Oxygen and light determine the pathways of nitrate reduction in a highly saline lake

Nitrate (NO3) removal from aquatic ecosystems involves several microbially mediated processes including denitrification, dissimilatory nitrate reduction to ammonium (DNRA), and anaerobic ammonium oxidation (anammox) 20 regulated by slight changes in environmental gradients. Saline lakes are prone to the accumulation of anthropogenic contaminants, making them highly vulnerable environments to NO3 pollution. We investigated nitrate removal pathways in mesocosm experiments using lacustrine, undisturbed, organic-rich sediments from Pétrola Lake (Spain), a highly saline waterbody subject to anthropogenic NO3 pollution. We used the revised N-isotope pairing technique (N-IPT) to determine NO3 sink processes. Our results demonstrate the coexistence of denitrification, DNRA, and anammox processes, 25 and their contribution was determined by environmental conditions (oxygen and light). DNRA and N2O-denitrification were the dominant nitrogen (N) removal pathways when oxygen and/or light were present (up to 82%). In contrast, anoxia and darkness promoted NO3 reduction by DNRA (52%) and N loss by anammox (28%). Our results highlight the role of coupled DNRA-anammox, as yet has never been investigated in hypersaline lake ecosystems. We conclude that anoxia and darkness favored DNRA and anammox processes over denitrification and therefore reduce N2O emissions to the atmosphere. 30

N budget, significantly increasing the global fixed N pool and nitrous oxide (N2O) emissions to the atmosphere (Canfield et al., 2010). Among N species, nitrate (NO3 − ) is a widespread compound responsible for water degradation due to excessive 35 fertilizer use in agriculture (Spalding and Exner, 1993). NO3 − accumulation can increase primary production in surface waters and, as a consequence, can trigger oxygen deficiency and promote eutrophication of surface waterbodies (Vitousek et al., 1997). Among aquatic ecosystems, saline lakes are highly vulnerable to NO3 − pollution. These ecosystems are mainly located in closed hydrological systems from arid and semi-arid regions, which combined with low precipitation and high evaporation rates typical of arid climates, lead to accumulate and biomagnify many pollutants compared to freshwater 40 systems (Williams, 2002).
Microbial processes are controlling the Earth's N cycle for ~2.7 billion years and have been widely studied in aquatic ecosystems. Certain microorganisms (diazotrophs) are able to fix N2 into a biologically useful form. In shallow aquatic ecosystems, cyanobacteria often dominate benthic N2 fixation, which is tightly linked with light availability and photosynthesis (Lu et al., 2018). Moreover, the role of cyanobacterial blooms altering N attenuation within aquatic 45 ecosystems has been recently reported (Zilius et al., 2018). In contrast to diazotrophs, microbes performing denitrification and anaerobic ammonium oxidation (anammox) are responsible for removing fixed N by producing N2 gas (Canfield et al., 2010).
Denitrification is considered to be the primary process of NO3 − removal from aquatic environments whose main end product is dinitrogen gas (hereafter referred to as N2-denitrification) through a multi-step reduction process (Harrison et al., 2009;50 Fernandes et al., 2016;Kuypers et al., 2018). N2O is an obligate intermediate of denitrification which can be also its main end product (hereafter referred to as N2O-denitrification) and N2O comprises an important atmospheric greenhouse gas (310 times more potent than carbon dioxide) (Trogler, 1999). However, the nitrification process (NH4 + oxidation to NO3 -) is also a significant source of atmospheric N2O as a by-product (Bremner and Blackmer, 1978;Dore et al., 1998;Löscher et al., 2012). A large diversity and potential activity of denitrifying bacteria have been previously shown for saline lakes (Kulp et 55 al., 2007;Lipsewers et al., 2016), as well as the existence of denitrification at the field scale (Doi et al., 2004;Gómez-Alday et al., 2014). In such ecosystems, variable redox conditions and the supply of organic matter (OM) and nutrients lead to increased N2O production by denitrification (Huttunen et al., 2003;Liu et al., 2015). In fact, N2O reduction to N2 seems to be a rate-limiting step during denitrification at extremely high salinities (Shapovalova et al., 2008). Denitrification is strongly affected by oxygen availability. Despite nitrous oxide reductase activity has been always considered to be inhibited already 60 at relatively low oxygen concentrations (0.25 mg/L) (Bonin and Gilewicz, 1991), recent studies showed that Clade II N2O reductases (NosZ), abundant in many biomes, are highly expressed and able to reduce N2O in the presence of low O2 concentrations (Yoon et al., 2016;Hallin et al., 2018).
Under anaerobic conditions, anammox couples ammonium (NH4 + ) oxidation to nitrite (NO2 − ) reduction to produce N2 (Van de Graaf et al., 1995). The activity of anammox bacteria has been described in marine ecosystems (Thamdrup and Dalsgaard,65 systems (Yang et al., 2012;Lipsewers et al., 2016), with a totally different community structure than described in freshwater lakes (Wang et al., 2015). So far, however, very little attention has been paid to the role of anammox processes in saline lakes. 70 When oxygen becomes limiting or unavailable, specific microorganisms can perform dissimilatory nitrate reduction to ammonium (DNRA) to obtain energy, retaining a more assimilable N source in the system. Thereby, by supplying NH4 + DNRA can promote anammox (coupled DNRA-anammox) . In coastal marine environments, DNRA may be as important as, or even more important than, denitrification for NO3 − reduction (Giblin et al., 2013), even causing eutrophication of coastal ecosystems during warm periods (Bernard et al., 2015). DNRA is promoted when NO3 − availability 75 is low or limited regarding available electron donors (Tiedje, 1988;Van den Berg et al., 2015). Therefore, light can impact on coupled DNRA-anammox as light will enhance primary production and the production of dissolved oxygen, which can strongly reduce the coupled DNRA-anammox process. However, the role of coupled DNRA-anammox in saline lakes has not yet been investigated.
In addition, sediment core incubations have been frequently used to quantify denitrification, DNRA, and anammox rates by 80 applying the 15 N isotope pairing technique ( 15 N-IPT) (Risgaard-Petersen et al., 2003;Roland et al., 2017). The 15 N-IPT was firstly applied on sediment cores to quantify N2 production derived from denitrification (Nielsen, 1992). Since then, many studies have focused on discriminating the relative contribution of N processes using 15 N-IPT, including DNRA, and recently coupled DNRA-anammox (Risgaard-Petersen et al., 2003;Holtappels et al., 2011;Hsu and Kao, 2013;Robertson et al., 2019). Hence, processes such as anammox have been traditionally underestimated. Recently, a set of equations for 15 N-85 IPT have been provided allowing to estimate the contribution of nitrous oxide production by N2O-denitrification and the contribution of DNRA to NO3 − reduction Salk et al., 2017). Prior to this revised methodology, coupled DNRA-anammox was indistinguishable from denitrification based on isotope tracer experiments (Francis et al., 2007).
Henceforth, these process estimates and the new IPT approaches are necessary for a complete N balance estimation.
Here, we tested the hypothesis that oxygen and light conditions in the water column can alter the balance between nitrate 90 removal pathways (denitrification, DNRA, and anammox) during sediment incubations. For this purpose, we incubated lacustrine sediments from a eutrophic lake (Pétrola Lake, Spain) and applied the revised 15 N-IPT to confirm and quantify Ncycling rates.

Study site 95
Samples were collected from Pétrola Lake (38º 50' 14'' N,1º 33' 40'' W), 35 km southwest of Albacete, Spain. Pétrola Lake is the main wetland in the endorheic Pétrola-Corral-Rubio-La Higuera Saline Complex, which covers a total area of about 275 km 2 and is located in the southeastern Castilla-La Mancha Region, in the High Segura river basin. Pétrola Lake is a terminal lake occupying about 1.76 km 2 in the lowest topographic position of the Pétrola endorheic basin (43 km 2 ), which https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License.
consists mainly of Mesozoic materials (Valiente et al., 2017). The endorheic basin is located in a zone vulnerable to 100 eutrophication, and fertilizer use is restricted (Order 2011/7/2 CMA). The lake is shallow (maximum depth 2 m) with major water volume oscillations depending on seasonal precipitation. The hydrofacies varies between Mg-Cl-SO4 (early spring) and Mg-Na-Cl-SO4 (early fall). The piezometric level of the aquifer is close to the topographic surface. Accordingly, various small springs and streams drain the aquifer and discharge into the lake following a radial pattern. The lake has been classified as a heavily modified water body due to the inputs of agricultural pollutants as well as untreated wastewater 105 directly spilled from Pétrola Village. The Pétrola endorheic basin was declared vulnerable to NO3 − pollution by the Regional Government of Castilla-La Mancha in 1998. Excess of N in the lake-aquifer system is mainly derived from inorganic synthetic fertilizers (Valiente et al., 2018).
The field survey was conducted in July 2015. The sampling site (control point 2651; Valiente et al., 2018) was located close to the lake's depocenter and is not affected by direct inputs of polluted freshwater streams or wastewaters. To evaluate initial 110 natural conditions (NC), water samples were collected and stored at 4 °C in darkness prior to further analyses. Furthermore, sediment cores (n=3) were taken from the upper 20 cm lacustrine sediment using acrylic coring tubes (5 cm inner diameter, 20 cm length). Coring tubes were capped at top and bottom with silicone rubber stoppers, cooled, and transported to the laboratory. Once there, the top 5 cm of each core was sliced and used for inorganic N-species extraction. Afterwards, these slices were then frozen at -20 ºC for further analysis. 115 Mesocosm preparation for core incubations was adapted from previous works (Welti et al., 2012). For this purpose, acrylic mesocosms (40 cm in length, 20 cm in diameter, containing a total volume of 12.6 L) were used for sampling and incubation to guarantee minimal disturbance of the sediment during sampling (n=9). The mesocosm tubes were acid-prewashed and then drilled into the sediment down to approximately 20 cm depth. Then, mesocosms were filled with 2 L of lake water to maintain sediment saturation during transport. Additional lake water was collected from the sampling point and stored at 4 120 ºC to fill the mesocosms in the lab. Black plastic sheets were used to cover the mesocosms to prevent light penetration during transport.

Sediment incubations
In the lab, the remaining volume of each mesocosm was filled with lake water (approximately 4 L), avoiding an air space over the water column. All mesocosms were sealed gas-tight. The bottom of each mesocosm was firmly sealed with a plate 125 bolted to the mesocosm tube. The top screw cap was equipped with two holes: one to inject air or argon, and the other to collect samples from the water column. To maintain oxic conditions in the water column, air was bubbled using an air compressor connected to a Teflon tube (7 mm inner diameter). From this tube, branches were connected to each mesocosm using a rubber cap. The end of each tube was assembled with a glass Pasteur pipette and placed in the upper water column.
For argon bubbling, a 25 L gas cylinder (Carburos Metalicos, Spain) with a pressure regulator was used. The argon 130 connection to each mesocosm was similar to that used for air bubbling.
For sample collection, a Teflon tube (4 mm inner diameter) was installed through the hole using a rubber cap. The tube inlet was placed 1 cm over the sediment surface, whereas the tube outlet was closed to the atmosphere with a three-way valve. In order to maintain water circulation inside each mesocosm, a small pump was installed in the inner wall. Mesocosms were placed in a temperature-controlled room (25 ºC) with no exposure to direct sunlight. 135 Three different treatments were studied in triplicate. Treatment 1 (OL; oxygen + light) mimicked field conditions by means of atmospheric air bubbling, to provide oxygen, and normal dark-light cycles (~ 14 h of light per day; no additional light source was used). Mesocosms of treatment 1 (n=3) were placed close to the room window. OL is henceforth considered as control. For treatment 2 (OD; oxygen + darkness), oxic conditions in the water column were preserved via atmospheric air bubbling. However, each mesocosm was covered with aluminum foil to protect it from light. Finally, treatment 3 (AD; 140 anoxia + darkness) maintained anoxic conditions by bubbling argon and mesocosms were shielded from light. The bubbling fluxes applied in the experiments were established based on the maximum solubility values of N2 (Hamme and Emerson, 2004) and N2O (Weiss and Price, 1980) in seawater using a salinity value of 50 g/L, similar to the one previously reported in Pétrola Lake (Valiente et al., 2018). Mesocosms were stabilized in the laboratory until constant N-NO3 − and N-NO2 − concentrations in the water column were reached. During the stabilization period, physico-chemical parameters, and 145 inorganic N-species were monitored at 12 h intervals, starting 12 h after collection of the sediment cores (time -36), and finishing 48 h after field sampling (time 0).
In order to apply the 15 N-IPT approach to quantify NO3 − transformation processes inside the mesocosms, 15 N-labeled nitrate (K 15 NO3) was added once mesocosm stabilization was reached (time 0). This involved spiking with 250 µmol of 15 N-NO3, reaching a water column concentration of about 40 µM N-NO3 − . After labeled nitrate addition, the sampling frequency and 150 incubation times were calculated following the NICE handbook (Dalsgaard et al., 2000). Thus, 30 min intervals were adopted as the initial sampling rate: this was calculated as the optimal time to enable denitrification to reach 90% of its steady state value, assuming a sediment penetration depth of oxygen of 1 mm based on previous works (Valiente et al., 2017).

Physico-chemical analyses
Physico-chemical parameters measured included temperature, pH, electrical conductivity (EC), total dissolved solids (TDS), redox potential (Eh), and dissolved oxygen (DO). These parameters were determined directly in the surface water from site https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License. 2651 using a HQ40d Portable Multi-Parameter Meter (Hach Company, USA). During sediment incubations, physico-165 chemical parameters were measured in the collected water samples. Collected water samples were immediately filtered through a 0.45 μm nylon Millipore® filter. Inorganic N-species were determined directly after collection at the Institute for Regional Development (University of Castilla-La Mancha). Determination of NO2 − and NO3 − concentration was achieved by UV-VIS spectrophotometry via the modified Griess reaction assay as described by García-Robledo et al. (2014). NH4 + concentrations were quantified by UV-VIS spectrophotometry using the modified indophenol method, as described by Institute of Inorganic Chemistry of the University of Vienna, as described by Nelson and Sommers (1996).

Isotope composition of N species 180
The isotopic composition of N-NH4 + in the water column was determined by a microdiffusion method . Sample aliquots (10 mL) were transferred to 20 mL HDPE vials with pre-weighed MgO (100 mg). Then, acid traps were added. They trapped the ammonia gas produced from ammonium when hydration of MgO increased the pH to >9.5.
Acid traps consisted of glass fiber filter discs (5 mm in diameter) placed on a strip of Teflon tape; 5 µL 2.5 M KHSO4 was pipetted onto the filter discs and the Teflon tapes folded and closed. Microdiffusion vials were then closed and placed on an 185 orbital shaker at room temperature for 2 days. Subsequently, each acid trap was transferred into a 1.5 mL reaction tube.
Tubes were placed into a desiccator containing concentrated H2SO4 for at least 24 h until further processing. To measure the isotopic composition of N-NO3 − , nitrate was isolated from the previously microdiffused extracts by a reductionmicrodiffusion method after conversion by Devarda's alloy to N-NH4 + (Prommer et al., 2014). The recovery efficiency of the conversion was expected to be ≥ 95% (Sørensen and Jensen, 1991;Mulvaney et al., 1997). Immediately after adding 190 Devarda's alloy, a new acid trap was added to each vial to retain ammonia gas deriving from nitrate reduction; the subsequent procedure was the same as described above. The filter discs from the acid traps were finally transferred into tin capsules and directly analyzed for N content and at % 15 N by EA-IRMS using an elemental analyzer (EA 1110, CE Instruments) connected via a ConFlo III interface (Thermo Fisher) to a DELTA plus IRMS (Finnigan MAT) in the SILVER
reactions. Each vial was completely filled with water sample avoiding any gas headspace. All vials were stored and shipped to the Mediterranean Institute of Oceanography (Aix-Marseille Université) for the analysis of N2 ( 29 N2 and 30 N2) and N2O isotopic species concentrations ( 44 N2O, 45 N2O, and 46 N2O) using GC-MS (Stevens et al., 1993). Dissolved N2 and N2O were 200 extracted from the samples in the Exetainer vials by introducing a 6 mL He headspace while simultaneously removing 6 mL of sample water. Sample injection was performed using a modified head-space autosampler (TriPlus 300, Thermo Fisher) that involves gas-equilibration at 65 °C for 6 min whilst shaking vigorously, so that more than 98% of the N2 and N2O equilibrium concentration was attained (Weiss, 1970). GC-MS analysis was performed using an Interscience Compact GC system equipped with AS9-HC and AG9-HCT columns. The GC conditions were as follows: injector temperature, 140 °C; 205 oven temperature, 60 °C; carrier gas flow rate, He 15 mL/min; interface temperature, 60 °C. The mass spectrometer was used in electron ionization mode, with an electron energy of 70 eV. Data were acquired in full-scan (m/z 2-200) and selected ion monitoring (SIM) mode (m/z 29, 30 monitored for N2; m/z 44, 45, 46 monitored for N2O). Ar (m/z = 40) was used as an internal standard. Data were acquired and analyzed using Excalibur software. Quantification of N isotopes in both gases was performed at the Mediterranean Institute of Oceanography (Aix-Marseille Université). Finally, isotopic mass balance 210 calculations were performed using discrete time points compared to the originally added amount of 15 N-NO3. Starting from the initial amount spiked (250 µmol K 15 NO3), N concentrations and atom percent enrichments were used to calculate the percentage of 15 N recovery in specific N forms and overall.

Denitrification, DNRA, and anammox activity measurements
For 15 N-IPT modeling, the revised 15 N-IPT calculation procedure (Salk et al., 2017) was applied. A detailed description of 215 parameters and equations is included in Table 1. For this purpose, our incubations were assumed to be intact core incubations. The probabilities of NO3 − reduction via denitrification, DNRA, and anammox were assumed to be equal . Genuine N2 production via denitrification (D14) and anammox (A14), as well as N2O production via denitrification, were calculated for each time step. Production rates were calculated according to Salk et al. (2017) for each time point after the addition of the labelled 15 NO3 − . Non-linear increments in the 15 N content were taken into account by 220 calculating the N production rates (i.e. 15 NH4 + , 29 N2, 30 N2, 45 N2O, 46 N2O) from the slope of the initial time point and each specific time point rather than a slope of all time points. Thus, a total of 20 rates of each process were calculated for each mesocosm. Ratios of 14 NO3 − : 15 NO3 − (r14) and 14 NH4 + : 15 NH4 + (r14a) were calculated and used as base parameters for activity calculations. The applied methodology allowed distinguishing between N2 production via coupled DNRA-anammox and via canonical anammox. DNRA rates were calculated using the production of 15 NH4 + , and of 30 N2 for anammox, over time. 225 However, this model cannot discriminate between 15 NO3 − assimilation and subsequent remineralization of OM to 15 NH4 + , and DNRA. Thus, the DNRA rate may include both processes. The sum of N2 production by denitrification and anammox, together with N2O production via denitrification, is designated as 'Total N loss'. The 'Total NO3 − reduction' adds the DNRA rate to the previous estimate.

Statistical analysis 230
Changes in chemistry and rates of N-loss processes over time as well as at the end of the incubation were assessed using oneway analysis of variance (ANOVA), followed by the Tukey's post hoc test (homogeneous variances) or by the Games-Howell post hoc test (heterogeneous variances). To assess differences in the hydrochemical conditions between initial (n=1) and final conditions (n=9), one-sample two-tailed t-tests were used. Results of statistical tests were considered to be significant at the confidence level 95% (α = 0.05). All tests were performed using SPSS-IBM Statistics software. 235

Differences between treatments in chemical parameters and rates of N-loss processes
Differences between initial (NC-48, time -48 h) and final conditions (EX72, time 72 h) were assessed for the three treatment groups (Table 2). Salinity, as TDS values, was around the hypersaline limit (50 g/L), with values ranging from 45. Regarding N-loss processes, mean (± standard deviation) and maximum rates are presented in Table 3, whereas a complete record of rates can be found in the Supplementary Information (Table S1). Among treatments, significant differences were only found for DNRA (F(2,161) = 10.0). Games-Howell post hoc tests indicated DNRA depends on oxygen in the water https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License. column, distinguishing between AD (2.80 ± 2.56 mmol N m -2 h -1 ) and OL (1.54 ± 1.53 mmol N m -2 h -1 ) or OD (1.35 ± 1.20 260 mmol N m -2 h -1 ) treatments.
Within each treatment, significant differences were found over time among processes. DNRA and N2O-denitrification showed significant time-related differences in the OL treatment (F(5,47) of 5.70 and 3.82, respectively). These processes, together with N2-anammox in the interval 3-6 h of incubation, were shown as the dominant ones according to Game-Howell post hoc tests. In the OD treatment, significant differences among processes were found in the interval 3-24 h of incubation. 265 At that time, DNRA and N2O-denitrification rates differed from the other processes (F(5,49) of 6.89 and 3.53, respectively).
Games-Howell post hoc tests showed that DNRA was the dominant process in the OD treatment between 3 and 6 h of incubation, and then, up to 24 h of incubation, DNRA was co-dominant with N2O-denitrification. Finally, significant differences were found in the AD treatment between DNRA and the other processes from 3 h of incubation onwards (F(5,50) = 3.32). Games-Howell post hoc tests indicated that DNRA was the dominant process up to 48 h. 270

Hydrogeochemical dynamics during sediment incubations
Three different treatments were applied during sediment incubations by modifying oxygen and light conditions in the water column (see Table 2 for details). The evolution of N parameters is shown in Figure 1 (N-species) and Figure 2 ( 15 N). In Figure 1, the evolution of N-NO3 − and N-NO2 − showed a well-defined nitrate-reduction pattern in all treatments. N-NO3 − concentrations decreased to below the limit of detection during the first stage (S1) in all treatments. Immediately after tracer 275 addition, N-NO3 − increased markedly and then gradually decreased (stage S2), the decline being fastest in the AD treatment (anoxia and darkness). This trend was also observed in Figure 2: 15 NO3 − reached maximum concentrations at 6 h (OL treatment), 12 h (OD treatment), and 1 h (AD treatment), and was completely removed from the water column within the first 36 h (OL and OD) and 12 h (AD). This decrease in N-NO3 − concentrations suggests the existence of assimilatory and/or dissimilatory nitrate reduction processes. 280 During the final stage (S3), N-NO3 − was below LOD, which may be the result of either the absence of complete nitrification in the water column, or the rate of NO3 − consumption being higher than the rate of NO3 − production. N-NO3 − in the sediment was significantly higher at the end of the incubations in all treatments. The existence of NO3 − reduction pathways, mainly during S2, is the most plausible explanation according to the isotopic data. However, other major issues for 15 N-IPT studies such as uptake and intracellular storage should not be discarded (Robertson et al., 2019). Significant inputs of NO3 − may also 285 promote blooms of diatoms (frequent in Pétrola Lake), which are physiologically adapted to grow rapidly under nitrate-rich conditions (Bronk et al., 2007). A phytoplankton bloom was observed after 15 N-NO3 − addition in the light treatment (OL), with a subsequent decay. Even though we cannot prove it, the role of diatom-bacteria aggregates in removing NO3 − from the water column subsequently fueling benthic anaerobic N-cycling, should be considered (Kamp et al., 2016).
N-NO2 − peaked during the second stage (S2), paralleling the decrease in N-NO3 − (Figure 1). Subsequently, N-NO2 − 290 decreased faster in treatment AD than in treatments OL (oxygen and light) and OD (oxygen and darkness). The intermittent conversion of nitrate to nitrite suggests active NO3 − reduction processes. In contrast, N-NH4 + in the water column increased https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License. over time in all treatments. The concentration moderately increased during S1 in all treatments (Figure 1). From the addition of the labelled NO3 − , concentration of N-NH4 + increased (with small oscillations) coupled with a constant increase in 15 NH4 + ( Figure 2) up to 18 h of incubation. This increase was more pronounced in AD than in OD and OL treatments. From 24 h 295 until the end of the incubation, N-NH4 + concentration increased whereas 15 NH4 + tended to stabilize. According to these results, N-NH4 + accumulation in the water column can be explained by DNRA (0 -18 h), OM remineralization, and sedimentary NH4 + release (Kalvelage et al., 2013). During the S1 stage, the absence of NO3 − hindered the activity of DNRA, and the increase of NH4 + in the water column must therefore be a consequence of rapid release from decaying cyanobacteria, as demonstrated by others (Gao et al., 2013). The small oscillations observed through S2 were the result of fluctuations in N-300 NH4 + production (DNRA, water-column OM remineralization) and consumption (anammox, NH4 + assimilation, and nitrification). Thus, NH4 + accumulation in the water column during S3 can be attributed to sedimentary OM remineralization after bloom collapse (García-Robledo et al., 2011), as the values did not significantly differ from initial conditions.
Concentrations of N2 were measured from the addition of the 15 N-NO3 − . In general, Figure 1 showed a stable concentration of N2 over time with small peaks in the first 12 h of incubation (positive for OL and AD treatments, negative for OD 305 treatment). However, a sudden drop in N2 concentration was found at 5 h in treatment OD, probably due to an occasional episode of N2 stripping, even though steady state gas concentrations were kept throughout the incubations. By comparing these data with 15 N-N2 evolution data (Figure 2) small variations in both 29 N2 and 30 N2 were observed after the tracer addition, where the sharp increase of N2 in the OD treatment at 48 h coincided with an abrupt rise in 30 N2. During the hours after the tracer addition (stage S2), the production of 30 N2 can be attributed either to denitrification or to coupled DNRA-310 anammox, by combining the DNRA substrate ( 15 NO2 -) with the DNRA product ( 15 NH4 + ) (Holtappels et al., 2011).
Considering N2O evolution, a different pattern was observed than that described for N2. Thus, an increasing trend was observed in all three treatments (Figure 1). N2O accumulated towards the end of the incubations in treatments OD and AD, with concentrations above 2.0 mmol/L. 45 N2O and 46 N2O evolution also provided evidence of N2O-denitrification during the S2 stage, both increasing over time ( Figure 2). However, once the original NO3 − had been consumed, increases in both 45 N2O 315 and 46 N2O for the oxic treatments can be attributed to other processes such as 15 N recirculation by coupled DNRAnitrification (DNRA fueling nitrification to N2O), a process whose importance has recently been highlighted in estuarine sediments (Dunn et al., 2009;Murphy et al., 2016). Finally, the solubility of N2O at 50 g/L of salinity and 25 ºC was 14.25 mmol/L, whereas the solubility for N2 at the same conditions was significantly lower (0.43 mmol/L). Therefore, N2 oversaturation is observed in the water column (Figure 1), clear evidence for the presence of the above-described NO3 − 320 reduction processes (Wenk et al., 2013;. Concentrations of N2 remained almost constant throughout the incubations (≈ 6 mmol/L), regardless of whether atmospheric air (OL and OD) or argon (AD) was bubbling in the mesocosm. Thus, N2 oversaturation is probably the result of N reduction and further accumulation in the water column, which has not yet reached atmospheric equilibrium for nitrogen (Weiss and Craig, 1973). About 29 N2 and 30 N2 concentrations, used for the IPT calculations, they remained almost unchanged and below the maximum solubility value (Figure 2). In addition, the 15 N mass 325 balance was calculated to detect whether gas bubbling (atmospheric air or argon to maintain mesocosms in an aerobic or https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License.
anoxic state) and differences in solubility may strip 29 N2 and 30 N2 faster than 45 N2O and 46 N2O (Figure 3). Mean 15 N recoveries were 92% for OL (from 79 to 108%), 94% for OD (from 67 to 125%), and 93% for AD (from 73 to 126 %). 15 N losses of 6-8% based on whole-system 15 N recoveries seem very small and may derive mainly from the accumulation of errors in the measurements (concentrations and at% 15 N enrichments) of 4-5 dissolved and gaseous N pools. Therefore, we 330 consider that there were no significant N losses deriving from gas bubbling, but if so the experiment would have only underestimated N2 production processes (i.e. full denitrification and anammox). Figure 4 shows the evolution of physico-chemical parameters. The evolution of DOC and DON in the water column showed stable concentrations during S1, followed by a sharp increase in S2 after tracer addition (Figure 4). Diatom blooms can disrupt the ecological balance, causing the breakdown of cyanobacterial populations, and the release of large amounts of 335 dissolved OM (Xue et al., 2017). Afterwards, DOC and DON decreased as a result of heterotrophic metabolism. In S3, we explain the observed DOC changes first by consumption and finally by accumulation after bloom collapse. DON increases are likely related to phytoplankton decay with subsequent organic N mineralization. Decreasing percentages of DON:DNb underline the role of OM remineralization throughout the incubation. Compared to initial conditions, pH decreased in OL, OD and AD treatments. The decrease of pH values is probably due to the release of organic acids and CO2, both produced 340 from carbon sources during microbial metabolism. pH values were within the optimal range for denitrification (Knowles, 1982) and DNRA (Van den Berg et al., 2015). Eh abruptly dropped in S1 and maintained negative values during S2 and S3, with a small rise after tracer addition (Figure 4).

Nitrous oxide production
The contribution of each process to total N removal was calculated for each mesocosm and treatment. ANOVA results 345 provided evidence of a (co-)dominant role of N2O-denitrification in OL and OD treatments, with 82% and 81% of N removal, respectively ( Figure 5). Contribution of N2O-denitrification to total N loss was significantly higher than reported for aquatic sediments (< 8.6%; Risgaard-Petersen et al., 2003;McCrackin and Elser, 2010). Treatments OL and OD showed mean N2O-denitrification rates of 1.76 (± 2.58) and 2.10 (± 2.24) mmol N m -2 h -1 , respectively. Such high values have been reported previously only in tropical wetland soils (up to 1.56 mmol N m -2 h -1 ; Liengaard, et al., 2014) and estuarine 350 sediments affected by agricultural activities (up to 4.85 mmol N m -2 h -1 ; Salahudeen et al., 2018) (Table 4). These results support evidence from previous observations (Huttunen et al., 2003), which showed that lakes subjected to elevated N inputs are an important source of N2O emissions. However, studies involving the role of N2O-denitrification in saline aquatic environments are mainly restricted to marine ecosystems. Our high measured rates may be explained by the high biological activity after 15 N-NO3 − addition, in the absence of nutrient limitation. Nonetheless, the N2O production in field studies is 355 most probably limited by N availability. Another potential source of N2O production is partial nitrification. In treatments OD and OL, the conditions are met for this process to take place. No NO3 − was measured after the tracer addition and further consumption, but N2O was produced over time. In fact, Figure 6 shows that N2O-denitrification was the main pathway in the https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License.
S3 stage for such treatments, which may mask the contribution of nitrification fueled by DNRA. Finally, abiotic contribution to N2O production may also contribute to produce N2O in hypersaline environments (Samarkin et al., 2010). 360 The AD treatment showed a similar average value of N2O-denitrification (1.87 ± 3.99 mmol N m -2 h -1 ) than treatments OL and OD, being similar to rates reported for pristine mangrove sediments (up to 0.67 mmol N m -2 h -1 ; Fernandes et al., 2010), but higher rates of N2-denitrification than OL and OD. Therefore, N2O-denitrification showed a smaller yet still dominant contribution to total N removal in the AD treatment. A possible explanation for this pattern is that N2O reductase activity is sensitive towards oxygen (Bonin and Gilewicz, 1991), being partially inhibited in treatments OL and OD in the presence of 365 dissolved O2 (~ 6.4 mg/L in the water column), thereby decreasing N2-denitrification in aerated treatments. Overall, N2Odenitrification showed a significant contribution to NO3 − reduction during the whole sediment incubations together with DNRA ( Figure 6). In terms of NO3 − reduction, when N2O-denitrification was of greater importance, DNRA and anammox showed a smaller contribution to NO3 − reduction, and vice versa. As suggested above, the influence of partial nitrification as a source of N2O cannot be discarded. Unfortunately, additional experiments would have been necessary which were not the 370 main focus of this study (pathways of nitrate removal).
N2-denitrification showed the highest rates at the beginning of the incubation (AD, ≤ 13 mmol N m -2 h -1 ). Mean measured N2 production rates attributed to denitrification in the OL treatment was 0.05 mmol N m -2 h -1 , in accordance with intact estuarine sediments (0.036 -0.155 mmol N m -2 h -1 ; Trimmer et al., 2003) and contributed on average 4% to total N removal ( Figure 3). N2-denitrification played a greater role in NO3 − reduction under darkness: 11% and 13% of the total N removal in 375 OD and AD treatments, respectively These results agree with earlier observations (Risgaard-Petersen et al., 1994) which showed reduced denitrification rates associated with light exposure and photosynthesis by benthic microphytes. In the OD treatment, mean production rate was 0.41 (± 1.57) mmol N m -2 h -1 by N2-denitrification (Table 3). Our values were similar to those reported in marine environments like Heron Island (0.48 mmol N m -2 h -1 ; Eyre and Ferguson, 2008) and Randers Fjord (0.34 mmol N m -2 h -1 ; Risgaard-Petersen et al., 2004) (Table 4). Highest N2-denitrification rates were found in the AD 380 treatment with an average value of 0.80 (± 2.61) mmol N m -2 h -1 . These results were close to those reported by Erler et al. (2008) (0.652 -0.966 mmol N m -2 h -1 ), where denitrifiers coexisted with anammox bacteria in a constructed wetland which received secondary treated sewage effluents. The largest contribution of N2-denitrification was detected at the initial stages of incubation, coupled to higher DOC concentrations, but also during late stages of incubation in OD (~ 30%) (Figure 4).
These results suggest the dominance of heterotrophic denitrification linked to the breakdown of biomass (Xue et al., 2017). 385

Close coupling between DNRA and anammox
Total N removal and NO3 − reduction reached highest values under anoxia and darkness conditions (mean of 3.63 ± 5.30 mmol N m -2 h -1 and 6.43 ± 6.56 mmol N m -2 h -1 , respectively; Table 3). As discussed above, under those conditions DNRA was the dominant process. These results are consistent with hydrochemical data, which showed a significant accumulation of N-NH4 + in the water column in the AD treatment. 390 https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License.
Previous research showed favorable conditions for DNRA activity in sediments from Pétrola Lake, such as high organic C:N ratios or the presence of microorganisms capable of performing DNRA (Valiente et al., 2017;Valiente et al., 2018). Average DNRA rates in OL and OD treatments (~ 1.4 mmol N m -2 h -1 ) were similar to those reported for anoxic estuarine sediments where DNRA was the dominant process (1.140 mmol N m -2 h -1 ; Dong et al., 2011). In the AD treatment, mean DNRA rates (2.80 ± 2.56 mmol N m -2 h -1 ) were similar to those observed in nutrient enriched environments like fringing wetlands (up to 395 6.13 mmol N m -2 h -1 ; Tobias et al., 2011) or eutrophic shelf seas (up to 3.58 mmol N m -2 h -1 ; Song et al., 2013) (Table 4). In line with the data obtained in the ANOVA tests, the 15 N-IPT data showed that NO3 − reduction by DNRA was significantly higher in AD (52%) than in OL (41%) and OD (35%) ( Figure 5). The contribution of DNRA was in the same range as reported for estuarine and salt marsh sediments (Dong et al., 2009;Koop-Jakobsen and Giblin, 2010), fostering the retention of reactive nitrogen in the system. Compared to denitrification, DNRA contributed more to NO3 − reduction after the initial 400 incubation phase, approximately from time 2.5 h onwards (Figure 4). Recent studies also demonstrated that DNRA is stimulated in the presence of H2S at the expense of denitrification (Roland et al., 2017). Our results support those findings: AD provided the most favorable conditions for bacterial sulfate-reduction (Table 2), and H2S production in Pétrola sediments (Valiente et al., 2017) can reach values up to 0.024 nmol/cm 3 •s.
DNRA therefore plays the key role in increasing N-NH4 + contents in the water column, being more relevant than OM 405 remineralization and sedimentary release. Existing NH4 + may be oxidized to NO2 − both under aerobic and anaerobic conditions (Schmidt et al., 2002), contributing to a temporary increase of N-NO2 − and promoting NO2 − and NH4 + consumption by anammox bacteria. Moreover, N-NH4 + release does fuel N loss from the system via coupled DNRAanammox. Therefore, DNRA and anammox bacteria acting together may have an energetic advantage over denitrifiers in the competition for substrates under low oxygen conditions . The close reliance of anammox on DNRA has 410 been reported in marine ecosystems with high N loss via anammox, mainly linked to the availability of OM (Kalvelage et al., 2013). In Pétrola Lake sediment incubations, anammox seems to be fueled by DNRA. This interpretation is based on the similar trend in the contribution of both processes to total NO3 − reduction (AD>OL>OD; Figure 3). Coupled DNRAanammox showed a higher contribution in all treatments than canonical anammox (Table 3), corroborating the key role of DNRA in fueling N loss pathways. 415 The isotope data clearly confirm the presence of anammox (Table 3). The mean rates of N loss via anammox in OL and OD treatments (~0.4 mmol N m -2 h -1 ) were in the range of previous studies in eutrophic sediments (up to 0.413 mmol N m -2 h -1 ; Han and Li, 2016), but significantly lower than those found in the AD treatment (0.96 mmol N m -2 h -1 ). These results agree with recent studies showing the importance of anammox activity in the presence of H2S in freshwater lakes (Roland et al., 2017), conditions which are given for the highly saline lake studied here. On average, the contribution of anammox to total 420 N loss ranged from 8% (OD) to 28% (AD) (Figure 3). This range corresponds with studies performed in continental shelf sediments  (28%), intertidal sediments (Hsu and Kao, 2013) (12%), and is close to the global mean value including inland waters (Trimmer and Engström, 2011) (23%). Finally, the higher the participation of anammox in total N https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License. removal, the lower the relevance of N2O-denitrification, leading to decreased N2O emissions from hypersaline lake ecosystems. 425

Conclusions
The purpose of the current study was to determine the influence of oxygen and light in the water column on the balance between nitrate removal pathways during incubations of lacustrine organic-rich sediments. Our findings provide evidence for the coexistence of denitrification, DNRA, and anammox in a highly saline lake. The application of the revised 15 N-IPT highlighted the importance of coupled DNRA-anammox during our incubations. We showed here that DNRA and N2O-430 denitrification played a predominant role in N removal under the studied conditions, showing unexpectedly high N2O emission rates compared to previous studies. However, DNRA was the key process when oxygen and light were absent from the water column. Then, anammox also had a greater influence on total N removal with markedly high rates (up to 0.96 mmol N m -2 h -1 ). It seems that anoxia and darkness promoted DNRA, the critical process which fuels anammox. As a result, these conditions reduced N2O emissions to the atmosphere. As far as we know, the role of coupled DNRA-anammox in such 435 saline ecosystems has not yet been explored, and therefore, anammox was typically underestimated. Further research is required to fully understand the role of coupled DNRA-anammox in N cycling in lake ecosystems, as well as the influence that coupled DNRA-nitrification can exert on N2O production.      https://doi.org/10.5194/bg-2020-20 Preprint. Discussion started: 31 January 2020 c Author(s) 2020. CC BY 4.0 License.  (%) (µmol L -1 ) (µmol L -1 ) (µmol L -1 ) (mmol L -1 ) (mmol L -1 ) (%) (µmol kg -1 ) (mmol kg -1 ) (µmol kg -1 ) Table 3. Mean (±SD) and maximum rates of N-loss processes after 72 h of mesocosm incubations.

Mesocosms
Total N removal Total NO3reduction N2-Denitrification N2O-Denitrification DNRA-Anammox Canonical Anammox N2-Anammox DNRA mean max mean max mean max mean max mean max mean max mean max mean max