinguish between N 2 O formation by abiotic chemodenitrification and by microbially catalyzed processes, we inactivated the microbial community in the sediment by gamma-radiation. Gamma sterilization destroys enzymes (RNAses) that are required for RNA degradation 41 . However, even in sterilized samples traces of RNA were found (Fig. 3), which might result from incomplete sterilization. Still, the sterilized samples can be considered as a valid control setup. In order to evaluate the nitrate-reducing Fe(II)-oxidizing and the Fe(III)-reducing potential of the remaining RNA in the sterilize sediments, participation in nitrogen-converting metabolisms was checked using the KEGG database (KEGG pathways). Based on a previous strategic study on the efficiency of sterilization we can still consider gamma-radiation as the most efficient sterilization method for sediment samples 41 .

Discussion
Our results show that chemodenitrification can account for up to 15-25% of the total N 2 O production in the marine sediments from Norsminde Fjord (Fig. 1). High N 2 O release in the presence of Fe(II) and nitrate/nitrite has been observed before in different environments 27,30,42 . For Norsminde Fjord previous studies reported N 2 O concentrations at the sediment-water/atmosphere interface of up to 0.49-4.9 µM 43-45 . Although, N 2 O emission rates are in a similarly high range compared to our data, the production of N 2 O has never been related to chemo-denitrification. The source of N 2 O formation has mainly been described to originate from heterotrophic denitrification or mixotrophic nitrate-dependent Fe(II) oxidation 26 . For the latter process, it has recently been shown that the oxidation of Fe(II) in the presence of nitrate and dissolved organic carbon, is a coupled abiotic-biotic reaction network (e.g. in Acidovorax sp. BoFeN1 12 ). This means that nitrate gets microbially (heterotrophic) reduced and intermediate reactive nitrogen species (e.g. nitrite) are involved in chemodenitrification which produces N 2 O upon the oxidation of Fe(II) 12 . Few studies hypothesized that chemodenitrification might play a much bigger role in the N 2 O production patterns in ecological systems than previously thought 25,30 . Wankel et al. 30 showed that N 2 O increases in stratified marine coastal sediments as soon as oxygen is consumed along the redox gradient. These results in combination with our data imply the importance of N 2 O production due to chemodenitrification.
Anoxic conditions influence the expression of genes involved in the reduction of N oxides (Figs 3 and 4). Although the presence of elevated N 2 O concentrations was thought not to upregulate the expression of nosZ genes (within clade I of nosZ-harboring bacteria) 46 , recently, Harter et al. 47 suggested that the increase of typical nosZ expression under nitrate-reducing conditions might be due to locally enriched N 2 O. In our study, 46% more N 2 O was formed (after 91.5 hours of incubation) in microbially active microcosms that were amended with Fe(II) and nitrite (Fig. 1), compared to the microcosms that were amended with Fe(II) and nitrate only. In these microcosms elevated expression of typical nosZ was detected (Fig. 3). High N 2 O concentrations, as well as high nitrite levels, were described to have a toxic effect on the microorganisms, which might trigger the expression of typical nosZ genes as a detoxification mechanism 48-51 . The measured N 2 O and nitrite levels in our experiments only represent the net concentrations, and potentially the total N 2 O formed is underestimated. Based on thermodynamic considerations, the reduction of N 2 O to N 2 provides a higher energy yield than for the other denitrification steps 48,52 . Therefore, microorganisms that are capable of N 2 O reduction (e.g. Shewanella spp., Marinobacter spp., and Pseudomonas spp., which were enriched and active in our microcosms (Fig. S3)), might gain an energetic advantage over species performing the full denitrification pathway (with nirK/S, nosZ) 53,54 . This advantage is only available as long as anoxic and no sulfidic conditions are prevailing, as the presence of sulfide and oxygen can inhibit the expression of nosZ genes and the concurrent reduction of N 2 O to N 2 . Recently, a second clade of nosZ (atypical nosZ; clade II nosZ) was discovered and atypical nosZ were shown to be expressed in non-denitrifying N 2 O-reducing microorganisms 49,54 . Hallin et al. 49 proposed that clade II bacteria have a N 2 O respiratory chain that allows more efficient free energy conservation compared to the clade I system 49 . The energetic benefit from N 2 O reduction might explain the high expression of atypical nosZ in all setups. However, only bacteria with the full denitrification capacity are upregulating typical nosZ in the setup where high N 2 O concentrations (due to chemodenitrification) accumulate to cytotoxic levels 47 .
Our data show that nitrite-induced chemodenitrification has substantial consequences for the active N-and Fe-cycling microbial community (based on 16S rRNA gene sequences) (Figs 4 and S2, S3). Community members that carry the potential to trigger heterotrophic Fe(III) and nitrate reduction were enriched and active (Fig. S3), which is also supported by the slight decrease of DOC in nitrate and Fe(II) amended microcosms (Fig. 1). The stimulating effects of chemodenitrification on active Fe-cycling microorganisms (based on 16S rRNA gene sequences) were obvious by an increase of potential Fe(III)-reducers (Arcobacter, Sulfospirillum, and Shewanella in nitrate-amended setups, and Psychrilyobacter in nitrite-amended setups). In addition, an increase in the relative abundance of active Fe(II)-oxidizing bacteria [Sulfurimonas (with nitrate amendment), and Marinobacter (2019) 9:10691 | https://doi.org/10.1038/s41598-019-47172-x www.nature.com/scientificreports www.nature.com/scientificreports/ and Pseudomonas (with nitrite amendment)] has been confirmed by 16S rRNA gene sequence analysis (Fig. S3)). Several Marinobacter species have been found in Fe-rich habitats 55 , and their metabolic potential to utilize nitrate as terminal electron acceptors and iron (i.e. FeS 2 and CuFeS 2 ) as an electron donor has been demonstrated previously 40,56 . Potential Fe(III)-reducers (e.g. Desulfobulbus, Desulfomusa, Sulfospirillum, Shewanella) were more abundant and active than bacteria that might be enrolled in Fe(II) oxidation. These results are in line with previous studies that quantified significantly more Fe(III)-reducing bacteria (up to 2.8%) compared to Fe(II)-oxidizing bacteria (in particular nitrate-reducing Fe(II)-oxidizers with 0.3%) in the same sediment 57,58 . Potential denitrifying bacteria with typical nosZ such as Shewanella spp. (which are metabolically flexible, i.e. they can use either Fe(III), nitrate or nitrite, and were detected in nitrate-amended microcosms), Pseudomonas spp. and Marinobacter spp. (in nitrite-amended microcosms) were stimulated by nitrate/nitrite addition and chemodenitrification reactions (Fig. S3, selection of typical nosZ bacteria from Norsminde Fjord sediment).
The surviving community (Xanthomondales clade JTB255, Candidatus Isobeggiatoa divolgata, Oceanospirillales, Sandaracinaceae, and Bacteroidetes at t 0 ) and enriched microorganisms (e.g. Synechococcus and Cyanobium sp. at t end ) in the gamma-sterilized microcosms (analyzed at both t 0 and t end ) (Fig. S2) have the ability of assimilatory nitrate reduction (reduction of nitrate to nitrite) but no genetic potential (e.g. no nirK/S and no norB genes) for further nitrite reduction. In addition, we did not see an increase of nitrite (Fig. 1) in microbially inactivated sediment. We can therefore conclude that all N 2 O produced in gamma-radiated sediments can be attributed to an abiotic mechanism, namely chemodenitrification.
In this study we observed that up to 15-25% of total N 2 O production might be caused by chemodenitrification, whereas 75-85% of N 2 O was produced by denitrification and other microbial processes. 1.2 mM dissolved Fe(II) was consumed in microcosms with sterilized sediment, that were amended with 2 mM Fe(II) and 4 mM NO 2 − (Fig. 1). Following the stoichiometry of Equation 1, 1.2 mM Fe(II) can abiotically reduce 0.6 mM NO 2 − . The measured decrease in NO 2 − was approx. 0.5 mM NO 2 − (Fig. 1), and we can therefore consider chemodenitrification as the main process in the microcosm setup with sterilized marine sediment. In the microbially active setups 1.4 mM Fe(II) and 1.3 mM NO 2 − was consumed. Based on the stoichiometry of chemodenitrification the consumption of only 0.7 mM NO 2 − can be attributed to this process. The consumption of the remaining 0.7 mM NO 2 − can be related to various microbial processes in the sediment. However, transferring our data to environmental systems we need to keep in mind several rate limiting factors for chemodenitrification. Compared to in situ conditions our setups do not suffer from substrate limitations, i.e. Fe 2+ and nitrite production rates. Also, Fe-organic matter complexes, as they potentially occur in nature, will influence the reaction kinetics and the fate of Fe and N species 59 . In addition to that, in nature other processes, Figure 4. Cause and effects of nitrite-and nitrate-induced (chemo)denitrification-based N 2 O formation in coastal marine sediment. When Fe(II) and nitrite/nitrate was added to the marine sediment, Fe(III) was formed among different process by chemodenitrification which is stimulating Fe(III)-reducing bacteria (FeRed). The FeRed bacteria then produce Fe(II) and stimulate Fe(II)-oxidizing bacteria. Therefore chemodenitrification has a significant impact on Fe-cycling in general. Simultaneously, nitrite/nitrate was reduced to NO and further to N 2 O by chemodenitrification and denitrifying bacteria. In addition, the presence of high nitrite/nitrate concentration leads to a high typical nosZ gene expression in denitrifying bacteria which is responsible for the reduction of N 2 O into N 2 .
www.nature.com/scientificreports www.nature.com/scientificreports/ such as anammox, DNRA, and nitrate reduction by sulfur oxidation will impact the N 2 O emission rate 60,61 . Finally, the incubation temperature of 25 °C represents conditions that can be reached in summer months, especially when the overlying water column is very shallow 62,63 . Although these conditions do not represent average field site conditions over the whole year, they represent the environmental circumstances at which maximum N 2 O emission via chemodenitrification can be expected.
Coastal marine environments have been recognized as an important source of N 2 O to the atmosphere 28,29 . Sandy sediments, such as the one we used for our study, are representative for about 70% of all global shelf sediments 64 and are expected to release up to 1.9 Tg N 2 O-N yr −1 28 . Assuming a contribution of 15-25% N 2 O release by chemodentrification (data from this study) shelf sediments might account for 0.3-0.5 Tg N 2 O-N yr −1 abiotically produced N 2 O. In the future the N 2 O release might even increase due to higher fertilizing activities in close-by agricultural fields. Leaching and runoff transport of nitrate into oceans leads to coastal marine eutrophication 65 and increases the potential for additional N 2 O production by (chemo)denitrification. There is a high chance that similarly high N 2 O emission rates that have been observed in many other environments, including soils 26 and sediments [28][29][30] , are triggered by chemodenitrification.

Methods
Site description and sampling procedure. Marine sediment samples were taken in February 2015 April 2016, and March 2017 from the Aarhus Bay area (Denmark) (geochemical characterization is described in Laufer et al. 58 and Table 1). Sediment from the shallow marine estuary Norsminde Fjord (NS) was collected at 0.5 m water depth near its narrow entrance from Aarhus Bay (N 56°01.171′; E 010°15.390′). Bulk sediment from the upper 3 cm were collected, transported to the laboratory at 4 °C, and the sediment was stored for a few weeks at 4 °C until microcosm experiments were set up.
Geochemical analyses. Temperature, pH, salinity, and oxygen concentration of the water column were analyzed in the field with a multimeter (WTW, Multi 3430). The geochemical parameters of sediment water content, DOC, and nitrate concentrations in porewater as well as Fe(II) and Fe(III) concentrations in porewater and sediment were determined in the laboratory. All analyses are described in detail in Laufer et al. 58 .
Microcosm experiment. Bulk sediment from Norsminde Fjord was homogenized before setting up microcosm incubations. Microcosm incubations were set up in 100 ml serum vials that were wrapped in aluminum foil for dark incubations. Five g of homogenized bulk sediment and 50 ml anoxic filtered seawater were used for each microcosm. The headspace of the microcosms was N 2 /CO 2 (90:10). For preparation of the media, seawater was made anoxic by flushing with N 2 for at least 1 h per liter, filtering it through a 0.22 µm filter (EMD Millipore Steritop TM ) and replacing the headspace by N 2 /CO 2 (90:10), followed by adding 20 mM NaHCO 3 as buffer (final pH of the microcosms: 7.2 ± 0.1). To inhibit the activity of sulfate-reducing bacteria (and therefore inhibit reactions of sulfur species with Fe(II)) an anoxic and sterile filtered 1 M Na 2 MoO 4 solution was added to a final concentration of 20 mM 58 Table S1.
Abiotic control microcosms with gamma-sterilized sediment were incubated under the same conditions as the microbially active ones. Sediment for gamma-sterilization (sterile microcosm experiments) was filled into plastic bags, sent to Synergy Health Allershausen, Germany, and radiated at 52 kGy with 5% radiation tolerance. All microcosm experiments were setup in triplicates and incubated in dark at 25 °C. The chosen temperature represents conditions at which the observed processes will have a maximum environmental impact. Although the yearly average temperature in the field is significantly lower, the chosen temperature will be reached in summer months, especially when the water column is very shallow 62,63 . www.nature.com/scientificreports www.nature.com/scientificreports/ Quantification of Fe(II), nitrate, nitrite, and DOC. For measuring Fe(II) and extractable Fe, 1 ml of slurry from each microcosm was sampled with a syringe inside an anoxic glovebox (100% N 2 atmosphere). 100 µl of this slurry sample was added to 900 µl 40 mM sulfamic acid in 1 M HCl and placed on a shaker (150 rpm) for 1 h (see Laufer et al. 58 ). The samples were then centrifuged (5 min, 7000 g, Eppendorf 5430R) and the supernatant was used for Fe(II) and extractable Fe determination with the spectrophotometric Ferrozine assay 21 . The remaining part of the initial 1 ml sample of the slurry was centrifuged and the supernatant was used for analyses of the dissolved phase. The Fe-extraction with sulfamic acid 21 avoids the abiotic oxidation of Fe(II) by nitrite during acidic Fe-extraction. For nitrate and nitrite, 100 µl of supernatant was added to 900 µl Milli-Q H 2 O and stored anoxically at 4 °C until analysis by flow injection analysis (FIA) 58 . For analysis of DOC in microcosm incubations, 20 ml of sample was necessary 58 and the contents of sacrificial microcosms were centrifuged for 15 min at 5000 g (Hermle 7300 Germany). Afterwards, the supernatant was filtered through a 0.45 µm filter (MF-Millipore MCE membrane) and the DOC concentration was measured with a carbon analyzer (High TOC, Elementar, Germany).
Nitrous oxide quantification. For the determination of N 2 O concentrations, the pressure of the microcosm bottles was brought to normal pressure (1 bar) before sampling via a water trap. Headspace gas samples of 0.5 ml were taken from the microcosms 0, 15, 39, 63, and 87 hours after the beginning of the incubation and transferred into 22.5 ml evacuated gas chromatograph (GC) vials using a gas-tight syringe (1100TLL 100 ml Gastight, Hamilton, Reno, NV, USA). The trace gas concentrations in the GC vials (22.5 ml) were measured using a GC equipped with an electron capture detector ( 63 Ni-ECD) for N 2 O and CO 2 (Hewlett Packard, 5890 Series II). The GC setup and configuration was described in detail previously 66 . Gas fluxes were calculated using the slope of the temporal change in concentration of the closed microcosms according to the equations published in Ruser et al. 66 . The result of 17% of total N 2 O by chemodenitrification is an average of three independent experiments and we show results of only one experiment.
DNA and RNA extraction, DNA digestion, reverse transcription, and amplification. Total DNA and RNA was extracted using the PowerSoil ® RNA and DNA isolation kit as directed by the manufacturer (MO BIO Laboratories, Carlsbad, CA, USA), with the following modifications: 0.8 g to 2 g sediment was used from each sediment slice; 5 min bead-beating; centrifugation steps at maximal speed (7000 g) at 4 °C; and a longer incubation times at −20 °C (1.5-2.0 hours). RNA and DNA were eluted in 50 µl 10 mM Tris buffer. DNA and RNA concentrations were determined using a Qubit ® 2.0 Fluorometer with DNA and RNA HS kits (Life Technologies, Carlsbad, CA, USA). RNA extracts were digested with the Ambion Turbo DNA-free TM kit as directed by the manufacturer (Life technologies, Carlsbad, CA, USA). Successful DNA removal was confirmed by PCR using general bacterial primers (see Supplementary Information). Subsequent reverse transcription reactions were done using a reverse transcriptase (Invitrogen, Life Technologies) as described in the supporting information. Microbial 16S rRNA genes were amplified using primers 515F and 806R (see Supplementary Information) targeting the V4 region. Quality and quantity of the purified amplicons were determined using agarose gel electrophoresis and Nanodrop (NanoDrop 1000, Thermo Scientific, Waltham, MA, USA). Subsequent library preparation steps and sequencing were performed by Microsynth AG (Balgach, Switzerland). A sequence library was prepared and sequence adapters added using the Nextera kit. Sequencing was performed on an Illumina MiSeq sequencing system (Illumina, San Diego, CA, USA) using the 2 × 250 bp MiSeq Reagent Kit v2 (500 cycles kit) (Illumina, San Diego, CA, USA). The MiSeq reporter software v2.6 (Illumina, San Diego, CA, USA) was used for signal processing, de-multiplexing, and trimming of adapter sequences. The quality of the reads was checked with the software FastQC version 0.11.5 and the primers trimmed using cutadapt v1.14. Amplicon reads (accession number: SRP132652) have been deposited in the NCBI Genbank database (bioproject: PRJNA431287).

Sequence analysis.
Demultiplexed and trimmed reads were further analyzed using QIIME (v1.9.1) (Caporaso reference). Paired end reads were joined using default settings and were further quality filtered and only those with a minimum Phred quality score of Q20 were used. Chimeric sequences were identified using usearch6.1 67 and removed. OTUs were picked using the QIIME workflow script "pick_de_novo_otus.py" and taxonomically identified at 97% similarity using the SILVA 128 reference database 68 . Singletons were removed from the OTU table prior to further analysis. We used the quality-filtered reads for the graphs (Figs S2 and S3) and checked also with <0.1% filtered OTU tables (data not shown). Rarefaction curves, diversity indices (Shannon diversity, Simpson diversity), richness (Chao1, ACE), and coverage estimators (Good's coverage) were calculated using QIIME workflow scripts.
Quantitative PCR of bacterial 16S rRNA genes and nitrogen cycle genes. Quantification PCR (qPCR) specific for phylogenetic and functional marker genes [16S rRNA gene (Bacteria), amoA (Archaea), nirK (nitrite reductase), qnorB and cnorB (NO reductases), typical nosZ (clade I N 2 O reductase in denitrifiers) and atypical nosZ (clade II N 2 O reductase in non-denitrifiers)] was carried out using the SsoFast EvaGreen Supermix (Bio-Rad Laboratories, Hercules, CA, USA), an iQ5 real-time PCR detection system (iQ5 optical system software, version 2.0, Bio-Rad Laboratories), and gene-specific primers. For details on plasmid standards, gene-specific qPCR primers, reaction mixtures and thermal programs, please refer to Table S2 in the Supplementary Information.

Data Availability
The microcosm datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.