Intensive nitrogen loss over the Omani Shelf due to anammox coupled with dissimilatory nitrite reduction to ammonium

Abstract

A combination of stable isotopes (¹⁵N) and molecular ecological approaches was used to investigate the vertical distribution and mechanisms of biological N₂ production along a transect from the Omani coast to the central–northeastern (NE) Arabian Sea. The Arabian Sea harbors the thickest oxygen minimum zone (OMZ) in the world's oceans, and is considered to be a major site of oceanic nitrogen (N) loss. Short (<48 h) anoxic incubations with ¹⁵N-labeled substrates and functional gene expression analyses showed that the anammox process was highly active, whereas denitrification was hardly detectable in the OMZ over the Omani shelf at least at the time of our sampling. Anammox was coupled with dissimilatory nitrite reduction to ammonium (DNRA), resulting in the production of double-¹⁵N-labeled N₂ from ¹⁵NO₂⁻, a signal often taken as the lone evidence for denitrification in the past. Although the central–NE Arabian Sea has conventionally been regarded as the primary N-loss region, low potential N-loss rates at sporadic depths were detected at best. N-loss activities in this region likely experience high spatiotemporal variabilities as linked to the availability of organic matter. Our finding of greater N-loss associated with the more productive Omani upwelling region is consistent with results from other major OMZs. The close reliance of anammox on DNRA also highlights the need to take into account the effects of coupling N-transformations on oceanic N-loss and subsequent N-balance estimates.

Introduction

Nitrate is the highest energy-conserving oxidant for respiration after oxygen. In oxygen-deficient seawater, nitrate is generally believed to be reduced in a stepwise manner to N₂O or N₂ (2NO₃⁻ → 2NO₂⁻ → 2NO → N₂O → N₂) for the respiratory oxidation of organic matter. Known as heterotrophic denitrification, this process has been considered an important factor in the global N-loss from the ocean (Naqvi, 1987; Gruber and Sarmiento, 1997; Codispoti et al., 2001), affecting indirectly biological productivity and thus sequestration of atmospheric CO₂ (Capone, 2000; Gruber, 2004). The major oxygen minimum zones (OMZs) in the world, found in the Arabian Sea and the eastern tropical North and South Pacific, are responsible for 20–40% of global N-loss from the oceans, whereas 10–20% is thought to occur in the Arabian Sea (Naqvi, 1987; Codispoti et al., 2001; Gruber, 2004; Devol et al., 2006). Recent findings from incubation experiments with ¹⁵N-labeled compounds indicated substantial N-loss due to anammox (NH₄⁺+NO₂⁻ → N₂+2H₂O) in the OMZs off Peru, Chile and Namibia, whereas denitrification was generally below detection (Kuypers et al., 2005; Thamdrup et al., 2006; Hamersley et al., 2007). These studies strongly question the traditional view that heterotrophic denitrification is the only significant driver of oceanic N-loss.

In the Arabian Sea OMZ, however, N-loss rates have rarely been measured directly. Despite the detection of signature lipid biomarker (Jaeschke et al., 2007) and gene sequences (Woebken et al., 2008) of anammox bacteria in the Arabian Sea OMZ, anammox activities had only been inferred (Devol et al., 2006) until recently. Nicholls et al. (2007) carried out incubation experiments using ¹⁵N-labeled substrates and reported indications of very low anammox rates at three depths in the central–northeastern (NE) Arabian Sea. A few years later, anammox rates were measured in the same region (Ward et al., 2009; Bulow et al., 2010). Interestingly, in the latter study, denitrification was reported to be the dominant and often the sole active N-loss process in most samples investigated. Nonetheless, as pointed out a few decades ago (Richards, 1965), if denitrification is the only active N-loss pathway, as a heterotrophic and presumably the major remineralization process in the OMZs, denitrification should liberate large amounts of ammonium (NH₄⁺) from the oxidation of organic matter. On the contrary, such expected NH₄⁺ accumulations have not been observed in the OMZs (Cline and Richards, 1972; Codispoti and Christensen, 1985; Naqvi, 1987). Thus, the occurrence of anammox, which removes NH₄⁺ while producing N₂, is highly probable.

Anammox has been reported to co-occur with multiple N-cycling processes, including nitrification, nitrate reduction and dissimilatory nitrate/nitrite reduction to ammonium (DNRA), in OMZs off the coasts of Peru and Namibia where denitrification was not detectable (Kartal et al., 2007a; Lam et al., 2009). Coupling of these reactions with anammox may in fact lead to the production of double-¹⁵N-labeled N₂ (¹⁵N¹⁵N) and complicate data interpretation of isotope-pairing experiments. For instance, coupling between ammonia oxidation to nitrite and anammox can generate ¹⁵N¹⁵N in ¹⁵NH₄⁺ incubations in the presence of little ambient NO₂⁻ (Lam et al., 2007; Jensen et al., 2008). It is also theoretically possible to have ¹⁵N¹⁵N produced from the interactions between anammox bacteria and organisms capable of DNRA in ¹⁵NO₂⁻ incubations, with results mimicking the signature of denitrification (Kartal et al., 2007a). In this case, ¹⁵NO₂⁻ can be reduced via DNRA to ¹⁵NH₄⁺, which is then combined with ¹⁵NO₂⁻ at a 1:1 ratio through anammox to generate ¹⁵N¹⁵N (Supplementary Figure S1). In the Arabian Sea OMZ, the occurrence of DNRA has been suggested on the basis of an excess production of single-¹⁵N-labeled N₂ (¹⁴N¹⁵N) that could not be explained by either denitrification or anammox (Nicholls et al., 2007). However, the actual occurrence of DNRA and the presence of DNRA-capable microorganisms in the Arabian Sea remain unexplored.

Unlike the eastern tropical South Pacific where there is perennial upwelling except for El Niño years, biogeochemical processes in the Arabian Sea are heavily influenced by the seasonal reversal of monsoonal winds (Wiggert et al., 2005). Surface production is enhanced by upwelling along the western boundary and to a lesser extent in the central region during the Southwest monsoon (June–August), and by deep convective mixing in the north during the Northeast monsoon (December–February; Marra and Barber, 2005). In the two intermonsoonal periods, conditions in the Arabian Sea approach oligotrophy. Because persistent oxygen-deficient conditions overlap with nitrite accumulations and large N-deficits primarily in the central–NE Arabian Sea, most though not necessarily exclusive N-loss activities have been postulated to occur in this part of the basin (Naqvi, 1991; Bange et al., 2005). In comparison, the mid-water column underlying the productive Omani upwelling zone rarely reaches complete anoxia, and so has not been considered important for N-loss (Naqvi, 1991; Bange et al., 2005). This apparent decoupling of N-loss and surface production in the Arabian Sea contrasts with other OMZs of eastern tropical Pacific and South Atlantic. However, no actual N-loss measurements have been made over the Omani shelf to date.

In this study, we examined the activities and mechanisms of N₂ production in the Arabian Sea OMZ, through a combination of various ¹⁵N-stable-isotope-pairing experiments, expression analyses of biomarker functional genes and the enumeration of anammox bacterial cells. We conducted experiments for six depths through the vertical expanse of the OMZ (100–900 m) at seven stations, encompassing the Omani shelf and the central–NE Arabian Sea, to compare the relative importance of N-loss from these two regions.

Materials and methods

Water sampling and analyses of nutrients, particulate organic carbon and particulate nitrogen

Sampling took place in September/October 2007, just at the beginning of the autumn intermonsoon (Table 1). We followed the cruise track of the former US Joint Global Ocean Flux Study (Morrison et al., 1999; Figure 1). Water sampling was conducted with a conductivity–temperature–depth rosette equipped with 10 l bottles on board the R/V Meteor. The standard oxygen sensor of the conductivity–temperature–depth (Seabird) and an additionally mounted microsensor were used together to identify the OMZ. Besides, we deployed the highly sensitive STOX (Switchable Trace amount OXygen) sensor (Revsbech et al., 2009; detection limit: 90 nM during our deployments) at stations 956–958 in the central–NE Arabian Sea. Water samples were analyzed for NH₄⁺, NO₂⁻, NO₃⁻ and PO₄³⁻ (detection limits 20, 30, 100 and 100 nM, respectively) at 10 to 25-m intervals for 12 stations (Figure 1). NH₄⁺ and NO₂⁻ concentrations were measured immediately after sampling using fluorometric and spectrophotometric techniques, respectively (Grasshoff, 1983; Holmes et al., 1999). Samples for NO₃⁻ and PO₄³⁻ were stored frozen and measured spectrophotometrically (Grasshoff et al., 1999) with an autoanalyzer (TRAACS 800, Bran & Lubbe, Hamburg, Germany) in a shore-based laboratory. Particulate organic carbon and nitrogen were collected by filtering 3–30 l of seawater onto pre-combusted glass fiber filters. Particulate organic carbon and nitrogen were quantified on a flush combustion CNS Carlo Erba 1500 analyzer (Carlo Erba, Milan, Italy) after carbonate removal with 1 M HCl.

Table 1 Summary of ¹⁵N-incubation experiments. Numbers in parentheses in the last column refer to the final concentrations of the ¹⁵N- and ¹⁴N-species added into the incubations

Figure 1

Distributions of (a) density, (b) oxygen and (c) N^* along the cruise track from the Omani coast to the central–northeastern Arabian Sea. Gray-filled circles denote sampling depths. Maps show sampling area, sampling locations and station numbers along our cruise track. Station numbers in red indicate sampling stations where ¹⁵N-incubation experiments were conducted.

¹⁵N-isotope pairing experiments

Time-series ¹⁵N-incubation experiments were carried out to determine N-loss rates with various ¹⁵N- and ¹⁴N-amendments, following previously described protocols (Dalsgaard et al., 2003; Kuypers et al., 2005). Different combinations of ¹⁵N-labeled and unlabeled NH₄⁺ and/or NO₂⁻ were added to helium-purged seawater samples, and dispensed into Exetainer vials (Labco, High Wycombe, Buckinghamshire, UK) for incubation (Table 1). Purging with helium removed O₂ to below the detection limit of an oxygen microsensor (⩽0.5 μM; MPI, Bremen, Germany). Samples were incubated at in situ temperature and in the dark for about 0, 6, 12, 24 and 48 h. Isotopic compositions and concentrations of the N₂ and N₂O produced were analyzed on a gas chromatography-isotope ratio mass spectrometer (Fisons VG Optima, Fison, Manchester, UK). Afterwards, net DNRA rates were determined as the net accumulation of ¹⁵NH₄⁺ from the same ¹⁵NO₂⁻ incubations. The isotopic composition of NH₄⁺ was analyzed after chemical conversion into N₂ with alkaline hypobromite (NaOBr; Warembourg, 1993). Further details are provided in Supplementary SI Text.

All rates were derived from linear regressions of ¹⁵N-production (N₂, N₂O or NH₄⁺) as a function of time, and only productions remaining linear without lag-phase and with slopes significantly greater than zero (one- or two-tailed t-tests, P<0.05) were reported (unless otherwise stated; detection limits: 150–200 pM per day, 0.5 nM per day and 0.5 nM per day for N₂, N₂O and NH₄⁺, respectively). Anammox and denitrification rates were calculated from the regression slopes and the mole fractions of ¹⁵N in substrate pools (Thamdrup and Dalsgaard, 2002; Thamdrup et al., 2006; Supplementary SI Text). In the case of DNRA–anammox coupling, total anammox rates were calculated in a similar manner as in nitrification–anammox coupling (Jensen et al., 2008; Supplementary SI Text).

Molecular ecological analyses

Anammox bacteria were enumerated via catalyzed reporter deposition–fluorescence in situ hybridization (CARD–FISH; Pernthaler et al., 2002) with 16S rRNA targeted probes BS820/BS820c (Hamersley et al., 2007; Supplementary SI Text). For simultaneous DNA and RNA extractions, seawater samples (10–15 l) were filtered through Sterivex filters (0.22 μm pore size, Millipore GmbH, Schwalbach, Germany) and stored at −80 °C until extraction (Supplementary SI Text). The biomarker functional genes for anammox (Scalindua-type cd₁-nitrite reductase or nirS), denitrification (denitrifier-nirS) and DNRA (cytochrome c nitrite reductase, or nrfA) were quantified along with their expressions (transcription as mRNA) via real-time PCR, as previously described (Lam et al., 2009). Detailed protocols are provided in the Supplementary SI Text and Table S1. Clone libraries were also constructed for the expressed Scalindua-type and denitrifier-nirS, and the complementary DNA inserts were sequenced for phylogenetic analyses (Supplementary SI Text). Active gene transcription in unmanipulated seawater samples may provide independent support for an active process detected via manipulated incubation experiments, though the relationships between rates and gene expressions are not necessarily straightforward. This is because the different measurement types have vastly different detection limits, and that gene expression is also affected by physiological states, stresses and other (post)transcriptional factors. As a lot remains unexplored in the immense oceanic microbiome, we do not claim an exhaustive coverage of the functional gene targets by the selected primers.

Results and discussion

Physicochemical settings

Consistent with previous observations (Morrison et al., 1999), the water column at the time of our sampling became more strongly stratified from the western boundary towards the central basin (Figures 1a and b). Enhanced surface chlorophyll a, lower sea surface temperatures and shoaling of isopycnals westwards indicate residual upwelling near the Omani shelf (Figure 1a, Supplementary Figure S2). Oxygen concentrations declined sharply below ∼25–50 m, and reached <10 μM at about ∼85–100 m. This oxygen-deficient OMZ extended to ∼1000 m depth (Figure 1b). Using the highly sensitive STOX sensor at stations 956–958, apparent anoxic conditions (<90 nM) were found between ∼110 and 800 m water depths (Supplementary Figure S3). Strong deficits of combined inorganic nitrogen relative to phosphate (PO₄³⁻; Redfield et al., 1963), shown here as strongly negative N^* (in μM=[NO₃⁻+NO₂⁻+NH₄⁺]−16 × [PO₄³⁻]+2.9 μmol kg⁻¹ × density in kg l⁻¹; Gruber and Sarmiento, 1997), were the most pronounced in the upper part of the central–NE OMZ (Figure 1c), overlapping with the strong secondary nitrite maximum (Lam et al., 2011). Meanwhile, contrary to previous reports (Naqvi, 1991; Bange et al., 2005), N-deficits were also observed over the Omani shelf from 200 m water depth to the seafloor, coincident with apparently anoxic conditions (<1–2 μM O₂ measured by a microsensor; Figures 1 and 2b), indicating past and/or present substantial N-loss from these waters.

Figure 2

Vertical distribution of (a) potential density anomaly (σ_θ) and N^*, (b) oxygen, (c) inorganic nitrogen, (d) ¹⁴N¹⁵N and ¹⁵N¹⁵N production rates (mean±s.e.) in various ¹⁵N-incubations, (e) anammox bacterial abundance (mean±s.e.) measured by CARD–FISH and the expression (mean±s.e.) of anammox (Scalindua-type) and denitrifier cd₁-containing nitrite reductase nirS genes, and (f) DNRA rates (mean±s.e.) as net ¹⁵NH₄⁺ production from ¹⁵NO₂⁻ and expression of the corresponding biomarker functional gene, nrfA (mean±s.e.)—all measured at station 946.

High N-loss via anammox over the Omani Shelf

Indeed, significant N-loss activity was detected at the shelf stations (944 and 946). In both incubations with ¹⁵NH₄⁺ and ¹⁵NO₂⁻, the production of single ¹⁵N-labeled N₂ (¹⁴N¹⁵N) occurred right from the start and increased linearly with time over 48 h (for example, Figure 3). At the upper four sampled depths at station 946, and 120 m at station 944, all of the produced ¹⁵N-labeled N₂ in the ¹⁵NH₄⁺ incubations was recovered as ¹⁴N¹⁵N, whereas no production of ¹⁵N¹⁵N was detected (Figures 2d and 3a, Supplementary Figure S4a). Such exclusive production of ¹⁴N¹⁵N without delay clearly indicated active occurrence of anammox. In parallel, there was comparable production of ¹⁴N¹⁵N in the ¹⁵NO₂⁻ incubations, further verifying anammox activity (Figures 2d and 3b).

Figure 3

Examples of the linear production of ¹⁵N-labeled N₂ during incubations (a) with ¹⁵NH₄⁺ and (b) with ¹⁵NO₂⁻. (c) DNRA measured as linear production of ¹⁵NH₄⁺ in the same ¹⁵NO₂⁻-incubations. Production rates of ¹⁴N¹⁵N, ¹⁵N¹⁵N and ¹⁵NH₄⁺ were calculated from the slopes in linear regression analyses. Data came from 125 m at station 946 over the Omani shelf.

Similar to the findings in the Peruvian and Chilean OMZ (Thamdrup et al., 2006; Hamersley et al., 2007), anammox rates determined from both ¹⁵N-incubations were the highest in the upper OMZ over the shelf (∼2.8 nmol of N₂ l⁻¹ per day at station 944; 21–39 nmol of N₂ l⁻¹ per day at station 946), coincident with local maxima of NH₄⁺ and NO₂⁻ (Figures 2c; Supplementary Table S2). Anammox rates dropped considerably to ∼1 nmol of N₂ l⁻¹ per day below 200 m at station 946, where NO₂⁻ returned to low levels. Anammox bacterial abundance, as measured via CARD–FISH, showed a similar vertical distribution to anammox rates (Figure 2e,d) (Spearman R=0.766, P<0.05 and R=0.755, P<0.05 for rates from ¹⁵NH₄⁺ and ¹⁵NO₂⁻ incubations, respectively). Expression of the anammox functional gene biomarker, Scalindua-nirS, which encodes for the anammox (Scalindua)-specific cytochrome cd₁-containing nitrite reductase (Strous et al., 2006), was detectable at most depths examined, with the highest level found in the upper OMZ (Figure 2e, Supplementary Figure S5). Phylogenetic analyses revealed that the Arabian Sea expressed Scalindua-nirS sequences formed two tight clusters with some South China Sea sediment clones, whereas they were clearly distinct from the Peruvian OMZ cluster (71–90% similarity) and the cultured Candidatus Scalindua sp. T23 (81–91%; Supplementary Figure S6). This pattern is consistent with the 16S rRNA-based phylogeny (Woebken et al., 2008). In fact, with support from internal transcribed spacer sequence analyses, Arabian Sea anammox bacteria seem to form a novel species Candidatus Scalindua arabica (Woebken et al., 2008) that cannot be targeted by the oligonucleotide Amx820 (Schmid et al., 2000), commonly used to quantify anammox bacteria in environmental studies (for example, Ward et al., 2009).

Interestingly, there was substantial production of ¹⁵N¹⁵N without lag in the ¹⁵NO₂⁻ incubations for most depths at station 946, which is normally taken as an evidence for concurrent denitrification activity (Figures 2d and 3b). The ¹⁵N¹⁵N production rates decreased with depth, from 8.7 nmol of N₂ l⁻¹ per day at 100 m to 0.41 nmol of N₂ l⁻¹ per day at 300 m (Figure 2d). Nevertheless, if denitrification and anammox co-occurred, anammox bacteria likely depended on denitrification for the released NH₄⁺ from organic matter. Consequently, the proportional contribution of anammox to total N₂ production would be constrained by the stoichiometric ratios of organic matter remineralized during denitrification (Dalsgaard et al., 2003). Assuming a C:N:P ratio of 106:16:1 in the organic matter (Redfield et al., 1963), anammox should in theory account for 29% of the total N₂ production (Supplementary SI text), or ⩽50% of total N₂ production in the extreme case when only proteins were remineralized (Van Mooy et al., 2002; Dalsgaard et al., 2003). Instead, anammox contributed to 41–77% of total N₂ production in our samples, suggesting that additional or alternative pathways were at play. The expression of the denitrifier-nirS, a biomarker functional gene encoding cd₁-containing nitrite reductase specific for denitrifiers, was not detectable at most depths at station 946 (Figure 2e). Neither detectable was the production of ¹⁵N-N₂O from ¹⁵NO₂⁻, further pointing to only low, if any, denitrification activity. Similarly, although ¹⁵N¹⁵N production was detected at 120 m at station 944, it only occurred after a 14-h lag (Supplementary Table S1), whereas denitrifier-nirS expression and ¹⁵N-N₂O were not detected.

DNRA–anammox coupling

Substantial ¹⁵NH₄⁺ production from the ¹⁵NO₂⁻ pool indicates high net DNRA rates over the Omani shelf, with ⩽40±5.5 and 3.55±0.40 nmol of N l⁻¹ per day measured at station 946 (Figures 2d,f and 3c) and station 944 (at 120 m with ∼5 h lag), respectively, where high anammox rates were also measured (Supplementary Table S2–3). These net DNRA rates were in the upper range of those reported for the Peruvian OMZ (Lam et al., 2009). Further evidence for DNRA came from the significantly correlating (Spearman R=0.538, P<0.01) expression of nrfA, the biomarker functional gene encoding the enzyme cytochrome c nitrite reductase that is essential in the DNRA reaction (Figure 2f, Supplementary Figure S5c, f). Together, the relatively high net DNRA rates, the consistently detectable active expression of nrfA and Scalindua-nirS versus the sporadically detected low denitrifier-nirS expression, strongly suggest that the ¹⁵N¹⁵N production resulted from the reduction of ¹⁵NO₂⁻ to ¹⁵NH₄⁺, followed by a one-to-one combination of the produced ¹⁵NH₄⁺ and the added ¹⁵NO₂⁻ through anammox (Supplementary Figure S1). Although heterotrophic denitrification is thermodynamically more favorable than DNRA, a recent study using pure cultures of denitrifying and ammonifying (DNRA capable) bacteria revealed that the former conserved less energy and so had lower growth yields than ammonifying bacteria (4.6–6 g biomass per mol of NO₂⁻ by denitrifiers versus 8.4 g by ammonifiers; Strohm et al., 2007). Hence, DNRA and anammox bacteria (growth yield 1.3–2 g biomass per mol e⁻-acceptor; Strous et al., 2006) acting together could have an energetic advantage over denitrifiers in the competition for substrates in oceanic OMZs.

Both measured DNRA rates and nrfA expression increased with depth in the upper OMZ at station 946, corresponding to a decrease in ambient NH₄⁺ levels. The ¹⁵N mole fraction of NH₄⁺ resulting from DNRA is estimated to increase from 24% at 100 m to 61% at 150 m, which reflects an increase in the proportion of anammox activity fueled by DNRA (Supplementary SI text). In other words, when ambient NH₄⁺ is low and DNRA is highly active, the generated NH₄⁺ by DNRA becomes more important for anammox. A tight DNRA–anammox coupling can then produce ¹⁵N¹⁵N from ¹⁵NO_x⁻, a signal that could easily be mistaken as a signature of denitrification only (Kartal et al., 2007a). Meanwhile, deeper in the water column where NH₄⁺ was scarce and no N-loss activity was detected, the calculated NH₄⁺ turnover (generation) due to DNRA alone was rather short (2–3 days). This might indicate additional in situ NH₄⁺ assimilation, or large advective losses due to stronger current regimes over shelf regions that have not been simulated in our incubation experiments, whereas substrate stimulation in DNRA rate measurements were unlikely given the high availability of ambient nitrate.

Nevertheless, at two out of six depths (100 and 200 m) at station 946, DNRA–anammox coupling was insufficient to fully explain the observed % of ¹⁵N¹⁵N in total N₂ production from ¹⁵NO₂⁻. The observed 23 and 74% at 100 and 200 m, respectively, were double those of the expected 10 and 34% based on DNRA and anammox rates and the ambient NH₄⁺ concentrations (Supplementary SI Text). Part of this ¹⁵N¹⁵N production could probably be explained by intracellular DNRA–anammox coupling by anammox bacteria themselves, which has not been accounted for in the above calculations. Such a capability has been shown in cultures (Kartal et al., 2007a) though it awaits verification in environmental studies. Alternatively, it might be partly attributable to denitrification, which would be in concordance with the detectable denitrifier-nirS expression at 150 and 200 m. Sequence analyses of these expressed denitrifier-nirS, as amplified with the primers nirS1F-6R (Braker et al., 1998) and subsequently cloned, revealed relatively low diversity (only six OTUs based on 97% nucleotide sequence cutoff) out of the 87 obtained sequences (Supplementary Figure S6), in contrast with the high diversity observed at the DNA level in a nearby region (Jayakumar et al., 2009). Most sequences were affiliated with Paracoccus denitrificans and Roseobacter denitrificans, whereas three OTUs were more closely related to other environmental sequences, including some previously retrieved from the Arabian Sea.

Low and sporadic N-loss in the central–NE OMZ

In the central–NE Arabian Sea OMZ (stations 950, 953, 955 and 957), contrary to the expected high N-loss activities (Naqvi, 1991; Bange et al., 2005), N₂ production was mostly undetectable except for perhaps 1–2 sporadic depths at each station (Supplementary Figure S4 c–f). Only at station 949, just outside the prominent secondary nitrite maximum, was anammox detected right below the oxycline (150 m) as significant ¹⁴N¹⁵N productions in all four sets of incubations: ¹⁵NH₄⁺, ¹⁵NO₂⁻, ¹⁵NH₄⁺+¹⁴NO₂⁻ and ¹⁵NO₂⁻+¹⁴NH₄⁺ (Supplementary Figure S4b, Supplementary Table S2). The detectable N₂ production at the other central–NE stations only appeared in the form of ¹⁴N¹⁵N, and lacked consistency over the four different ¹⁵N-treatments (Supplementary Figure S4 c–f). Although experiments with ¹⁵NH₄⁺ alone almost never yielded any ¹⁵N-labelled N₂ to unambiguously confirm anammox activity, there was no production of ¹⁵N¹⁵N from ¹⁵NO₂⁻ (±¹⁴NH₄⁺) without initial time lag to verify in situ denitrification either. When detected, the rates (⩽1 nmol of ¹⁴N¹⁵N l⁻¹ per day) were close to the detection limit, and were often found with both substrates (NH₄⁺ and NO₂⁻) added. Hence, these N-loss rates in the central–NE Arabian Sea should be regarded as potential rather than in situ rates. Anammox potentials were further corroborated by the consistently detectable expression of Scalindua-nirS (Supplementary Figure S5), and the presence of anammox bacteria identified via CARD–FISH, though their cellular abundance was too low to give reliable microscopic cell counts (<0.1% total microbial abundance). These results are consistent with another recent study conducted during the same intermonsoon period, when patchy anammox activity was reported at two stations nearby (Ward et al., 2009; Bulow et al., 2010). DNRA–anammox coupling was not evident in the central–NE basin at the time of our sampling. DNRA rates were assessed at two central–NE stations (957 and 955), and were only detectable at two depths (200 m at station 955 and 900 m at station 957) with very low rates near detection limit.

Despite the greater abundance of denitrifiers over anammox bacteria detected at the gene level, which agreed with the recent study in the central–NE Arabian Sea (Ward et al., 2009), there was only low to undetectable active expression of the dentrifier-nirS gene (Supplementary Figure S5). Although the primers (Michotey et al., 2000; Throbäck et al., 2004) used for quantification (quantitative PCR) may not cover all environmental nirS sequences and thereby underestimated denitrifiers at the gene level, these primers have been tested to exhibit high specificity (Throbäck et al., 2004) and cover the vast majority of the expressed complementary DNA sequences except for two OTUs (six sequences) retrieved in our study. In contrast, the primer nirS3R (Braker et al., 1998) used to quantify denitrifiers by Ward et al. (2009) has sequence mismatches with all complementary DNA sequences obtained in our study, as well as with a fair number of cultured strains (Braker et al., 1998) and published environmental sequences (for example, 67 out of the 113 denitrifier sequences shown in Supplementary Figure S6), including some retrieved from the Arabian Sea. If the stringency of quantitative PCR conditions is relaxed to cover more sequences, this primer may additionally bind to a second region 18 bases downstream and result in double or unspecific quantification. Therefore, the primer nirS3R is unsuitable and not chosen for use in quantitative PCR. Overall, our gene expression analyses showed little indication of actively expressed denitrifier-nirS in these waters, albeit new designs for more universal nirS-targeting primers suitable for quantitative PCR, or the use of PCR-independent methods, such as metagenomics/metatranscriptomics, are necessary for a more comprehensive appraisal.

Altogether, the lack of readily detectable denitrifier-nirS expression at most stations and depths in the central–NE Arabian Sea, was in line with the absence of any significant ¹⁵N¹⁵N production detected from ¹⁵NO₂⁻ incubations, as well as the lack of measurable ¹⁵N-labeled N₂O production. Interestingly, in the incubations with ¹⁵NO₂⁻ (±¹⁴NH₄⁺) at these stations, an abrupt N₂ production often occurred after 24–50 h time lag (Supplementary Figure S7). This sharp delayed N₂ production included both ¹⁴N¹⁵N and ¹⁵N¹⁵N. Although denitrification might be inhibited by ⩾2–4 μM of O₂ (Devol, 1978) and anammox by ⩾13 μM (Jensen et al., 2008), <90 nM of O₂ was detected in the core of the central–NE OMZ in situ with the STOX sensor, and ⩽0.5±0.09 μM O₂ measured in the exetainers with an oxygen microsensor in the current and previous studies (<0.2 μM by Dalsgaard et al., 2003 and Jensen et al., 2008). Hence, oxygen inhibition of N-loss processes at sub-micromolar levels can most likely be discounted. Instead, although reflecting an apparent artifact after prolonged incubation and not counted as rates here, the lagged ¹⁵N¹⁵N production from ¹⁵NO₂⁻ (±¹⁴NH₄⁺) likely indicated stimulated denitrification after long incubation hours, and would agree with the observed high gene abundance but inconsistent expression of denitrifier-nirS. Similarly, time lags could be observed in the recent study by Ward et al. (2009) that was further detailed in Bulow et al. (2010). In that study, there was no significant ¹⁵N-labeled N₂ production from ¹⁵NO₂⁻ nor from ⁴⁶N₂O in the first 24 h of incubation, at least for the representative sampled depth shown and for which their highest denitrification rate (25±9.1 nmol of N₂ l⁻¹ per day) was reported. In other words, despite the persistent denitrification potentials present in the central–NE Arabian Sea OMZ, there is limited evidence for their in situ activities. However, active denitrification cannot be excluded at other times, as it is likely subjected to strong spatiotemporal variabilities.

Distribution and variability of N-loss in the Arabian Sea

Considering the entire basin of the Arabian Sea, our measured anammox rates ranged from undetectable to 27±1.1 nmol of N₂ l⁻¹ per day (from ¹⁵NH₄⁺ incubations), or up to 38±1.9 nmol of N₂ l⁻¹ per day taking into account the coupling with DNRA (from ¹⁵NO₂⁻ incubations; Table 2). When integrated over the depth range of the OMZ, the highest N-loss due to anammox occurred over the Omani shelf at a rate of up to 4.5±0.4 mmol of N m⁻² per day (Figure 4), comparable with those reported for the Namibian and Peruvian shelves (Kuypers et al., 2005; Hamersley et al., 2007). Together with previous findings of substantial anammox, but low to undetectable denitrification in other marine oxygen-deficient waters (Kuypers et al., 2005; Thamdrup et al., 2006; Hamersley et al., 2007; Jensen et al., 2008), the current study confirms the importance of anammox in global oceanic N-loss. At just about 500 km offshore, the depth-integrated N-loss was reduced by two orders of magnitude, and down to only potential rates in the central–NE basin (Figure 4). Even if the measured potential rate profiles were integrated for the central–NE OMZ, there would be at most 0.3–0.6 mmol of N m⁻² per day N-loss. This finding of greater N-loss associated with the more productive upwelling waters toward the coast, on one hand coheres with observations from other OMZs that are also linked to upwelling systems (Kuypers et al., 2005; Hamersley et al., 2007). On the other hand, it contradicts with the conventional presumption of high N-loss in the central–NE Arabian Sea deduced from nitrite accumulation or N-deficits (Naqvi, 1991; Bange et al., 2005), and the moderate denitrification rates measured previously (Ward et al., 2009; Bulow et al., 2010).

Table 2 Summary of volumetric and depth-integrated rates of anammox and denitrification in various studies of oceanic OMZs.

Figure 4

Estimates of depth-integrated anammox rates (±s.e.) measured in Arabian Sea OMZ in this study. Rates are calculated from incubations with ¹⁵NO₂⁻, and only rates that showed consistency across various isotope combinations have been included. The dashed line marks the approximate geographical extent of the prominent secondary nitrite maximum (⩾0.5 μM) as the postulated major denitrification region in the Arabian Sea (Naqvi, 1991).

The fact that high N-loss rates were obtained over the more productive Omani shelf waters suggested a close link between N-loss and the availability of organic matter. Indeed, our measured N-loss rates were significantly and positively correlated with surface particulate organic carbon (Spearman correlation R=0.50, P<0.0001) and particulate nitrogen (Spearman R=0.53, P<0.0005; Supplementary Figure S8). Both anammox and denitrification have to rely on organic matter, either indirectly for the source of remineralized NH₄⁺ by the former, or directly for sources of carbon and electrons by the latter or both (Kartal et al., 2007b). Thus, organic matter would be a logical controlling factor for N-loss via either pathway. This is consistent with a study from the eastern tropical South Pacific where surface primary production and not oxygen appeared to be the primary factor regulating N-loss (Lipschultz et al., 1990). Surface production in the Arabian Sea is heavily influenced by the seasonal monsoons, and nutrients are delivered to the open ocean via mesoscale eddies more than open-ocean Ekman pumping (Marra and Barber, 2005; Wiggert et al., 2005). Consequently, high spatiotemporal variabilities are especially apparent in the central–NE region, as reflected in the patchiness in surface chlorophyll a during our sampling period (Supplementary Figure S2; Lam et al., 2011). Over just a 2-month period around our sampling, a high degree of week-to-week patchiness could be seen especially in the central–NE Arabian Sea; whereas the effects of upwelling persisted westward until early October 2007 (Supplementary Figure S2b). The recent study by Ward et al. (2009) sampled in close proximity to stations 950 and 955 at 1 week after and before this study, respectively. They reported moderate N-loss rates through mainly denitrification. Although their observations might not have truly differed much from ours if the same criteria were used for rate calculations (for example, exclusion of data series with significant initial time lags), the disparities between the two studies may also be attributed to the high spatiotemporal variabilities in the Arabian Sea. The detected N-loss activity might reflect N-loss in response to episodic pulses of organic matter sinking from surface water. Therefore, the spatiotemporal variabilities of N-loss and N-cycling in the Arabian Sea certainly require further assessment.

Conclusions and perspectives

Our current study reports for the first time a direct link between DNRA and anammox in the ocean through a production of ¹⁵N¹⁵N from ¹⁵NO₂⁻—signal easily mistaken as a signature for denitrification only. This DNRA–anammox coupling mediates the highest N-loss rates reported at this time in the Arabian Sea, and was found in a hitherto overlooked region over the productive Omani shelf. By contrast, only low potential rates at best were detected in the presumed active N-loss zone in the central–NE OMZ. Despite being the world's thickest oceanic OMZ, the calculated depth-integrated N-loss from the Arabian Sea OMZ was only a fraction of that in the Peruvian OMZ (Table 2). This suggests that the common estimate of the Arabian Sea accounting for ∼20% of global oceanic N-loss is perhaps too high, and that the strongly negative N^*, as a time-integrated signal, might reflect an accumulation of low or episodic N-loss within a poorly ventilated water layer in the central–NE basin (Lam et al., 2011). Although we might have only captured a snapshot in space and time, if N-loss is indeed intimately linked to organic matter availability, considering the Omani upwelling being the most productive waters in the basin on an annual basis, the detected N-loss therein might represent a major contribution to the yearly N-budget that has so far been ignored. The observed DNRA–anammox coupling may also occur at the eastern and northern boundaries of the Arabian Sea, where the OMZ impinges on the Indian and Pakistani shelves underlying some relatively productive waters at least on a seasonal basis. These shelf regions may together be responsible for significant total N-loss that needs to be taken into N-budget calculations. Further studies with higher temporal and spatial resolutions are necessary to elucidate the true spatial extent of N-loss, and how exactly N-loss and the overall N-cycling respond to seasonality and short-term variabilities in the Arabian Sea, and whether there is a dominance or succession of N-loss mechanisms. Besides, our results highlight the need to examine possibly concurrent N-cycling processes together with N-loss in oxygen-deficient environments, as these concurrent processes may complicate reliable estimates of total N-loss.