Vertical segregation among pathways mediating nitrogen loss (N-2 and N2O production) across the oxygen gradient in a coastal upwelling ecosystem

. The upwelling system off central Chile (36.5 ◦ S) is seasonally subjected to oxygen (O 2 ) -deﬁcient waters, with a strong vertical gradient in O 2 (from oxic to anoxic matter and the fuel for the production of other electron donors (i.e. ammonium) and acceptors (i.e. nitrate and nitrite) after its remineralization. These results highlight the links between several pathways involved in N loss. They also establish that different mechanisms supported by alternative N substrates


Introduction
It is widely accepted that fixed nitrogen (N) availability influences marine primary productivity (Falkowski et al., 1998). Therefore, fixed-N loss from the ocean has an important influence on ecosystem functioning and global biogeochemical cycles. A significant fraction of the global fixed-N removal (30-50 %) occurs in oxygen minimum zones (OMZs), where dissolved oxygen (DO) levels as low as 2-4 µmol L −1 or lower activate nitrate-based anaerobic metabolisms (Devol, 1978;Dalsgaard et al., 2014). Fixed-N removal occurs predominantly in the form of N 2 , generated by canonical denitrification and anammox processes (Gruber and Sarmiento, 1997;Codispoti et al., 2001;Devol, 2003). The steep oxygen (O 2 ) gradients observed at the OMZ's boundaries further trigger a substantial production and accumulation of nitrous oxide -N 2 O (Naqvi et al., 2010;Dalsgaard et al., 2012), which may eventually be emitted to the atmosphere, thus contributing to the overall N loss. OMZs are notable features of eastern boundary ecosystems (i.e. South and North Pacific, South Atlantic) and the Arabian Sea, where seasonal or permanent wind-driven upwelling sustains remarkably high levels of biological production, and subsequently a high respiration rate from subsurface organic matter (Helly and Levin, 2004).
The eastern South Pacific (ESP) is influenced by the shallow poleward Peru-Chile undercurrent (e.g. Blanco et al., 2001), which transports nutrient-rich, high-saline, and O 2 -depleted Equatorial Subsurface Waters -ESSWs (Strub et al., 1998). During austral spring-summer (September-April), a marked upwelling-favourable period driven by persistent southwesterly winds, ESSWs arrive at the coastal area off central Chile, impinging on one of the widest continental shelves in Chile (Sobarzo and Djurfeldt, 2004;Sobarzo et al., 2007). Throughout the upwelling season high primary production rates (Daneri et al., 2000) occur along with a large downward flux of organic matter, leading to an increase in respiration rates at depth (Montero et al., 2007). These conditions, coupled with the presence of already O 2deficient waters, and greater availability of electron donors and acceptors (as a result of the advection of inorganic substrates, local organic matter respiration, and autotrophic redox reactions), create steep gradients for O 2 and other chemicals, with the establishment of hypoxic-anoxic conditions at middle-bottom depths. This leads to an active removal of fixed N and greenhouse gas production (Farías et al., 2009;Galán et al., 2014). Likewise, the dense upwelled waters cause increased stratification in the water column, decelerating the downward flux of organic particles formed at the surface (Charpentier et al., 2007), facilitating their accumulation, and thus creating anoxic microbial hotspots within the oxycline that favour the N-based anaerobic microbial metabolisms (Klawonn et al., 2015;Stief et al., 2016).
As previously mentioned, OMZs are also characterized by sustaining an important production and outgassing of N 2 O (e.g. Nevison et al., 2004;Naqvi et al., 2010), which is clearly evident in this system . N 2 O is a potent greenhouse gas, with a ∼ 300 times greater radiative effect than CO 2 , and also contributes to the destruction of the stratospheric ozone (Ravishankara et al., 2009;Portmann et al., 2012). The distribution and accumulation of N 2 O is highly sensitive to O 2 levels, and thus far both heterotrophic and autotrophic processes are known to contribute to its cycling (Ritchie and Nicholas, 1972;Wrage et al., 2001;Codispoti, 2010). Nevertheless, the mechanisms responsible for N 2 O accumulation in both the upper and lower OMZ oxyclines remain unclear. Under anoxia, N 2 O produced is completely reduced to N 2 . This reduction is, however, inhibited with increasing O 2 levels and N 2 O accumulates and in some cases becomes the main product of denitrification processes (Kester et al., 1977;Goreau et al., 1980;Bonin et al., 1989; but see also Qu et al., 2016). N 2 O can also be produced during ammonium (NH + 4 ) oxidation, a chemolithoautotrophic process by which NH + 4 is oxidized aerobically either by bacteria or archaea. However, the pathways of N 2 O production appear to differ between ammonium oxidizers from these two domains (Blainey et al., 2011;Kim et al., 2011;Spang et al., 2012;Kozlowski et al., 2016). While N 2 O is a minor product in ammonium oxidation (up to 10 % relative to NO − 2 production; Goreau et al., 1980), the yield of N 2 O produced by nitrifiers relative to NO − 2 increases ∼ 20 times as O 2 saturation in the atmosphere diminishes to 1 % (Kester et al., 1977). N 2 O is formed as a byproduct of hydroxylamine (NH 2 OH) and nitric oxide (NO) precursors in ammonium oxidizing bacteria (Wrage et al., 2001;Arp and Stein, 2003;Stein, 2011), and during situations of O 2 stress, this bacterial group could potentially reduce NO − 2 produced into N 2 O via NO, through nitrifier-denitrification process (Ritchie and Nicholas, 1972;Poth and Focht, 1985;Shaw et al., 2006). Conversely, and despite several reports suggesting that archaeal ammonium oxidizers are an important source of N 2 O based on their ubiquity, abundance, activity and high NH + 4 affinity (Könneke et al., 2005;Wuchter et al., 2006;Martens-Habbena et al., 2009;Santoro et al., 2011;Jung et al., 2014), the pathway through which the N 2 O is produced by this group is still not fully understood. Nonetheless, it was recently demonstrated that archaea involved in ammonium oxidation do not have the enzymatic capacity to reduce NO to N 2 O through nitrifier denitrification, and N 2 O formation was suggested to occur through a hybrid pathway pairing nitrogen  (Stieglmeier et al., 2014;Kozlowski et al., 2016). Although the atmospheric N 2 O concentration is increasing by ∼ 0.25 % annually (IPCC, 2014), estimates of the marine contribution to N 2 O budget are highly uncertain. This is mainly due to the lack of coastal upwelling areas included in the global estimates of oceanic N 2 O emissions, although the contribution from these regions is potentially significant (Bange et al., 1996;Nevison et al., 2004). While several decades of research have focused on measuring (or estimating) the products of N loss (as N 2 and N 2 O), few studies have tracked the sources to determine the dominant process or to establish whether loss products are formed as the result of coupling among multiple N-cycling pathways. Despite both denitrification and anammox being considered as major loss terms for available N (as either N 2 and/or N 2 O), canonical nitrification or nitrifier denitrification may also contribute to removal in systems with high O 2 variability. Thus, when low O 2 concentrations prevail, it is possible that multiple microbial N-loss processes operate simultaneously. However, the relative contribution of each metabolic pathway, and direct and indirect processes involved in N loss, remains unsolved. This study investigates variations in the N-based metabolisms of the microbial community, coupled with N loss, in the seasonally stratified and O 2 -limited waters of the coastal upwelling system off central Chile, during the spring transition and summer, across a gradient from oxic to anoxic conditions. We used a combination of 15 N labelled activity measurements, inhibitory assays, and analysis of natural nitrate isotope abundances. The results obtained from this coastal ecosystem, where extreme O 2 variability is seasonally observed, could help to understand the ecological and biogeochemical implications of global warming, which is expected to cause intensification and expansion of oceanic OMZs (Keeling et al., 2010;Deutsch et al., 2011;Helm et al., 2011).

Hydrographic analyses
Seawater samples for discrete hydrographic measurements were taken in triplicate. Filtered (0.7 µm, GF/F glass fibre filter) water was analysed for NO − 3 , NO − 2 and PO 3− 4 using standard colorimetric methods (Grasshoff et al., 1983) in an automatic analyser (Seal Analytical). NH + 4 was measured fluorometrically from fresh samples (40 mL), using the orthopthaldialdehyde method (Holmes et al., 1999) on a Turner Designs Trilogy Laboratory Fluorometer (Turner Designs). The relative standard error for NO − 3 , NO − 2 , PO 3− 4 and NH + 4 was lower than ±10, ±3, ±3, and ±5 %, respectively. N 2 O was analysed from the equilibrated headspace (5 mL of pure He) in 20 mL crimp-cap bottles by gas chromatography (Varian 3380) using an electron capture detector maintained at 350 • C. The coefficient of variation of the dissolved N 2 O measurements was < 3 % (for more details, see Cornejo and Farías, 2012).

Satellite-derived chlorophyll-a
High-resolution (1 km) MODIS (Moderate Resolution Imaging Spectroradiometer) imagery of Chlorophyll-a (Chl-a) were processed, based on the default Chl-a algorithm for MODIS (OC3), using NASA's SeaDAS (SeaWIFS Data Analysis System) software for the coastal region off Concepción. MODIS level 1 files were obtained from NASA (https: //oceancolor.gsfc.nasa.gov/data/) and processed using an improved atmospheric correction method for coastal turbid waters (Wang and Shi, 2007). Further details regarding processing options and flags are found in Saldías et al. (2012). Composites of 3 days were generated to reduce the cloud cover and increase the number of pixels available during the sampling dates.

Satellite-derived wind stress and in situ data
Satellite-derived, level 2 coastal ocean surface wind vectors were retrieved from the Advanced Scatterometer (AS-CAT) on MetOp-A, at 12.5 km sampling resolution. Data files are obtained from the Royal Netherlands Meteorological Institute (http://projects.knmi.nl/scatterometer/). For the study area, particular swaths presented wind measurement data from ∼ 20-25 km from the coast, with improved wind vectors over the continental shelf around station 18. In situ wind data were obtained from the Carriel Sur Airport meteorological station. Daily means of wind stress were derived following Large and Pond (1981). Considering the synoptic character of the upwelling events, and to obtain the water mass residence time at the sampling station, box plots representing the boundaries of statistical values, and the error of in situ wind stress were calculated for the 5 days previous to sampling. To calculate the onset (i.e. spring transition) and extension of the upwelling season, cumulative alongshore (south-north) wind stress was estimated as described by Barth et al. (2007).

15 N tracer and inhibitor experiments
Experimental assays with 15 N-labelled tracers and specificpathway inhibitors were carried out with water from the oxycline and the bottom layer (25 and 85 m depth, respectively) in order to evaluate the contribution and coupling of the different processes involved in N loss through the water column. The 15 N-tracer incubations followed the basic procedures as previously described by Thamdrup et al. (2006), with some modifications. Basically, 250 mL glass bottles were filled immediately following water sampling with the Niskin bottles, to avoid oxygenation, allowing overflow until the volume was replaced three times. The bottles were closed with butyl rubber septa and crimps, and kept in the dark at a low temperature until later processing in the laboratory. Time from sampling until incubation start was < 2 h. 15 N-labelled fixed-N substrates were added through the septa into separate bottles as follows (the final concentration is stated in µmol L −1 in parentheses): 15 NO − 3 (15), 15 NO − 2 (5), and 15 NH + 4 (5). In order to determine the 15 N mole fraction, concentrations of the substrates were measured before and after addition of the 15 N-labelled compounds. Treatments from the oxycline during January were evaluated under in situ O 2 conditions, while the rest of the samples were purged with He during 15 min for anoxic incubations (both depths during September and the bottom samples in January). Anoxic conditions after helium purging were confirmed by the switchable trace O 2 (STOX) microsensor (Revsbech et al., 2009). After adding the substrates, mixing, and purging (if necessary), water was dispensed into 12.6 mL glass vials (Exetainers, Labco), until they were completely filled, allowing the volume to be replaced three times to avoid oxygen contamination. Exe-Biogeosciences, 14, 4795-4813, 2017 www.biogeosciences.net/14/4795/2017/ tainers were incubated at in situ temperatures (10-11 • C) for up to 48 h. Subsamples were sacrificed in triplicate at four instances during the incubation (0, 15, 24, 48 h), and biological activity was halted by injection of 50 µL of a 50 % (w/v) zinc chloride (ZnCl 2 ) solution into each Exetainer. The N-isotopic composition of 15 N 2 produced by anammox and/or denitrification was measured by gas chromatography continuous-flow isotope ratio mass spectrometry (GC-CF-IRMS; Finnigan Delta Plus) using N 2 as standard (Dalsgaard et al., 2012). Subsequently, 15 N in N 2 O were analysed from the same samples following the chromatographic separation of N 2 O on a GC column. Specific inhibitors were used to block the activity of target pathways/microorganisms and discriminate the contribution of the different processes to N cycling and N 2 and N 2 O production. Allylthiourea -ATU (at a final concentration of 86 µmol L −1 ), a specific bacterial inhibitor of the ammonium monooxygenase (AMO) enzyme (Ginestet et al., 1998), and N1-Guanyl-1, 7 Diaminoheptane -GC7 (at a final concentration of 100 µmol L −1 ), a phylogenetic inhibitor of the archaeal growth (Jansson et al., 2000), were used in 15 NH + 4 incubations to separate the bacterial and archaeal contribution to ammonium oxidation (AO), and to indicate any possible competition for 15 NH + 4 between ammonium oxidizers and anammox during both sampling periods. Acetylene (final concentration of 10 % v/v), which inhibits the AO and the final step of denitrification (i.e. N 2 O to N 2 ; Balderston et al., 1976), was added to the 15 NO − 2 treatments during January to quantify the gross N 2 O production from nitrite reduction (NiR). This experimental design is detailed in Table 1.

15 N-labelled dissolved inorganic N species
Pathways indirectly coupled to N-loss processes were measured in the same incubations from dissolved inorganic N compounds formed by redox reactions (i.e. 15 NH + 4 from 15 NO − x or 15 NO − 2 from 15 NH + 4 and 15 NO − 3 ). After analysing 15 N 2 and 15 N 2 O in the headspace of the Exetainers, the dissolved 15 N-labelled species ( 15 NH + 4 and 15 NO − 2 ) remaining in the liquid phase from the different treatments were transformed into 15 N 2 and analysed by GC-CF-IRMS as previously described. To determine DNiRA rates, 15 NH + 4 that accumulated in 15 NO − 2 treatments was converted to 15 N-N 2 using an alkaline hypobromite solution (Warembourg, 1993). Furthermore, rates of AO and nitrate reduction (NaR) were calculated as the mole fraction of 15 NO − 2 produced in 15 NH + 4 and 15 NO − 3 treatments, respectively, after conversion of the 15 NO − 2 produced by these pathways to 15 N-N 2 with sulfamic acid (McIlvin and Altabet, 2005).

Natural abundance of nitrate isotopes
The natural abundance isotopic composition of N in NO − 3 was analysed by the denitrifier method (Casciotti et al., 2002;Sigman et al., 2001) after removal of NO − 2 present in the samples by treatment with sulfamic acid (McIlvin and Altabet, 2005). The δ 15 N values were measured in the N 2 O quantitatively produced after nitrate reduction by denitrifying bacteria that lack N 2 O-reductase activity. The isotopic composition was analysed in a GC-CF-IRMS (Finnigan Delta Plus). Data are reported in delta notation as The standard deviation for nitrate isotope analysis ranged between 0.01 and 0.4 ‰ (Casciotti et al., 2002).

Calculations
N 2 produced by anammox and denitrification, and from the chemical conversion of NO − 2 and NH + 4 to N 2 , was determined from the excess of 15 N-labelled N 2 over the 15 N mole fraction in the source compounds according to Thamdrup et al. (2006), taking into account the random and hybrid isotope pairing patterns associated with denitrification and anammox, respectively. Rates were derived from the slope of the linear regression of 15 N-N 2 production ( 14 N 15 N and 15 N 15 N) as a function of incubation time. Only significant (p ≤ 0.05) and linearly developing rates without lag phase were considered. Rates of both NO − 2 and NH + 4 production were calculated from linear regression of the concentration of the 15 N-labelled product over time. Rates of N 2 O production are reported as separate rates of 45 N 2 O ( 14 N 15 NO) and 46 N 2 O ( 15 N 15 NO) accumulation, also from linear regression, as different pathways with different patterns of isotope pairing may contribute to N 2 O formation.
N deficits (N * ) were calculated as stoichiometric anomalies from the Redfield ratio using the relation between fixed inorganic N and phosphate concentrations (Gruber and Sarmiento, 1997).

Oceanographic setting
The hydrographic conditions over the continental shelf off Concepción showed the steepening of the O 2 gradients as well as other chemical species, during the development of the austral upwelling season (September to April). The onset of the upwelling season during spring (early September) was characterized by a clear shift in the wind direction to predominantly southerly winds (arrows in Fig. 1a and positive wind stress values denoted as red bars in Fig. 1b), and the winddriven incursion (up to 25 m depth) of cold, saline, NO − 3rich and O 2 -poor ESSWs (identified by the 11 • C isotherm and 34.6 isohaline) ( Fig. 2a and c). This onset was also indicated by a change to positive values in the cumulative alongshore wind stress, which initiates on 9 September 2013 (data not shown). This injection of nutrients combined with a high availability of photosynthetically active radiation (PAR) triggered the progressive enhancement of near-surface primary production (see the temporal increases in satellite-derived Chl-a in Fig. 1a). This situation is intensified during summer, with stronger southerly wind pulses (Fig. 1a) and further shoaling of the ESSW to 20 m (Fig. 2e). During this period the exhaustion of O 2 also intensified at the bottom (< 5 µmol L −1 ). Thus, three layers are established in the water column as a consequence of the influx of ESSW (Farías et al., 2009): a well oxygenated mixed layer, the oxycline characterized by a marked reduction of the dissolved O 2 content over a compressed vertical scale, and the bottom layer, where the oxygen deficiencies led to potentially anoxic conditions (Murillo et al., 2014). Furthermore, an increased availability of NO − 3 and NO − 2 as electron acceptors, and of organic matter, NO − 2 , NH + 4 , and H 2 S as donors, are also characterized by the intensification of upwelling through the two deeper layers .
During September, NO − 3 -rich waters (up to 30 µmol L −1 ) occupied most of the subsurface, from the bottom to the upper boundary of the oxycline (up to 25 m depth; Fig. 2c). Nevertheless, although similar high NO − 3 values remained in the oxycline during January (Fig. 2g), notable NO − 3 depletion (values as low as 19.5 µmol L −1 ) was observed close to the bottom and associated with extreme O 2 exhaustion (∼ 4 µmol O 2 L −1 ), with subsequent NH + 4 and NO − 2 accumulation (up to 1.3 and 1.6 µmol L −1 , respectively). N deficiencies (N * ) generally increased with depth ( Fig. 2d and h), with the greatest depletion (−24 µmol L −1 ) observed in bottom waters during January, which seems to be associated with the NO − 3 consumption, while similar N * values were observed in the oxycline during both periods (∼ −10 µmol L −1 ). The vertical distribution of N 2 O was bimodal with similar peak values at the oxycline and in bottom waters, with concentrations slightly increasing from September to January (∼ 25 to 27 nmol L −1 , respectively; Fig. 2d and h).
During September, the maximum concentration and mean particle diameter were observed in the surface layer (120 µL L −1 and up to 77 µm, respectively), possibly due to the build up of phytoplanktonic biomass, with a secondary peak occurring close to the base of the oxycline (36 µL L −1 and 47 µm, respectively; ∼ 25 m depth). During January, the particle concentration decreased relative to September, but showed a noteworthy increase in size (an entire order of magnitude greater). During this period, particles exhibited a broad distribution along the oxycline with a maximum size and concentration at the upper boundary (up to 650 µm and 20 µL L −1 , respectively; ∼ 10 m depth), and a subsurface secondary peak (∼ 500 µm and 11 µL L −1 , respectively; ∼ 25 m depth) around the base of this layer ( Fig. 2b and f).

Natural abundance of nitrate isotopes
The vertical distribution of δ 15 N-NO − 3 showed contrasting patterns between the two sampling periods ( Fig. 2d and h). During September, the δ 15 N value was highest at the surface (14.6 ± 0.3 ‰) and decreased consistently through the water column to a minimum at the bottom (11.2 ± 0.2 ‰). In January, the δ 15 N values increased from surface to depth, ranging from 9.5 ± 0.3 to 14.5 ± 0.1 ‰, respectively. The oxycline showed relatively similar δ 15 N values for both periods, being slightly higher during September (around 12 and 11 ‰, respectively).

15 N experiments
The production of labelled N-gaseous species (measured as N 2 and N 2 O) revealed the coexistence of nitrification, anammox, and denitrification in both the oxycline and bottom layer during the study period (Table 2 and Figs. 3 and 4), with a remarkable increase in the N removal between spring and summer. Interestingly, during both sampling times a higher activity in N loss was observed in the oxycline relative to the suboxic bottom waters, where activity was < 40 % of the oxycline values.

15 N 2 production
During September (Fig. 3a, b), denitrification was the main N removal pathway in the oxycline (18.9 ± 4.3 nmol L −1 d −1 ) when compared to the activity of anammox (2.8 ± 0.3 nmol L −1 d −1 ), with rates corresponding to 87 and 13 % of the total N 2 produced in this layer, respectively. On the contrary, anammox was the only process involved in N 2 production in bottom waters (4.5 ± 0.9 nmol L −1 d −1 ). During January (Fig. 3c, d), a similar vertical segregation in N 2 production processes was observed; however the magnitude of the rates changed. While at the oxycline, denitrification registered the highest N 2 production rate during this study (50.3 ± 10.3 nmol L −1 d −1 -73 % of total N 2 production at this layer), anammox' contribution to N 2 production in the oxycline increased substantially (18.7 ± 4.6 nmol L −1 d −1 -27 % of total N 2 production at this layer) relative to that observed during September. At depth, N 2 was only formed from anammox but at a very low rate (0.4 ± 0.02 nmol N 2 L −1 d −1 ), which represents a substantial reduction of this process there in comparison to September.
During January, anammox activity increased following the addition of the AO inhibitor ATU (Fig. 3c). During this period, anammox rates from 15 NH + 4 treatments at the oxycline increased from 17.6 ± 1.7 to 26.1 ± 4.6 nmol L −1 d −1 , in incubations amended with ATU. Similar trends were observed at depth, but with a relatively lower increase (from 0.4 ± 0.02 nmol L −1 d −1 from 15 NH + 4 to 0.5 ± 0.04 nmol L −1 d −1 from 15 NH + 4 + ATU). By contrast, no significant increase was observed after addition of GC7. At depth, anammox activity was not detected after the addition of this archaeal inhibitor.
3.3.2 15 N 2 O production N 2 O production from 15 N labelled substrates showed vertical and temporal changes with respect to substrate source (Table 2 and Fig. 4a, d). In both sampling periods in the oxycline, N 2 O was produced exclusively from AO, with slightly higher rates during September (3.8 ± 0.7 nmol L −1 d −1 ), relative to January (2.2 ± 0.01 nmol L −1 d −1 ). At depth, there was a change in the N 2 O source pathway over time; however it should be noted that during September acetylene (C 2 H 2 ) was not tested in the incubations. In September, N 2 O was only produced from ammonium (5.8 ± 0.6 nmol L −1 d −1 ), whereas production from ammonium was lower (0.02 ± 0.002 nmol L −1 d −1 ) relative to the N 2 O produced from nitrite (0.05 ± 0.005 nmol L −1 d −1 ) during January. Nevertheless, the addition of C 2 H 2 , which inhibits N 2 O reduction to N 2 , resulted in a higher production of N 2 O (up to 22.6 ± 3.5 nmol L −1 d −1 from 15 NO − 2 + C 2 H 2 ). This N 2 O accumulation was not observed in the oxycline during January, despite C 2 H 2 also being added to the NO − 2 incubations.

Dissolved inorganic 15 N-compounds cycling
The redox activity related to the production of NO − 2 and NH + 4 , substrates of denitrification and anammox, is summarized in Fig. 5 and Tables 3 and 4, respectively. In general, nitrite production was higher during January (Fig. 5b) than in September (Fig. 5a), but with variations observed in maximum production activity between layers. During September, the total NO − 2 produced (i.e. NO − 2 formed from AO plus NaR) was higher in the bottom waters (15.1 nmol NO − 2 L −1 d −1 ) relative to the oxycline (8.0 nmol NO − 2 L −1 d −1 ), with AO and NaR contributing similarly in both layers. On the contrary, in January nitrite production was higher in the oxycline with NaR as the main source (83 ± 5 nmol NO − 2 L −1 d −1 and 53 ± 16 nmol NO − 2 L −1 d −1 from NaR and AO -61 and 39 %, respectively), while AO was the principal source of NO − 2 at depth (38 ± 6 nmol NO − 2 L −1 d −1 and  Significance codes: 0.001 - * * * , 0.01 - * * , 0.05 - * , not significant -ns. Without inhibitor -WI. Degrees of freedom -DF. a -Number represent "partial rates" because it is not clear which pathway forms the N 2 O from AO. 16 ± 4 nmol NO − 2 L −1 d −1 from AO and NaR -70 and 30 %, respectively). It is important to remark that the assays developed during January with water from the oxycline were incubated at the in situ oxygen level.
During January, the addition of ATU into the incubations resulted in a significant reduction (> 70 %) in nitrite production from AO, both in the oxycline (14.4 ± 4.9 nmol NO − 2 L −1 d −1 ) and at depth (9.6 ± 0.7 nmol NO − 2 L −1 d −1 ), while GC7 caused only a slight but non-significant reduction in nitrite production rates from AO at both depths. By contrast, during September, no nitrite production activity was observed from AO for treatments with either ATU or GC7 addition, indicating that AO could be completely inhibited by these compounds (data not included in Table 3).
Similarly to the nitrite production scenario, over both depths the labelled NH + 4 formed by dissimilative reduction of NO − 2 (i.e. DNiRA) was higher during January ( Fig. 5d) relative to September (Fig. 5c), with marked variations in the yield of this reaction between depths through-out the upwelling season. In September, DNiRA was similarly active in the oxycline (59.9 ± 3.9 nmol NH + 4 L −1 d −1 ) and in the bottom layer (49.3 ± 5.3 nmol NH + 4 L −1 d −1 ). On the contrary, during January a notable production of NH + 4 was observed in the oxycline (up to 751 ± 81 nmol NH + 4 L −1 d −1 ) in comparison to the production at depth (82.8 ± 7.3 nmol NH + 4 L −1 d −1 ).

Environmental variability
The hydrographic properties during the study period reflect the already well-described onset of the upwelling season during the austral spring transition (September), and the subsequent development and intensification of upwelling conditions in the summer as upwelling progresses over the shelf off central Chile (Farías et al., 2009;Galán et al., 2012). Southerly wind pulses during spring-summer ( Fig. 1) draw ESSW with low O 2 (< 50 µmol L −1 ) and a high amount of  Significance codes: 0.001 - * * * , 0.01 - * * , 0.05 - * , not significant -ns.
NO − 3 (> 25 µmol L −1 ) onto the shelf from the southernmost extension of the permanent ESP OMZ . As the upwelling season progresses, heterotrophic processes in subsurface waters are stimulated by the large amount of organic matter produced at the surface that settles on the sediments. The increase in respiration rates creates extreme O 2 depletion at depth during the summer (∼ 4 µmol L −1 ; Fig. 2e) and occasionally anoxic conditions occur in the bottom waters of this system (Murillo et al., 2014). The O 2 limitation switches on anaerobic microbial metabolisms, which initially use NO − 3 as the terminal electron acceptor, generating other inorganic species of N (i.e. NH + 4 , NO − 2 , N 2 O; Fig. 2g, h), which in turn may support chemolitho-and chemoorganotrophic metabolisms including nitrification, anammox, and denitrification, contributing in this way to the loss of fixed N. The coupling between primary production at the surface and anaerobic N-based metabolisms in the subsurface seems to depend on alternations between upwellingfavourable wind pulses and relaxed or inverted wind events, which allows for phytoplankton blooms (Daneri et al., 2012;Evans et al., 2015), and in turn, the concomitant settling of these organic particles to the sediment. An optimal window (e.g. 3-10 days) of relaxation or reversal in the wind stress not only affects the primary production rate but also  , isotopic pairing in the gas produced, and inhibitor when used (acetylene -"C 2 H 2 "). Note that N 2 O produced in the incubations with C 2 H 2 represents gross production rates. Insets in (c) and (d) represent details of lower values at depth. the microbial community composition and the biomass accumulation (Rutllant and Montecino, 2002;Wilkerson et al., 2006;Du and Peterson, 2009), and could even explain the hot N 2 O moments (short periods of disproportionately high rates) observed in this system . In this regard, clear variations in the alongshore wind stress (intensity and direction) were observed between sampling times, considering the 5 days preceding the sampling and the sampling day itself. Thus, the system was coming from a pulse of inverted, downwelling-favourable winds during the sampling of the spring transition, in contrast to a transitional upwelling-favourable situation during the sampling in summer (see box plots in Fig. 1b). These variations in wind characteristics also determine the intensity of near-surface primary production, illustrated by the Chl-a distribution as a proxy of organic matter formation (Fig. 1a), and probably the availability, size, and vertical distribution of particles through the water column (Fig. 2b, f). Hence, during September a less productive system was observed, with a secondary peak of small particles more narrowly distributed around the oxycline base, whereas in January a more intense vertical advec-tion of ESSW supported an increase in surface Chl-a, with a wider distribution of larger particles around both oxycline limits.

15 N 2 production
An intense N cycling leading to N loss over the continental shelf off central Chile was demonstrated by the 15 N labelling results (Figs. 3, 4, and 5), with higher levels of activity observed not only as the upwelling season progressed but also in the oxycline (relative to the bottom waters) during both periods. This is the first time that anammox, denitrification, and nitrification -through N 2 O production -were found cocontributing to the removal of fixed N in this coastal system and at relatively high O 2 conditions as measured in the oxycline (∼ 95 µmol O 2 L −1 ). A plausible explanation for the observation of oxygen-sensitive processes (i.e. anammox and denitrification) operating under ambient oxic conditions is that the oxycline incubations for both sampling periods were carried out with water retrieved from the secondary peak of particles (∼ 25 m depth), which accumulate at this level due to high stratification (Fig. 2b, f), causing a deceleration in their downward flux. These organic particles may create anoxic microenvironments (Ploug et al., 1997) as well as generate chemical hotspots that provide substrates (organic and inorganic) to sustain the anaerobic metabolisms of the pelagic microbiota. The influence of anaerobic microniches on N 2 production has been previously proposed in scenarios where strong covariation between anammox rates and the abundance of particle-associated anammox cells have been found in incubations with water from the Namibian continental shelf, with in situ O 2 levels reaching 25 µmol L −1 (Kuypers et al., 2005;Woebken et al., 2007). Recently, several anaerobic N processes (e.g. nitrate reduction, denitrification, anammox, dissimilatory nitrate reduction to ammonium -DNRA, and N 2 O production) have been detected in suspended cyanobacterial and diatom anoxic-aggregates (traced with O 2 microsensors) incubated in oxic waters (Klawonn et al., 2015, andStief et al., 2016, respectively), supporting the idea that particles are able to operate as anoxic N-removal hotspots. Particles in this study were smaller (≤ 0.6 mm) than the cyanobacterial/algal aggregates (diameters ≥ 3 mm; Klawonn et al., 2015, andStief et al., 2016, respectively), but the static incubation may have facilitated the development of anoxic niches in our incubations by allowing the particles to settle. Activity of anoxic metabolism in these O 2depleted microenvironments is related to the size of the particles/aggregates and the ambient oxygen conditions due to the availability of substrate and the diffusion velocity, respectively. Larger particles in situations of low ambient O 2 levels offer more organic substrates that favour internal O 2 consumption through respiratory activity, facilitating the development of anoxia within the particles core (Ploug et al., 1997(Ploug et al., , 2008Klawonn et al., 2015). Thus, differences in particle size could explain differences in the N-loss activity observed at the oxycline between sampling periods.
Phylogenetic analyses of functional genes indicate that this coastal system maintains a great diversity of pelagic N cycling microorganisms, with heterotrophic denitrifying bacteria, aerobic ammonia oxidizers (either from bacteria and archaea domains), and anammox bacteria related to the Candidatus Scalindua clade, present not only in the oxygendepleted bottom waters but also around the oxycline (O. Ulloa, unpublished data). Furthermore, previous evidence of the presence of anammox cells in the study area, identified by CARD-FISH, showed that this bacterial population could be distributed throughout the water column at similarly higher numbers, when the maximal development of upwelling is reached (Galán et al., 2012).
The presence of O 2 (in situ) in particle-associated incubations from the oxycline during January favours coupling between aerobic mechanisms (e.g. AO and organic matter remineralization) with anaerobic N removal processes (i.e. anammox and denitrification), as illustrated by the high production rates of NO − 2 and NH + 4 (Fig. 5). Assuming that these aerobic processes also occur in association with the particles, it is proposed that aerobic metabolisms occurring at the periphery of the oxygenated particle fuel the internal anaerobic pathways by the diffusion of their products (e.g. nitrite produced by AO) to the respiration-mediated anoxic core, similarly to the scenarios described for various other microbial aggregates in wastewater treatment plants (Nielsen et al., 2005), man-made model systems (Stief et al., 2016) and natural environments (Klawonn et al., 2015). On the other hand, the detection of nitrification during anoxic incubations has been attributed to the persistence of traces of O 2 in the samples (e.g. Bristow et al., 2016;Ganesh et al., 2015), although those studies exhibited caution to O 2 contamination that was similar to this study (see Sect. 2.5 in Methods).
In January, N 2 production by heterotrophic denitrifier communities in the oxycline was higher than during September. This can be mostly accounted for by differences in the availability of reactive organic matter, consistent not only with the rise in Chl-a ( Fig. 1a) but also with the increase in the observed particle size (Fig. 2b, f). Increases in the organic matter availability, measured as particulate organic carbon during the development of the upwelling season in both the oxycline and bottom waters, have been previously reported for this coastal system (Graco et al., 2001;Galán et al., 2014). The influence of organic matter on denitrification activity has been extensively documented in OMZ waters (e.g. Ward et al., 2008Ward et al., , 2009Dalsgaard et al., 2012;Chang et al., 2014). Furthermore, the increased nitrite production rates observed in the oxycline during January (Fig. 5a, b) and the in situ accumulation of this substrate (see profile in Fig. 2g) could also favour denitrification, although the main electron acceptor for this process was NO − 15 NO − 2 and 15 NO − 3 amended treatments (Fig. 3b, d). Similarly to denitrification, anammox rates in the oxycline also increased from September to January. This temporal variability in anammox activity could be related to the availability of NO − 2 and NH + 4 , which is implied by the observation of a vast increase in the production of these substrates during January (Fig. 5b, d). The intense cycling of NO − 2 and NH + 4 via nitrification, anammox, and denitrification, among other processes (e.g. nitrite and ammonium assimilation; processes not determined in this study), should prevent a substantial in situ accumulation of these substrates (Fig. 2g). At depth, N 2 production rates by anammox and denitrification were either reduced or non-existent in comparison to the rate of activity observed in the oxycline. Furthermore, rates were lower that those previously observed in the same area . This reduction in N 2 production throughout the water column, and over time (intra-and interseasonally) is likely to be related to variations in the availability of O 2 and substrates that fuel both processes (i.e. organic matter, NO − 2 , and NH + 4 ). A greater depletion of ambient O 2 could explain the major N removal activity that was previously observed in the area , a scenario that has been widely observed elsewhere (Jensen et al., 2008;Kalvelage et al., 2011;Dalsgaard et al., 2012). On the other hand, vertical variations in the quality of the organic matter that sinks to the depths could affect microbial organoheterotrophic activity, including denitrification rates (Chang et al., 2014). In turn, this affects the delivery of remineralized compounds required for lithoautotrophic processes, such as anammox and nitrification, among others. Likewise, reduced anammox and organotrophic denitrification activity has been previously registered during the summer period at depth and was related to the presence of reduced sulfur compounds in the water column that inhibited the activity of the anammox cells ) and triggered a lithoautotrophic, H 2 S-based metabolism of the denitrifier community .

15 N 2 O production
It can be concluded that in this coastal system during January the main substrate source for N 2 O production changes along chemical gradients through the water column. Thus, N 2 O was produced from NH + 4 within the oxycline, while at depth it was produced from NO − 2 . However, it must be considered that during September, C 2 H 2 , a N 2 O reduction inhibitor after NiR, was not added to the incubations. Conversely, while the NiR pathway is associated with the production of N 2 O, based on the experimental design used here, it is not clear which pathway formed N 2 O coupled to the oxidation of NH + 4 . It is possible that N 2 O was produced by bacteria during the oxidation of NH + 4 to NO − 2 as a side product of the reaction (e.g. Goreau et al., 1980) or through the less understood nitrifier-denitrification reaction (e.g. Wrage et al., 2001). The latter pathway combines two labelled N atoms from NO or NO − 2 and should therefore produce mostly 46 N 2 O in our experiment if these substrates originate from AO, whereas 45 N 2 O dominated (Table 2 and Fig. 4). Ammonium oxidizing archaea may also participate in N 2 O production (Martens-Habbena et al., 2009;Santoro et al., 2010Santoro et al., , 2011Jung et al., 2014). In this sense, the large dominance of 45 N 2 O over 46 N 2 O observed could be explained by the recently described "hybrid" pathway (Kozlowski et al., 2016), in which labelled hydroxylamine ( 15 NH 2 OH) produced from AO might be combined with NO reduced from the NO − 2 present in the native pool of the incubations. The coupling between AO and NiR after the formation of NO − 2 is less plausible, because any labelled NO − 2 that is produced would be diluted by the existing pool of NO − 2 in the environment. Furthermore, experiments carried out with 15 NO − 2 in the oxycline during summer did not produce labelled N 2 O via NiR, despite the use of C 2 H 2 as was observed at depth, rejecting the idea that the coupling between AO and NiR is the pathway through which N 2 O is generated. Increases in N 2 O production within the oxycline have also previously been associated with AO (Codispoti, 2010).
It appears more likely that differences in N 2 O produced in the oxycline between seasons are related to O 2 concentration, rather than to NH + 4 availability, as the production of this substrate was consistently at least 1 order of magnitude higher than the ammonium-based N 2 O production rates (Fig. 4c, d). The O 2 -limited incubations carried out during September, relative to in situ O 2 conditions during the January experiments, could, together with a greater availability of particles occurring at the sampling depth, generate an O 2 -depleted environment, in which the yield of N 2 O formed through nitrifier denitrification Goreau et al., 1980;Lipschultz et al., 1981;de Wilde and de Bie, 2000) or through the "hybrid" pathway (Kozlowski et al., 2016) will increase. There is ample evidence that N 2 O production via nitrification in aquatic systems is increased under conditions of reduced oxygen and abundant ammonium availability McElroy et al., 1978;de Wilde and de Bie, 2000;Löscher et al., 2012).

Cycling of dissolved inorganic 15 N compounds
During January, at least for the duration of the sampling period of this study, bacteria seemed to play a major role in the modulation of AO compared to their archaeal counterpart based on the different effects of ATU and GC7 inhibitors on nitrite production (Fig. 5a, b). This result could also indicate an in situ influence of ammonium oxidizers on the N-loss activity through competition for NH + 4 with anammox cells (Fig. 3), and also possibly through the involvement in N 2 O production. This segregation between AO players was not observed during September, probably owing to differences in O 2 availability during the incubations that could suppress the activity of ammonium oxidizers as previously mentioned. However, a dominance of bacterial AO does not correspond to recent molecular and biogeochemical observations from the study area, particularly during the upwelling season, when a high abundance of Thaumarchaeota (previously Crenarchaeota; an archaeon that sustains itself chemolithoautotrophically through aerobic oxidization of NH + 4 ) account for a significant fraction of the microbial assemblage, while ammonium-oxidizing bacteria are scarce or undetectable (Levipan et al., 2007;Quiñones et al., 2009;Bristow et al., 2016). Likewise, the relative contribution to ammonium oxidation, quantified by the ammonia monooxygenase subunit A gene (Molina et al., 2010), and dark carbon assimilation rates (Farías et al., 2009), showed that archaea had a greater potential activity relative to the homologous bacteria in the system. Also, the Thaumarchaeota are mainly associated with the oxycline (Belmar et al., 2017), where N 2 O hotspots are frequently observed (Cornejo and Farías, 2012;Farías et al., 2015), further suggesting their possible participation in N 2 O production in this layer, particularly as the isotopic signature suggests ammonium oxidation by archaea is a major precursor of N 2 O in the ocean (Santoro et al., 2011). To reconcile the inhibitor results with all other evidence, we speculate that GC7, as a growth inhibitor of archaea, might not affect archaeal metabolism in short-term incubations, thus masking the relative importance of this domain for AO during this study.
During the sampling period, NH + 4 was produced at a high rate by dissimilative nitrite reduction (i.e. DNiRA), a process measured for the first time in this system. Rates were especially high during January in the oxycline. It should be noted, however, that the hypobromite method used to analyse 15 NH + 4 also oxidizes organic 15 N, which may have formed through assimilation of labelled NO − 2 . Nonetheless, it is remarkable that all the NH + 4 required for the lithoautotrophic community, i.e. anammox and nitrification (measured here as AO), and even for previous rates of light assimilation measured at the same station (spring-summer season average < 100 and < 50 nmol L −1 d −1 for 30 and 80 m, respectively; Fernández and Farías, 2012), could be supported by DNiRA at the two depths evaluated. Furthermore, the expected ammonium production from organic matter remineralization by canonical denitrification and nitrate and nitrite reduction, at the rates measured here and assuming that these are heterotrophic processes, should be an important additional source of this substrate. The DNRA process was suggested to be important in some previous surveys in the OMZ off Peru and in the Arabian Sea (Lam et al., 2009;Jensen et al., 2011; but see also Kalvelage et al., 2013) and was suggested to supply anammox with ammonium in previous OMZ studies in which heterotrophic denitrification was not detected (Thamdrup et al., 2006;Hamersley et al., 2007;Galán et al., 2009;Lam et al., 2009). Nevertheless, the NO − 2 required just to support the DNiRA, disregarding the NO − 2 demand from other processes (e.g. anammox and denitrifiers, or for nitrite assimilation and nitrite oxidation -not evaluated in this investigation), was not fully supported by the measured nitrite production through nitrate reduction and ammonium oxidation. This uncoupling between microbial sources and sinks for this substrate indicates that NO − 2 should be a limiting factor and hence a pivotal compound in this system. Thus, allochthonous NO − 2 is needed to cope with this imbalance. Considering this, upwelled NO − 2 from the oceanic OMZ is the most probable source (e.g. Galán et al., 2009), with an extensive vertical distribution starting close to the upper OMZ boundary and reaching concentrations up to 10 µmol L −1 . Alternatively, fluxes from sediments are a less probable source due to the increased consumption of nitrite in this environment (Devol, 2015).

Natural abundance of nitrate isotopes
The enriched δ 15 N-NO − 3 values observed (exceeding the average in the deep Pacific of 5 ‰; Sigman et al., 2000) indicate that nitrate-based biological mechanisms are the main modulators of the isotopic N composition in the subsurface waters over the shelf off central Chile during September and January. The general negative trend between NO − 3 concentration and δ 15 N provides evidence that light isotopes are preferentially used by biologically mediated NO − 3 consuming processes, leaving the residual NO − 3 pool enriched. Likewise, concurrence between the δ 15 N values and the N deficit (except for the surface sample) confirms previous reports (Voss et al., 2001;Sigman et al., 2000). The δ 15 N value (11.2 ‰) at depth during the spring transition is used as an estimate of the background source signal during the onset of the upwelling season. Thus, the progressive δ 15 N enrichment observed throughout the oxycline to the surface may result from locally occurring NO − 3 fractionation processes, including dissimilatory nitrate reduction processes and nitrate assimilation. During January, the association between the maximum δ 15 N-NO − 3 value (up to 14.5 ‰) at depth, the maximum N deficit (Fig. 2h), and a pronounced O 2 deficiency (< 5 µmol L −1 ) is likely to be the result of diverse dissimilative nitrate consumption processes (i.e. nitrate reduction to nitrite, DNRA, and denitrification through N 2 O production). The isotope N ratio was reduced in the oxycline (11.7 ‰; Fig. 2f), possibly due to the addition of isotopically light NO − 3 by the oxidation of recently fixed N, with an isotopic composition of ∼ −2 to 0 ‰ (Liu et al., 1996). Nitrogen fixation in this system occurs mainly during summer and autumn (Fernández et al., 2011), and the process may produce up to 20 % of the new N in surface waters associated with OMZs (Brandes et al., 1998). Based on the natural abundance of 15 N-NO − 3 found, the input of N by fixation seems to also be a minor N source in this system. Considering a starting point of 14.5 ‰ (as in surface during September; Fig. 2d) and assuming that N fixation adds fixed N at 0 ‰ and N from fixation is converted to NO − 3 without fractionation (i.e. all NH + 4 oxidized), 34 % of the NO − 3 would need to be added through fixation. With NO − tion of just 0.7 µmol L −1 of fixed N. In general, our observations are consistent with those reported previously for the Peruvian OMZ (Ryabenko et al., 2012), where high δ 15 N-NO − 3 values (> 10 ‰) were associated with denitrification (and in this case together with nitrate reduction processes), occurring at lower O 2 concentrations (< 5 µmol L −1 ) and higher N deficits. Furthermore, this is consistent with fractionation during nitrate assimilation by phytoplankton in surface waters (O 2 > 200 µmol L −1 ).

Conclusions
Considering that N removal processes are initially fuelled by organic matter and the concomitant production of electron donors (e.g. NH + 4 ) and acceptors (e.g. NO − 2 ) after its remineralization, the vertical and temporal differences in Nloss processes during this study and previous studies in the area  highlight the influence of the development of the upwelling season on the availability and vertical distribution of particulate organic matter. Furthermore, the accumulation of sinking organic particles around the oxycline considerably increases the volume for removing N processes in this coastal system. The great amount of regenerated ammonium produced after the oxidation of this organic load has a central role in regulating the generation of N 2 O at the oxycline. Nonetheless the main substrate that supports the N removal as either producing N 2 and N 2 O is the overabundance of preformed N, mostly in the form of nitrate, which fertilizes the system through the upwelling season. Thus, during this productive period, pulses of accumulation and consumption of different N substrates modulated the structure and activity of the microbial N-based assemblage of this coastal upwelling system. Data availability. Data from hydrographic profiles and from field experiments are available upon request by emailing Alexander Galán (agalan@ucsc.cl).