Oxygen Sensitivity of Anammox and Coupled N-Cycle Processes in Oxygen Minimum Zones

Nutrient measurements indicate that 30–50% of the total nitrogen (N) loss in the ocean occurs in oxygen minimum zones (OMZs). This pelagic N-removal takes place within only ∼0.1% of the ocean volume, hence moderate variations in the extent of OMZs due to global warming may have a large impact on the global N-cycle. We examined the effect of oxygen (O2) on anammox, NH3 oxidation and NO3 − reduction in 15N-labeling experiments with varying O2 concentrations (0–25 µmol L−1) in the Namibian and Peruvian OMZs. Our results show that O2 is a major controlling factor for anammox activity in OMZ waters. Based on our O2 assays we estimate the upper limit for anammox to be ∼20 µmol L−1. In contrast, NH3 oxidation to NO2 − and NO3 − reduction to NO2 − as the main NH4 + and NO2 − sources for anammox were only moderately affected by changing O2 concentrations. Intriguingly, aerobic NH3 oxidation was active at non-detectable concentrations of O2, while anaerobic NO3 − reduction was fully active up to at least 25 µmol L−1 O2. Hence, aerobic and anaerobic N-cycle pathways in OMZs can co-occur over a larger range of O2 concentrations than previously assumed. The zone where N-loss can occur is primarily controlled by the O2-sensitivity of anammox itself, and not by any effects of O2 on the tightly coupled pathways of aerobic NH3 oxidation and NO3 − reduction. With anammox bacteria in the marine environment being active at O2 levels ∼20 times higher than those known to inhibit their cultured counterparts, the oceanic volume potentially acting as a N-sink increases tenfold. The predicted expansion of OMZs may enlarge this volume even further. Our study provides the first robust estimates of O2 sensitivities for processes directly and indirectly connected with N-loss. These are essential to assess the effects of ocean de-oxygenation on oceanic N-cycling.


Introduction
Oxygen (O 2 ) is one of the key regulatory factors of major biogeochemical cycles in the marine environment [1]. The distribution of dissolved O 2 in the world's oceans is regulated by gas exchange between surface waters and the lower atmosphere, advective processes within the ocean, as well as the biological processes of photosynthesis and respiration. Oxygen, entering the ocean interior mainly at high latitudes, is distributed throughout the global ocean via thermohaline circulation. In the ocean's sunlit surface layer, phytoplankton produces O 2 and fixes carbon dioxide (CO 2 ) in to biomass. Near the base of the euphotic zone, concentrations of O 2 are generally at their lowest as photosynthesis diminishes or ceases altogether while the repiration of sinking organic matter by heterotrophic micro-organisms consumes O 2 at maximal rates.
Subsurface regions of severely reduced O 2 concentrations (O 2 #5 mmol L 21 ), the so-called oxygen minimum zones (OMZs), are found along the eastern boundaries of the ocean basins in the subtropics and tropics (e.g. off California, Namibia, Peru/Chile) and in the Arabian Sea. Typically in these regions, wind-driven circulation results in the upwelling of nutrient-rich deep waters, fueling high primary production in the euphotic zone. The high surface productivity results in high export of organic matter and thus strong respiration in subsurface waters. Combined with the poor ventilation of these water masses [2,3], this leads to permanently O 2 -depleted to anoxic conditions at mid-depths [4][5][6].
Although OMZs (if defined by O 2 #5 mmol L 21 ) account for only ,0.1% of the global ocean volume [7], they play a key role in controlling the oceans' nutrient inventory as 30-50% of the oceanic nitrogen (N) loss is estimated to occur therein [7,8]. The recharge of such N-deficient waters from these regions back to adjacent surface waters limits primary production and thus carbon (C) sequestration in large parts of the tropical oceans. N-loss as primarily the formation of gaseous dinitrogen (N 2 ) can occur via two pathways: (1) heterotrophic denitrification, the reduction of nitrate (NO 3 2 ) to gaseous dinitrogen (N 2 ) via a sequence of intermediates (NO 3 2 RNO 2 2 RNORN 2 ORN 2 ) and (2) anammox, the anaerobic oxidation of ammonium (NH 4 + ) with nitrite (NO 2 2 ) to N 2 . In the OMZs of Namibia and Peru/Chile, on which the current study focuses, anammox has been identified as the major N-loss pathway based on 15 N-labeling experiments, whereas heterotrophic denitrification was often not detectable or only measured sporadically [9][10][11].
In the course of global climate change and increasing anthropogenic pressures on the marine environment, coastal and open ocean OMZs have been expanding and intensifying in the last decades [12,13]. A continuing decline in dissolved O 2 due to reduced O 2 solubility and enhanced stratification [14], as well as coastal and open ocean eutrophication [15,16], is expected. Deoxygenation will have the greatest effect on water masses already deficient in O 2 as these are often at or near the thresholds for anaerobic processes such as anammox or denitrification. Deutsch et al. [17] calculated that a reduction of the mean upper ocean O 2 content by only 1% would mean a doubling of water masses with O 2 #5 mmol L 21 , thus significantly enlarging the ocean volume potentially affected by N-loss.
However, the sensitivities of anammox and denitrification to changes in dissolved O 2 and their upper O 2 limits in the marine environment are largely unknown. N-loss attributed to denitrification has been reported to occur at up to 20 mmol L 21 of O 2 [18]. Nonetheless, direct measurements of denitrification under controlled exposure to low O 2 concentrations in OMZs are lacking. Active anammox bacteria have been found to be abundant at O 2 concentrations up to 9 and 20 mmol L 21 in the Namibian and Peruvian upwelling systems, respectively [9,10], and it has been suggested that marine snow aggregates could provide suitable anoxic micro-niches at ambient O 2 concentrations up to 25 mmol L 21 [19,20]. Off Peru/Chile the measured anammox rates were often the highest at the base of the oxycline and in the upper OMZ [10,11,21], likely associated with intensified remineralization of organic matter in these water layers. This further indicates that, unlike their cultured counterparts, which are inhibited at O 2 concentrations as low as 1 mmol L 21 [22], marine anammox bacteria can tolerate O 2 concentrations higher than the upper O 2 limit (5 mmol L 21 ) often used to restrict anaerobic processes in biogeochemical models [23]. Recently, Jensen et al. [24] investigated the O 2 sensitivity of anammox in the near-anoxic zone of the Black Sea water column and showed that anammox bacteria remained active up to ,9 mmol L 21 of O 2 . Still unknown is whether this relatively high O 2 tolerance is widespread amongst anammox bacteria in the major OMZs of the world's oceans.
Although anammox is an autotrophic process, it relies on other N-cycling processes for the required reactive substrates NO 2 2 and NH 4 + , e.g. NH 3 oxidation to NO 2 2 and heterotrophic nitrate (NO 3 2 ) reduction to NO 2 . The co-occurrence of these aerobic and anaerobic processes together with anammox requires them to be adapted to a certain overlapping range of O 2 concentrations. Thus far, it remains unclear whether or not processes coupled to anammox can proceed in the same range of O 2 as assumed for anammox (0-20 mmol L 21 ), or if they show different O 2 sensitivities that might hence restrict N-loss to a narrower O 2 regime. Under anoxic conditions, NO 3 2 is the next thermodynamically favored electron acceptor, which can be used by a variety of micro-organisms to oxidize organic matter [25]. In OMZ waters, secondary NO 2 2 maxima are often interpreted as active NO 3 2 reduction [26,27]. The formation of NO 2 2 from NO 3 2 is the first step in both denitrification and dissimilatory nitrate reduction to ammonium (DNRA), but it can also be considered as a stand-alone process, as more micro-organisms are known capable of reducing NO 3 2 to NO 2 2 than to N 2 or NH 4 + [25,28]. Heterotrophic NO 3 2 reduction to NO 2 2 has been measured at high rates in the Peruvian OMZ [29,30], and has been estimated to account for approximately two thirds of the NO 2 2 required for anammox in this region [30]. At the same time, NO 3 2 reduction also provides an important source of NH 4 + released from oxidized organic matter [30,31]. Lipschultz et al. [29] investigated the effect of varying O 2 concentrations on NO 3 2 reduction to NO 2 2 in the Peruvian OMZ. They observed that NO 3 2 reduction rates doubled under anoxic conditions (N 2 atmosphere) compared to in situ conditions (2.5 mmol L 21 of O 2 ), while rates decreased by ,75% at 20 mmol L 21 of O 2 .
When O 2 is present, NO 2 2 can be produced aerobically by NH 3 oxidizing bacteria and archaea in the first step in nitrification. Rates of NH 3 oxidation are generally highest near the upper OMZ boundaries [32,33]. In the Peruvian OMZ, this is also where anammox bacteria are most active [10]. These bacteria are partly fueled by NH 3 oxidation in this zone [30]. A similarly tight coupling between anammox and NH 3 oxidation was shown earlier for the Black Sea [34]. The occurrence of NH 3 oxidizers is, however, not restricted to the upper OMZ. They have been found active at non-detectable concentrations of O 2 (,1-2 mmol L 21 ) in the core of OMZs [30,33,35] and are thus obviously well adapted to near-anoxic O 2 conditions. When Lipschultz et al. [29] investigated the O 2 sensitivity of NH 3 oxidation in the Peruvian OMZ, the inferred de-oxygenation of the samples only caused a ,50% decrease in activity relative to ambient O 2 (2.5 mmol L 21 ), whereas no stimulation was achieved by an increase to ,20 mmol L 21 of O 2 .
With anammox as well as NO 3 2 reduction being apparently tolerant to relatively high O 2 and NH 3 oxidation being apparently able to cope with severe O 2 depletion, an expansion of OMZs might indeed drive larger water masses to greater N-deficits. This would potentially exacerbate N-limitation of primary production in large parts of the ocean and thus affect the oceans' capacity to attenuate the rising atmospheric CO 2 . However, at present no study has systematically investigated the O 2 sensitivities of anammox and concurrent N-cycling processes in oceanic OMZs, and thus the future nutrient balance in these regions remains speculative at best.
In this paper, we present results for the Namibian and Peru/ Chile upwelling systems, two of the most productive regions in the worlds' oceans associated with massive N-loss, where we explored the effect of O 2 on anammox, NH 3 oxidation and NO 3 2 reduction throughout the OMZ.

Ethics Statement
The necessary permissions were obtained from the governments of Namibia and Peru to carry out research in their waters.

Water sampling and nutrient analyses
Samples were taken on two cruises to the OMZs off Namibia (M76/2) and Peru (M77/3), where upwelling persists year-round, onboard R/V Meteor in May/June 2008 and December/January 2008/2009, respectively ( Fig. 1). A pump-CTD system was used to collect water samples just below the oxycline, through the core of the OMZ, down to ,375 m depth off the coast of Peru. The pump CTD system was equipped with a conventional amperometric O 2 micro-sensor to obtain vertical profiles of dissolved O 2 . In addition, the recently developed STOX (Switchable Trace amount OXygen) sensor [6], which allows high-accuracy O 2 measurements in near-anoxic environments (detection limit: 50-100 nmol L 21 during our deployments), was deployed. At least five measuring cycles after $10 min sensor equilibration at a given sampling depth were used to calculate O 2 concentrations. Water samples were taken with a depth resolution of 1-2 m for nutrient analyses. NH 4 + was measured fluorometrically [36] and NO 2 2 was analyzed spectrophotometrically [37] on board. Water samples for NO 3 2 and PO 4 32 were stored frozen until spectrophotometric determination [37] with an autoanalyzer (TRAACS 800, Bran & Lubbe) in a shore-based laboratory. Detection limits for NH 4 + , NO 2 2 , NO 3 2 and PO 4 32 were 10, 10, 100 and 100 nmol L 21 , respectively. N-deficits were calculated from the measured fixed inorganic N-and PO 4 32 concentrations as N* (in mmol L 21 ) following Gruber and Sarmiento [8] Fig. 1 and Table 1). Based on O 2 profiles, three to six depths per station were chosen for a standard series of 15 Nlabeling experiments. The experimental procedure for 15 Nlabeling experiments has been described in detail previously [9,31,38]. Briefly, N-loss by either anammox or heterotrophic denitrification was measured as the production of 15 [39], or to N 2 by sulfamic acid [40,41]. Rates were calculated from the slope of linear regression of 15 N-production as a function of time. Only significant and linear production of 15 N-species without an initial lag-phase was considered (t-tests, p,0.05; R 2 .0.8). The net production rates presented here have been corrected for the mole fractions of 15 N in the original substrate pools but not for isotope dilution due to any other concurrent N-consumption or production processes in the course of the incubation.

Oxygen sensitivity experiments
In order to determine the effect of varying O 2 concentrations on N-cycle processes, one to two depths per station were sampled for additional O 2 sensitivity experiments. Samples were taken from the upper OMZ, where aerobic and anaerobic N-cycle processes have been shown to co-occur [30], except one sample taken deeper in the core of the Peruvian OMZ (St. 36). Samples were obtained in 250-mL serum bottles and purged with helium (He) for approximately 15 min to remove any initial O 2 and to lower the N 2 background in order to enhance the detection limit of 29 N 2 and 30 N 2 [38]. As a small sample volume was lost during He-purging, the bottles were then refilled with a second He-purged sample from the same depth to avoid headspace. Afterwards, air-saturated water from the same depth was added to the serum bottles in exchange for part of the de-oxygenated water to adjust samples to the desired O 2 concentration. At St. 206 and 252 (Namibian OMZ) three samples each were One sample, to which no airsaturated water was added, served as an anoxic control at all stations. After additions of either 15  2 , 14 N-species were added to all experiments to exclude substrate limitation, which would otherwise complicate the interpretation of any O 2 effects on the processes of interest. Moreover, keeping the 14 N-pool of the product of a certain reaction well above the expected concentrations produced from the added 15 N-substrate could minimize any further conversion of the newly formed 15 Nproducts by co-occurring processes. The rate measurements for the various processes were carried out as described above. To exclude formation of 29 N 2 due to coupled nitrification-denitrification in incubations amended with 15 NH 4 + we added allylthiourea (ATU; final concentration 84 mmol L 21 ) to an additional sample of the highest O 2 treatment (,11.5 mmol L 21 ) at St. 206 and 252. ATU is a specific inhibitor of aerobic NH 3 oxidation [42][43][44] and does not affect anammox activity shown at least in sediments [45]. Two sets of incubations were performed in parallel at St. 206 and 252 and one sample per time-point was sacrificed to measure dissolved O 2 . For the remaining stations, O 2 concentrations were determined only for the initial time-point in each 15 N-incubation experiment. We used a custom-built, fast-responding O 2 micro-sensor (Clark-type; MPI Bremen) for most measurements (detection limit: ,0.5 mmol L 21 of O 2 ), except at St. 206 where a STOX sensor was used for selected samples.

Data analysis
We applied least-squares fitting to each set of samples of the O 2 sensitivity experiments using Excel's solver function [46].

Hydrochemistry in the Namibian OMZ
The water column was poorly stratified over the Namibian shelf at St. 206 and 252 during the time of sampling, as indicated by a weak density gradient, along with the vertical profiles of dissolved O 2 and inorganic N-species ( Fig. 2A). At both stations O 2 declined gradually with depth, from ,200 mmol L 21 in the surface waters to less than 10 mmol L 21 at ,80 m. STOX measurements at the incubation depths revealed O 2 concentrations as low as 0.6060.11 mmol L 21 at St. 206. In the central OMZ at St. 252 (Table 1), the sensor was at its detection limit (100 nmol L 21 of O 2 during M76-2). Ammonium concentrations were typically in the range of 1-3 mmol L 21 in the oxic zone (,80 m) and decreased to 0.1-0.5 mmol L 21 at the base of the oxycline (Fig. 2B). Towards the sediment-water interface NH 4 + concentrations increased up

N-cycling in the Namibian and Peruvian OMZs
Distribution of anammox activity. Over the Namibian shelf a strong increase in the N-deficit was observed below the oxycline. Minimum values for N* (down to 219 mmol L 21 ) were found in the central OMZ, suggesting N-loss therein. We measured 15  incubations, the formation of 15 N-labeled N 2 was attributed to anammox activity and not denitrification. At both stations, anammox rates and N-loss inferred from N* increased with depth (Fig. 2C)

Oxygen sensitivity of anammox and coupled N-cycle processes
Oxygen sensitivity of anammox. Anammox activity, as indicated by 15   Oxygen sensitivity of nitrate reduction to nitrite. Nitrate reduction rates in the O 2 sensitivity assay carried out for the Namibian OMZ waters, decreased with increasing O 2 concentrations ( Table 2). The incubation experiments at St. 206 revealed a stronger negative response to elevated O 2 levels than those performed at St. 252. Activity at St. 206 was reduced to ,30% of the anoxic control in the highest O 2 treatment (7.3 mmol L 21 ), whereas a doubling of the O 2 concentration (14.7 mmol L 21 ) led to a decrease in NO 3 2 reduction rates to ,60% of the control experiment at St. 252 (Fig. 3B).
In the Peruvian OMZ, production of 15 Figure 3B).
Oxygen sensitivity of ammonia oxidation. Rates of NH 3 oxidation to NO 2 2 showed no significant difference over the range of the applied O 2 concentrations (,1-12 mmol L 21 ) in the Namibian OMZ samples ( Table 2). Activity varied by a maximum of ,15% among the different O 2 treatments but without any systematic trends (Fig. 3C).
Similar to the observations for the Namibian shelf, 15 (Fig. 3C).

Oxygen sensitivity of anammox in OMZ waters
In the investigated samples from both the Namibian and Peruvian OMZ, the only N 2 -forming pathway detected by 15 Nlabeling experiments was anammox. This confirms the results from earlier studies, which detected N-loss due to anammox but not denitrification in these regions [9][10][11]. The highest anammox rates (on the order of 500 nmol N L 21 d 21 ) were measured in the Namibian shelf waters. Off Peru, rates declined from ,50 nmol N L 21 d 21 over the shelf to ,10 nmol N L 21 d 21 at the open ocean sites. This may be explained by differences in surface productivity between the two upwelling systems [47] as well as between Peruvian coastal and open-ocean waters, since organic matter transport ultimately fuels all processes delivering NH 4 + and NO 2 2 for the anammox reaction [30,31]. Anammox often showed the highest rates in the upper OMZ, as seen in previous studies [10,11,21] probably in response to the high NH 4 + release from the enhanced remineralization of particulate organic matter at the base of the oxycline, below which all three activities decreased with depth. There were exceptions, however, particularly at depths close to the seafloor on the shelf, where exceptionally high rates were likely supported by NH 4 + diffusing out of the sediment [9,48,49] (S. Sommer, pers. comm.).
In the O 2 tolerance assays, N-loss due to anammox was in fact detectable at O 2 levels significantly higher (up to ,15 mmol L 21 ) than that generally used to define OMZs (,5 mmol L 21 of O 2 ). Anammox activity in samples taken at the shallow sites appeared the least affected by increasing O 2 . The rates therein remained measurable even at adjusted O 2 concentrations of 10 to 15 mmol L 21 . These are almost twice as high as the anammox O 2 -tolerance level previously determined in the Black Sea suboxic zone [24]. In comparison, anammox activity appeared increas- correlation between the measured rates and adjusted O 2 levels, the upper O 2 limit for anammox to proceed in the OMZs is estimated to be ,20 mmol L 21 (Table 3 & Fig. 3). The apparently higher O 2 tolerance at the shelf stations may be explained by an adaptation of anammox bacteria to fluctuations in dissolved O 2 due to the presence of a less stable oxycline at the upper boundary of the OMZ. Vertical mixing is usually enhanced in coastal upwelling regions. This was indicated by a weak density gradients and a gradual O 2 decline over the Namibian shelf, where the level of dissolved O 2 are known to be variable [50]. In the open-ocean off Peru, ventilation of the OMZ from above is hindered due to strong stratification [51]. The dissolved O 2 content is perhaps most stable within the core of the OMZ, where the highest O 2 sensitivity of anammox was measured in our current study (180 m at St. 36). With O 2 concentrations consistently below 1-2 mmol L 21 , anammox bacteria thriving therein are unlikely to have adapted to higher O 2 levels compared to their counterparts in more dynamic environments.
Alternatively, marine snow particles have been speculated to provide ''anoxic'' micro-environments in which O 2 is sufficiently depleted to favor N-loss at ambient O 2 levels ,25 mmol L 21 [9,20], while some anammox bacteria have been shown to be potentially particle-associated in the Namibian OMZ [20]. Hence, higher abundance of particles in coastal waters than further offshore or in the core of the OMZ might also explain the apparently higher O 2 tolerance by anammox bacteria near the coast.

Oxygen sensitivity of nitrate reduction in OMZ waters
The reduction of NO 3 2 to NO 2 2 , was detected at high rates at the shallow shelf stations both off Namibia and Peru (,100 to 360 nmol L 21 d 21 ) and decreased with increasing distance from the coast in the Peruvian OMZ (,10 to 50 nmol L 21 d 21 at St. 36). The rates measured off Peru are consistent with earlier results from 15 N-labeling experiments in the same region [29,30] and a similar rate distribution was recently reported for the Arabian Sea OMZ [52,53].
Reduction of NO 3 2 to NO 2 2 showed a high degree of variability in O 2 sensitivity amongst stations. No effect of increasing O 2 on NO 3 2 reduction was observed in the 120 m incubations at St. 36. At the remaining stations, the correlation between activity and adjusted O 2 concentrations was non-linear and could be best described by an exponential function, as determined by least-squares fitting (Table 3 & Fig. 3b). Our results from two shelf stations in the Namibian (St. 252) and Peruvian (St. 62) OMZs further confirmed earlier observations by Lipschultz et al. [29] that NO 3 2 reduction was only moderately affected by increasing O 2 . About 50% of NO 3 2 reduction activity remained when O 2 was adjusted to ,14 to 17 mmol L 21 in our abovementioned samples (Table 3). More pronounced sensitivity to O 2 was detected at St. 206 on the Namibian shelf and at 180 m at St. 36 off Peru, where rates were reduced by ,50% relative to the control already at ,4 mmol L 21 of O 2 .
The observation, that in general NO 3 2 reduction activity was only moderately affected by increasing concentrations of O 2 may at first seem at odds with the fact that NO 3 2 respiration is generally considered an anaerobic process. However, it has been reported from experiments with cultures and environmental samples that complete or partial denitrification can take place under aerobic conditions [54][55][56]. Moreover, the different enzymes involved in the step-wise reduction on NO 3 2 to N 2 during denitrification, differ in their O 2 sensitivity. In various bacterial strains the NO 2 2 and nitrous oxide (N 2 O) reductase appear to be most sensitive with respect to O 2 , whereas the NO 3 2 reductase is the most O 2 -tolerant enzyme [57][58][59]. This O 2 tolerance could explain the observation that even the highest O 2 additions did not lead to a full inhibition of NO 3 2 reduction in the samples taken from the Namibian and Peruvian OMZ waters. However, the detected variability in terms of O 2 sensitivity among the different incubation experiments and the lack of any response at 120 m at St. 36 remains puzzling. One possible explanation might be the high phylogenetic diversity and thus variable physiology of the NO 3 2 reducers inhabiting the OMZ waters [30,60].

Oxygen sensitivity of ammonia oxidation in OMZ waters
Ammonia oxidizing activity seemed widespread throughout the OMZ overlying the Namibian shelf, as indicated by high NO 2 2 production rates. Off Peru, nitrifying activity peaked at the base of the oxycline, where the highest NH 4 + release due to remineralization of sinking organic matter can be expected. Though O 2 was not always detectable in situ, NH 3 oxidation rates could be detected at these upper OMZ depths, consistent with previous studies [30,33,35].
In the O 2 sensitivity assays, NH 3 oxidation at most decreased slightly in the anoxic control (St. 54) when compared to the higher O 2 treatments. No stimulation at higher O 2 levels (20 to 25 mmol L 21 of O 2 ) was achieved. A similar observation was made by Lipschultz et al. [29], though they detected a 50% reduction of activity in their assumedly anoxic control. Our results suggest a relatively high O 2 affinity of aerobic NH 3 oxidizers in both OMZs investigated. It has been shown that cultured bacterial NH 3 oxidizers, including marine nitrifiers, are, in principle, able to cope with very low O 2 concentrations down to at least ,2 mmol L 21 [61][62][63]. The only cultured marine aerobic ammonia oxidizing archaea investigated so far appears to have a limited capacity to survive under near anoxic conditions [64]. However, a higher O 2 affinity of archaeal NH 3 oxidizers in the environment is indicated by results from the Peruvian OMZ, which suggest that both bacterial and archaeal NH 3 oxidizers are active at undetectable in situ O 2 levels (,1.5-2 mmol L 21 ) [30].
Based on our findings, the minimum O 2 concentration for NH 3 oxidizer to be active in OMZ waters is most likely in the nanomolar range. An adaptation of aerobic micro-organisms to extremely low O 2 has been shown in a recent study by Stolper et al. [65]. They demonstrated aerobic growth in a culture experiment at an O 2 concentration #3 nmol L 21 . Alternatively, when O 2 is scarce, NH 3 oxidizer may also grow anaerobically via the oxidation of NH 3 with gaseous nitrogen dioxide (NO 2 ) or tetraoxide (N 2 O 4 ) [66]. However, as these compounds are rare in the marine environment, it is unlikely that this is of major ecological significance.

Implications for N-loss in the future ocean and our understanding of N-cycling in modern OMZs
In summary, the current study shows that O 2 is a major controlling factor for anammox activity in OMZ waters. Based on our O 2 assays we estimate the upper limit for anammox to be ,20 mmol L 21 O 2 , which is significantly higher than previously shown for the Black Sea (Table 3 & Fig. 3). In contrast, NH 3 oxidation and NO 3 2 reduction as the main NH 4 + and NO 2 2 sources for anammox were little or only moderately affected by changing concentrations of dissolved O 2 . Intriguingly, aerobic NH 3 oxidation was active at non-detectable O 2 concentrations, while NO 3 2 reduction to NO 2 2 , which is generally considered to be an anaerobic process, was fully active up to at least 25 mmol L 21 O 2 . Hence, aerobic and anaerobic N-cycle pathways in OMZs can co-occur over a larger range of O 2 concentrations In mmol L 21 . Calculated from regression functions obtained by least-squares fitting of the data given in Table 2 than previously assumed. The zone where N-loss can occur is primarily controlled by the O 2 -senstivity of anammox and not by the O 2 -senstivity of the tightly coupled aerobic NH 3 oxidation and anaerobic NO 3 2 reduction. Additionally, our results indicate that N-loss and other Ncycling processes within such O 2 regimes would be controlled by other environmental factors such as substrate availability. For instance, the (near) anoxic conditions in the core of the OMZ do not confer the highest NO 3 2 reduction and anammox rates despite the ideal O 2 regime. Surface water productivity and therewith export of particulate organic matter into the OMZ might play an important role in controlling anammox activity. Sinking organic matter is the ultimate source of the required reactive substrates NO 2 2 and NH 4 + for anammox and it may also provide suitable anoxic micro-environments for anammox bacteria in zones of higher ambient O 2 [9,20].
The fact that anammox in the marine environment can proceed at O 2 levels ,20 times higher than those known to inhibit enrichment cultures of anammox bacteria (,1 mmol L 21 ) [22] enlarges the global oceanic volume potentially affected by N-loss from the previously estimated 0.1% tenfold to ,1% (O 2 #20 mmol L 21 ) [67]. In addition, recent reports show that OMZs have been expanding and intensifying worldwide, particularly in the tropical Atlantic and Pacific [13]. Such expansions of the OMZs would mean an even greater increase in ocean volume potentially subject to active N-loss processes in the coming years. In other words, progressively more fixed inorganic N may be removed from the oceans, and larger areas in the subtropics and tropics might experience enhanced N-limitation due to the recharge of N-deficient waters back to the surface in the future. In the long run, negative feedbacks might also ensue from increasing N-loss and ocean warming. Less productive surface waters would export less organic matter to subsurface waters and lead to reduced O 2 consumption rates. The stronger stratification due to the warming of the upper ocean might also hamper upwelling of nutrient-rich water to the surface, therewith reducing export production and the respiration of O 2 in OMZs.
The relative significance of these positive and negative feedback mechanisms, or how they may counteract each other and eventually influence global oceanic nutrient budgets, would require further investigations complemented with realistic global biogeochemical modeling. To date, the models used to develop future scenarios of the global ocean nutrient balance have rarely taken into account coupling N-cycling processes, and certainly not their respective O 2 sensitivities.
In light of the above presented results, the simple switching from aerobic to anaerobic respiration at ,5 mmol L 21 of O 2 often implemented in models [23] appears not realistic. The current study provides the first robust estimates of O 2 sensitivities for processes directly and indirectly connected with N-loss. These factors are necessary for biogeochemical models to collectively and accurately assess the effects of ocean de-oxygenation on N-cycling in OMZs and neighboring water masses, and hence global oceanic N-balance.