An axenic strain of the marine pelagic, chain-forming, diatom S. marinoi (CCMP 1332) was obtained from the Provasoli-Guillard National Center for Marine Algae and Microbiota (NCMA; formerly CCMP). This strain was formerly referred to as S. costatum and has previously been studied for its intracellular NO3− storage capacity (Kamp et al., 2011). The diatoms were cultured in F/2 medium plus silicate (Guillard and Ryther, 1962) prepared with filtered (0.45 μm) and autoclaved Baltic Sea water (salinity adjusted to 30). The cultivation temperature was 14°C, the light:dark cycle was 10:14 h, and the light intensity was 20 μmol photons m^-2 s^-1. S. marinoi was frequently checked for possible contaminations with bacteria by careful phase-contrast microscopy and by plating out subsamples of the cultures on nutrient agar plates. Contamination of the cultures was not detected at any point during the present study.

Skeletonema marinoi was investigated for possible pathways of dissimilatory NO3− reduction by ¹⁵N- stable isotope labeling of its intracellular NO3− pool and following the time course of the intracellular ¹⁵NO3− concentration and the extracellular ¹⁵NH4+, ¹⁵NO2−, N₂O, and ¹⁵N₂ concentrations in axenic S. marinoi cultures.

To prepare the S. marinoi cells for the experiment, their non-labeled intracellular NO3− (i.e., intracellular ¹⁴NO3−) was replaced with ¹⁵N-labeled intracellular NO3− (i.e., intracellular ¹⁵NO3−). The cells were washed three times with sterile NaCl solution (salinity 30; 5 min; 600 g) to remove ¹⁴NO3− from the growth medium, flushed with N₂ for 30 min to remove O₂ (the O₂ concentration was followed with optode spots; SensorSpot, Pyroscience, Germany), and incubated for 24 h in NO3−-free F/2 medium plus silicate, to make the cells use up their intracellular ¹⁴NO3−. After this “starvation procedure,” the cells were washed again and grown for 12 h in F/2 medium plus silicate, in which ¹⁴NO3− was replaced with ¹⁵NO3− (50 μmol L^-1; 98 atom %, Cambridge Isotope Laboratories), for subsequent accumulation of intracellular ¹⁵NO3−.

For the experimental incubation, the concentrations of NH4+ (added to meet assimilation requirements) and sodium acetate (added as an electron donor) were adjusted to 100 μmol L^-1 each in the growth medium and the S. marinoi culture was split in two.

(a) One half of the culture was transferred into a dark, gas-tight glass bottle, flushed with He for 30 min to introduce dark and anoxic conditions, thoroughly mixed, and distributed into 24 (+3) replicate 6-mL gas-tight incubation vials (Labco, UK) wrapped in aluminum foil. At time intervals of 0.5, 1, 1.5, 2, 3, 4, 5, and 6 h, a He headspace of 3 mL was set in each of three incubation vials, and the diatoms in the remaining 3 mL were killed with 100 μL ZnCl₂ (50% w/v). The vials were stored upside-down at room temperature until measurement of ¹⁵N₂ and N₂O on a gas chromatography-isotopic ratio mass spectrometer (GC-IRMS; 184 Thermo Delta V Plus, Thermo Scientific; for details see Dalsgaard et al., 2012; Stief et al., 2016) and a gas chromatograph (GC 7890, Agilent Technologies), respectively. The cell suspensions collected during setting the headspace were filled into centrifugation tubes and centrifuged (5 min; 600 g). To calculate the intracellular NO3− concentration, 25 μL of the well-mixed pellet was diluted 1:10 in NaCl solution (salinity 30) plus 4% formaldehyde for diatom cell counting in a Fuchs-Rosenthal counting chamber. The remaining pellet and 100 μL of the cell-free supernatant were separately frozen at -20°C for (intracellular) NO3− analyses with an NO_x analyser (CLD 66s, EcoPhysics; for details see Kamp et al., 2011, 2013). The remaining cell-free supernatant was frozen at -20°C for measurement of NH4+, ¹⁵NH4+, and ¹⁵NO2− concentrations (for details see Kamp et al., 2013; Stief et al., 2016). S. marinoi cells taken from three additional incubation vials were carefully investigated for bacterial contamination (see above) at the end of the experiment. No contamination was detected.

(b) The second half of the culture was kept under light and oxic conditions and sub-sampled at time intervals of 0, 1, 2, 3, 4, 5, and 6 h in triplicates each for the measurement of intracellular NO3− and extracellular ¹⁵NH4+ and total NH4+ concentrations.

Since the NO_x analyzer does not discriminate between ¹⁴NO3− and ¹⁵NO3−, and the intracellular ¹⁴NO3− pool of S. marinoi might not have been completely depleted prior to the incubation with ¹⁵NO3− (see “starvation procedure”), intracellular ¹⁵NO3− was also measured with the cadmium/sulfamic acid assay in three samples taken at t₀ (McIlvin and Altabet, 2005; Füssel et al., 2012). Together with the data obtained from the NO_x analyzer (i.e., ¹⁴⁺¹⁵NO3−), the ¹⁵NO3− data were used to calculate the labeling fraction of the intracellular NO3− pool at t₀, which amounted to 52.5 ± 0.7%. This initial labeling fraction was assumed to remain constant throughout the experimental incubation because S. marinoi does not replenish the intracellular NO3− pool under anoxia (Supplementary Figure S1) and isotope fractionation during passive NO3− leakage from diatom cells is very unlikely. The initial labeling fraction of the intracellular NO3− pool was therefore used to convert the measured ¹⁵N-concentrations to intracellular NO3−-derived N-concentrations (denoted as ^ICN) by dividing the ¹⁵N-concentrations by 0.525.

Free-living (axenic) S. marinoi cells as well as S. marinoi cells in diatom-bacteria aggregates were investigated for their NO3− storage capacity under light and oxic conditions and its correlation to the extracellular NO3− concentration, i.e., the NO3− concentration in the surrounding seawater. In total, three batches of free-living S. marinoi cells and four batches of aggregate-associated S. marinoi cells were investigated at different extracellular NO3− concentrations.

Free-living S. marinoi cells were cultivated in F/2 growth medium as described above, washed in nitrate-free NaCl solution (salinity 30), and then adjusted to 0, 10, 15, 25, 50 (n = 2), 60, 100 (n = 2), and 500 μmol L^-1 NO3− for 12–24 h, after which the intracellular NO3− content of the diatom cells was measured in up to three replicates as described above.

For the production of diatom-bacteria aggregates, 50 mL of a stationary-phase S. marinoi culture was mixed with 550 mL natural Baltic Sea water (salinity adjusted to 30) and filled into four glass bottles, and sealed without bubbles. These aggregate-production bottles were mounted on a plankton wheel (diameter: 60 cm) and continuously rotated to induce aggregate formation and to always keep the aggregates sinking (Stief et al., 2016). After aggregates had formed (2–3 days), the water in the four aggregate-production bottles was adjusted to 15, 75, 100, or 350 μmol L^-1 NO3− and the bottles were rotated on the plankton wheel for another 24 h. The aggregates were then individually harvested from the aggregate-production bottles with a glass tube, sized along the three axes with a ruler, and transferred into a centrifugation tube together with water from the aggregate-production bottle still adhering to the aggregate. After the aggregate had settled, 100 μL of the supernatant was sampled and immediately frozen at -20°C until NO3− analysis (see above). The remaining water was carefully removed without destroying the aggregate. The centrifugation tubes were vigorously mixed to arrive at a homogenous suspension. A subsample of the aggregate suspension was taken for diatom cell counts (see above), and the remaining aggregate suspension was frozen in liquid nitrogen, exposed to three freeze-thaw cycles to extract intracellular NO3− (Heisterkamp et al., 2012), and stored at -20°C until NO3− analysis (see above). The intracellular NO3− concentration in the aggregate-associated diatom cells was calculated from the NO3− concentrations in the supernatant and the extracted aggregate suspension, the diatom cell counts, and the aggregate volume (i.e., 3.5–6 mm³ for the four batches). The average cell volume of S. marinoi of 0.33 pL was taken from Kamp et al. (2011).

Diatom-bacteria aggregates formed under oxic conditions in a light:dark cycle were investigated for the consumption of intracellular NO3− after sudden shifts to dark and anoxic conditions, thereby mimicking conditions for an aggregate sinking through oxygen depleted waters.

Aggregates were produced in two separate batches (‘batch 1’ and ‘batch 2’) as described above and under light and oxic conditions, thereby mimicking aggregate formation and intracellular NO3− accumulation in the euphotic zone. Aggregates of ‘batch 1’ were pre-incubated in Baltic Sea water (salinity adjusted to 30, ¹⁴NO3− adjusted to 75 μmol L^-1) for 24 h. For the experimental incubation, seawater was flushed with N₂ to remove O₂ and filled into 6-mL incubation vials (Labco, UK). Aggregates of ‘batch 2’ were pre-incubated in 100 μmol L^{-1 15}NO3− for 12 h (see next section). For the experimental incubation, seawater was flushed with He to remove O₂ and to lower the N₂ background and filled into 6-mL incubation vials. Ammonium and acetate were not added because organic matter mineralization inside the aggregates was expected to cover the possible demands for N-assimilation and C-dissimilation by diatoms and bacteria. Single aggregates were harvested from the pre-incubation bottles, sized as described above, and transferred into the incubation vials. Incubation vials were sealed, wrapped in aluminum foil, and mounted on the plankton wheel (except for the t₀ samples) to keep the aggregates sinking. At time intervals of 0, 3, 6, 24, and 48 h, four incubation vials each were sacrificed and the samples were processed as described above for NO3− analysis in aggregate-associated diatom cells and whole aggregates.

The conversion of intracellular vs. extracellular NO3− to intermediates and products of dissimilatory NO3− reduction inside sinking diatom-bacteria aggregates was investigated with a ¹⁵N-stable isotope experiment.

Diatom-bacteria aggregates were produced as described above in natural Baltic Sea water (salinity adjusted to 30), but without adding ¹⁴NO3− above the natural background concentration of 15 μmol NO3− L^-1. Instead, 100 μmol L^{-1 15}NO3− was added to the aggregate-production bottle 24 h prior to the experimental incubation to allow diatoms to accumulate ¹⁵NO3− intracellularly. Ellipsoidal aggregates of 6.1 ± 2.1 mm³ formed within 3 days (aggregate ‘batch 2’). Aggregates of this batch were also analyzed together with the aggregates of ‘batch 1’ with respect to storage and consumption of intracellular NO3− (see previous section).

To remove all extracellular ¹⁵NO3− and to make intracellular ¹⁵NO3− the sole ¹⁵NO3− source in the experiment, aggregates were carefully washed in nitrate-free NaCl solution (salinity 30). After washing, the aggregates were sized with a ruler, and transferred into 6-mL incubation vials that were previously filled with anoxic Baltic Sea water (salinity adjusted to 30). Ammonium and acetate were not added for reasons given in the previous section. The incubation vials were sealed, wrapped in aluminum foil, and mounted on the rotating plankton wheel (except for the t₀ samples) to keep the aggregates sinking.

(a) At time intervals of 0, 3, 6, 24, 48, and 72 h (diatom-free) water and intact aggregates were taken from three incubation vials each, frozen in liquid nitrogen to stop all metabolic activities and stored at -20°C until analysis of intracellular ¹⁵NO3− and extracellular NO3−, NH4+, and NO2− (see above).

(b) At parallel time intervals of 3, 6, 24, and 72 h, a He headspace of 2.5 mL was set in three additional incubation vials each (note that this was done in additional incubation vials because setting the headspace tends to destroy the aggregates). The remaining sample volume of 3.5 mL was amended with 100 μL ZnCl₂ (50% w/v). The incubation vials were stored upside-down at room temperature until measurement of ¹⁵N₂ and N₂O concentrations (see above). The cell suspensions that were collected during setting the headspace were centrifuged (5 min; 600 g) and the supernatants were used for measuring extracellular ¹⁵NH4+ and ¹⁵NO2− concentrations.

The labeling fraction of the initial intracellular NO3− pool of the diatoms was determined as described above and in this case was 37.2 ± 9.3% (n = 3). Assuming that this labeling fraction remains constant during the incubation (see above), the measured ¹⁵N-concentrations of products (i.e., ¹⁵NO2−, ¹⁵NH4+, and ¹⁵N₂) were converted to intracellular NO3−-derived N-concentrations (i.e., ^ICNO2−, ^ICNH4+, and ^ICN₂) by dividing the ¹⁵N-concentrations by 0.372. Extracellular, NO3−-derived N-concentrations of NO2− and NH4+ (i.e., ^ECNO2− and ^ECNH4+) were calculated from the changes in measured total concentrations (i.e., ^TOTNO2− and ^TOTNH4+) minus the calculated ^ICN-concentrations. ^TOTN₂-concentrations were derived by using the principles of random isotope pairing, i.e., ^TOTN₂ = (²⁹N₂)²/(4 × ³⁰N₂) + ²⁹N₂ + ³⁰N₂ (Nielsen, 1992); and ^ECN₂-concentrations were calculated as described for ^ECNO2− and ^ECNH4+. Application of the isotope pairing technique is based on the assumptions that (a) denitrification is the only significant N₂-producing process in diatom aggregates (Stief et al., 2016) and (b) ¹⁵NO3− and ¹⁴NO3− are uniformly mixed at the site where denitrification takes place. The latter may not hold in the aggregates, if ¹⁵NO3− is uniformly released in the anoxic center, while ¹⁴NO3− decreases along a radial gradient into the center. This would result in an underestimation of ^TOTN₂ and thereby of ^ECN₂ concentrations.

The S. marinoi cells used in this experiment had stored intracellular NO3− at a concentration of 13.9 ± 0.4 mmol L^-1 (mean ± SD, n = 3). The time course of total intracellular NO3− and intracellular NO3−-derived ^ICNO2−, ^ICNH4+, and ^ICN₂ concentrations in axenic S. marinoi cultures after the sudden shift to dark and anoxic conditions revealed that NH4+ was the only product arising from intracellular NO3− consumption (Figure 1; Supplementary Table S1). A time-integrated mass balance calculation revealed that 90% of the intracellular NO3− was converted to ^ICNH4+ during the incubation period of 6 h. The initial ^ICNH4+ concentration of 3.3 μmol L^-1 is likely due to the production and release of ^ICNH4+ during the time period needed to shift the diatom culture to dark and anoxic conditions. Taken together, the data shows that S. marinoi is able to perform DNRA under dark and anoxic conditions (but see restrictive interpretation in the Introduction). Furthermore, the temporal coincidence of intracellular NO3− consumption and ^ICNH4+ production rules out the possibility of NO3− assimilation, followed by protein degradation and subsequent NH4+ release. Since ^ICNO2− and ^ICN₂ production were not observed, S. marinoi seems incapable of dissimilatory NO3− reduction to NO2− (DNRN) as a stand-alone process, denitrification, and anammox.

Under light/oxic conditions, the intracellular NO3− concentration decreased only slightly during the 6 h incubation (-0.06 ± 0.03 fmol NO3− cell^-1 h^-1 under light/oxic conditions vs. -0.36 ± 0.04 fmol NO3− cell^-1 h^-1 under dark/anoxic conditions; linear regression), probably due to assimilation and/or leakage of intracellular NO3−. In contrast to anoxic conditions, NH4+ was not released in measurable quantities by S. marinoi in oxic conditions (data not shown). Non-labeled NH4+ that was adjusted to 100 μmol L^-1 in the growth medium prior to the experiment, was not taken up under anoxic conditions, but decreased under oxic conditions from 96 ± 14 to 43 ± 4 μmol NH4+ L^-1 during the 6 h incubation, probably due to assimilation (data not shown).

Intracellular NO3− concentrations in free-living S. marinoi cells as well as in S. marinoi cells in diatom-bacteria aggregates grown or kept under light/oxic conditions were not correlated to extracellular NO3− concentrations, with the exception that no intracellular NO3− was detected at an extracellular NO3− concentration of 0 μmol L^-1 (Figure 2). Additionally, the NO3− storage capacity of free-living and aggregate-associated S. marinoi cells was in the same concentration range (i.e., 6–62 and 18–51 mmol L^-1, respectively).

The diatom-bacteria aggregates had accumulated NO3− at concentrations that were 2–3 orders of magnitude higher than extracellular NO3− concentrations. In aggregate ‘batch 1,’ intracellular NO3− was initially stored at concentrations of 51.0 ± 6.6 mmol L^-1 based on diatom cell volume, and 4.2 ± 1.1 mmol L^-1 based on aggregate volume [mean ± SD of 12 samples taken at 0–6 h (see below); Figures 3A,B]; in aggregate ‘batch 2,’ initial intracellular NO3− was 18.0 ± 3.3 mmol L^-1 per cell, and 1.4 ± 0.3 mmol L^-1 per aggregate (mean ± SD of nine samples taken at 0–6 h; Figures 3A,B). Assuming that the extracellular NO3− concentration in the porewater of the aggregate was similar to the adjusted extracellular NO3− concentration (i.e., 75 and 100 μmol L^-1 in ‘batch 1’ and ‘batch 2,’ respectively), the standing stock of NO3− inside diatom-bacteria aggregates was largely (i.e., ≥98%) contained in diatom cells.

Within the first 6 h of incubation under dark and anoxic conditions, the intracellular NO3− concentration of the diatom-bacteria aggregates did not change significantly over time (Figures 3A,B; Supplementary Table S2). The scatter in the data may, however, have masked a potential decrease in intracellular NO3− concentration during this short time period. After 6 h, however, intracellular NO3− was rapidly consumed in aggregate ‘batch 1’ at a rate of -0.37 ± 0.03 fmol NO3− diatom-cell^-1 h^-1 or -361.78 ± 95.12 nmol NO3− aggregate^-1 h^-1. The intracellular pool was completely consumed after 48 h. In aggregate ‘batch 2,’ intracellular NO3− consumption within the first 48 h was -0.09 ± 0.01 fmol NO3− diatom-cell^-1 h^-1 or -145.00 ± 35.79 nmol NO3− aggregate^-1 h^-1, but continued at a lower rate after 48 h and NO3− was not completely used up after 72 h (Figures 3A,B; Supplementary Table S2). Thus, irrespective of the different conditions that the aggregates experienced during the pre-incubation (regarding extracellular NO3− concentration and exposure time), the aggregates displayed the same temporal pattern of intracellular NO3− consumption during incubation under dark and anoxic conditions.

Only after 6 h of incubation, significant concentration changes of intra- and extracellular NO3− as well as of extracellular NO2−, N₂, and NH4+ were observed (Figures 4A,B; Supplementary Table S3). What may look like an increase in intracellular NO3− concentration within the first 6 h of incubation is actually not statistically significant (Figure 4A; Supplementary Table S3). However, scatter in the data may have masked a potential net turnover of intra- and extracellular NO3− during this initial time period. In contrast to the exclusive production of NH4+ in the axenic cultures, the consumption of intracellular NO3− during anoxic incubation of aggregates was accompanied by the production and release of ^ICNO2−, ^ICN₂, and ^ICNH4+ from the aggregates (Figure 4A). Extracellular NO3− present in the seawater showed a similar temporal pattern of consumption and was likewise accompanied by the production and release of ^ECNO2−, ^ECN₂, and ^ECNH4+ from the aggregates, though at higher concentrations than observed for the intracellular NO3−-derived products (Figure 4B). This likely reflects the higher supply rate of extracellular NO3− diffusing into the aggregate from the surrounding water compared to the presumably slow release of intracellular NO3− from the diatom cells into the aggregate. Both ^ICNO2− and ^ECNO2− concentrations peaked after 48 h incubation and decreased thereafter. In contrast, ^ICNH4+ and ^ECNH4+ concentrations increased significantly only after 48 h. ^ICN₂ and ^ECN₂ were produced from the onset and throughout the entire incubation period. A time-integrated budget shows that in total 95% of the intracellular NO3− was consumed within the first 48 h and retrieved as ^ICNO2− (59%), ^ICN₂ (31%), and ^ICNH4+ (5%). In contrast, only 49% of the extracellular NO3− that was consumed within the first 48 h was retrieved as ^ECNO2− (35%), ^ECN₂ (14%), and ^ECNH4+ (<1%).

Notably, in the diatom-bacteria aggregates, the production of ^ICNH4+ did not coincide with the consumption of intracellular NO3−, as was the case in the experiments with axenic S. marinoi (Figure 1; Supplementary Table S1), but occurred with a delay of ca. 2 days (Figures 4A,B; Supplementary Table S3). Additionally, ^ICNO2− and ^ICN₂ were important products of intracellular NO3− consumption in the diatom-bacteria aggregates, but not in the axenic S. marinoi cultures. Thus, intracellular NO3− must have been made accessible to other microorganisms inside the diatom-bacteria aggregates contributing to complex anaerobic nitrogen cycling during the dark and anoxic incubation.

The aggregate-forming diatom S. marinoi uses intracellularly stored NO3− for the dissimilatory NO3− reduction pathway DNRA under dark and anoxic conditions and is thus the third diatom species that has been positively tested for DNRA in axenic cultures. Previously, DNRA has been found in the benthic diatom A. coffeaeformis and the pelagic, aggregate-forming T. weissflogii (Kamp et al., 2011, 2013, 2015). Thus, DNRA might be a widely distributed anaerobic metabolic pathway used by diatoms that are (temporarily) exposed to anoxic conditions in the dark, where neither photosynthesis, nor aerobic respiration is possible.

Dissimilatory nitrate reduction to ammonium activity by S. marinoi is fueled by intracellular NO3− that has been accumulated under light and oxic conditions, where it is used for assimilation (e.g., Eppley and Rogers, 1970; Clark et al., 2002; Brown et al., 2009; Bender et al., 2012; Glibert et al., 2016) or stored for later use in assimilatory or dissimilatory pathways. Nitrate uptake under anoxic conditions has, to our knowledge, not been documented and was also not observed in this study (Supplementary Figure S1). The maximum intracellular NO3− concentration found in S. marinoi was 62.0 ± 0.7 mmol L^-1 and was measured in cells that had been exposed to an extracellular NO3− concentration of 58 μmol L^-1. Intracellular and extracellular NO3− concentrations were apparently not directly correlated for S. marinoi, at least not in the targeted concentration range (Figure 2). The NO3− uptake and/or storage capacity of S. marinoi might depend on factors other than the extracellular NO₃ concentration, such as temperature (Eppley and Coatsworth, 1968; Collos et al., 1992; Lomas and Glibert, 1999a; Tantanasarit et al., 2013), the physiological state of the cells, or the extracellular NH4+ concentration (Lomas and Glibert, 1999b; Glibert et al., 2016).

Interestingly, the depletion of the intracellular NO3− pool of S. marinoi under anoxic conditions may take more than 2 days (Figure 3), whereas A. coffeaeformis and T. weissflogii completely use up their intracellular NO3− pools within less than 24 h after exposure to darkness and anoxia (Kamp et al., 2011, 2013). This was supported by the observation that S. marinoi cells that were first subjected to the “starvation procedure” and then exposed to ¹⁵NO3− had an isotopically mixed intracellular NO3− pool (i.e., ¹⁴NO3− and ¹⁵NO3−). The intracellular NO3− consumption rate immediately after exposure of S. marinoi to darkness and anoxia was 3–25 times lower for S. marinoi (0.36 fmol NO3− cell^-1 h^-1, Figure 1) than for T. weissflogii (1.15–7.47 fmol NO3− cell^-1 h^-1; Kamp et al., 2013), and A. coffeaeformis (9.1 fmol NO3− cell^-1 h^-1; Kamp et al., 2011). Likewise, the absolute amount of NO3− stored per cell was 7–35 times lower for S. marinoi (1.9–20.4 fmol NO3− cell^-1, Figure 2) than for T. weissflogii (113 fmol NO3− cell^-1; Kamp et al., 2011) and A. coffeaeformis (129 fmol NO3− cell^-1; Kamp et al., 2011). Thus, despite the roughly similar ratio between pool size and consumption rate of intracellular NO3− among the three investigated diatom species, the intracellular NO3− pool in S. marinoi can sustain DNRA for a longer period than T. weissflogii and A. coffeaeformis. The intracellular NO3− consumption rate of S. marinoi depends, however, on the initial intracellular NO3− concentration. The more NO3− that is stored, the higher the consumption rate because NO3− is largely consumed within a certain time frame (approximately 1–2 days for S. marinoi). The consumption rate might actually be overestimated, if NO3− leaks out of the cell, e.g., if the high concentration gradient between intra- and extracellular NO3− is not maintainable in darkness and anoxia. Nitrate leakage may also partly explain that only 90% of the intracellular NO3− lost during the 6-h incubation was retrieved as NH4+.

Intracellular NO3− consumption rates of foraminifera that are capable of denitrification in the absence of O₂ are ∼100 times higher (1.7–83 pmol cell^-1 h^-1; Risgaard-Petersen et al., 2006; Høgslund et al., 2008; Piña-Ochoa et al., 2010a,b; Bernhard et al., 2012) than those of diatoms capable of DNRA, while the absolute amounts of NO3− stored per cell are up to ∼10⁵ times higher (18 nmol NO3− cell^-1; Risgaard-Petersen et al., 2006). Therefore, it is not surprising that the intracellular NO3− stores of foraminifera were estimated to sustain denitrification and thus survival under anoxic conditions for much longer periods (i.e., over a month; Risgaard-Petersen et al., 2006; Glud et al., 2009). In contrast, it was hypothesized that diatoms use intracellular NO3−-fueled DNRA for entering a resting stage rather than for long-term survival with an active anaerobic metabolism (Kamp et al., 2011). Diatoms are generally known to survive dark and anoxic conditions in marine sediments for several years or even decades as resting spores (Lewis et al., 1999; McQuoid et al., 2002; Jewson et al., 2006; Härnström et al., 2011). Given the large phylogenetic diversity of diatoms of up to 100.000 species (Leblanc et al., 2012) and the pronounced genetic variation in diatoms (Armbrust et al., 2004; Bowler et al., 2008; Prihoda et al., 2012), it seems likely that diatoms have also evolved mechanisms other than DNRA to remain metabolically active during darkness and anoxia, or hitherto unrecognized pathways of anaerobic metabolism.

Diatom-bacteria aggregates produced from axenic S. marinoi and the natural bacterial community of coastal seawater contained high amounts of intracellular NO3− at the end of the aggregate-production phase in a diel light:dark cycle. It has previously been shown for S. marinoi aggregates that the total intracellular NO3− content increases with aggregate volume, whereas the aggregate-volume-specific intracellular NO3− content decreases with aggregate volume (Stief et al., 2016). The first observation is consistent with this study, with a larger number of diatom cells in larger aggregates that can store more intracellular NO3− in absolute amounts (Supplementary Figure S2). Aggregates larger than 2 μL in volume had a lower diatom cell density than the smaller aggregates, which is consistent with the previously observed lower volume-specific intracellular NO3− content of larger aggregates. This may indicate that the diatom cells in the outer shell of the aggregates mainly take up NO3− from the surrounding water because NO3− transport into the center of the aggregates is diffusion-limited. However, direct comparison of free-living and aggregate-associated S. marinoi cells did not reveal any difference in NO3− storage capacity (6–63 and 18–51 mmol L^-1, respectively, Figure 2), rendering a strong influence of diffusion limitation on NO3− transport unlikely. Additionally, the high cell densities in the aggregates and the possible competition for NO3− do not seem to lower the ability to accumulate NO3− in aggregate-associated S. marinoi cells.

When diatom-bacteria aggregates were exposed to dark and anoxic conditions, their intracellular NO3− content was used up within 2–3 days at similar cellular rates as observed in the axenic, free-living S. marinoi. Based on the cell-specific intracellular NO3− consumption rate of axenic S. marinoi (0.36 fmol NO3− cell^-1 h^-1) and the total number of S. marinoi cells in aggregates (0.5–2.1 × 10⁶ cells aggregate^-1), an intracellular NO3− consumption rate of 0.18–0.76 nmol NO3− aggregate^-1 h^-1 can be projected. This agrees reasonably well with the measured intracellular NO3− consumption rates of 0.14–0.36 nmol NO3− aggregate^-1 h^-1. In sharp contrast to axenic S. marinoi cultures, however, the consumption of intracellular NO3− by aggregate-associated S. marinoi cells was not accompanied by the concurrent release of NH4+, which argues against DNRA by aggregate-associated S. marinoi as a major nitrate sink. Instead, the ¹⁵N-labeling experiment clearly indicated that much of the NO3− initially stored by S. marinoi was used for dissimilatory NO3− reduction by the diverse microbial community of the aggregates. This interpretation is supported by the observations that (a) only a small fraction of the intracellular NO3− was converted to ^ICNH4+ (the pathway that can be carried out by axenic S. marinoi), while a much larger fraction was converted to ^ICNO2− and ^ICN₂ (that were not produced by axenic S. marinoi), (b) the aggregates produced ^ICNH4+ with a delay of 2 days, whereas axenic S. marinoi produced ^ICNH4+ immediately after the onset of dark and anoxic conditions, and (c) the time course and the products of dissimilatory NO3− reduction inside the aggregates were very similar, irrespective of whether driven by intracellular or extracellular NO3−.

The mechanism of intracellular NO3− transfer from the diatom cells to the microbial community of the aggregates is currently unknown. Living diatom cells inside the aggregates may continuously leak NO3− under anoxic conditions, especially if the aggregate porewater is NO3−-depleted and the diatoms are unable to maintain the steep concentration gradient across the plasma membrane. Under dark and anoxic conditions, diatoms are apparently unable to refill their intracellular NO3− stores (see above), which would then lead to a net loss of intracellular NO3−. However, NO3− leakage was never observed in axenic cultures of S. marinoi, nor in T. weissflogii or A. coffeaeformis, since the extracellular NO3− concentration does not increase while the intracellular NO3− is consumed under dark and anoxic conditions (Kamp et al., 2011, 2013). Alternatively, decaying or lysing diatom cells may gradually release intracellular NO3− into the aggregate if, for instance, cell lysis is triggered by viral infection (Kimura and Tomarua, 2015; Kim et al., 2015). Viral-mediated mortality and cellular lysis could be especially high in the aggregates with high cell densities (Brussaard, 2004). S. marinoi, like many other microalgae, is also known to undergo programmed cell death (PCD) when stressed (Bidle and Falkowski, 2004; Orefice et al., 2015). Anoxia inside the aggregates might act as the stressor triggering PCD in diatoms.

An alternative scenario could be that aggregate-associated S. marinoi only reduce NO3− to NO2−, and the NO2− is then immediately excreted by the cell to be further reduced to N₂ and NH4+ by the microbial community of the aggregates. S. marinoi is known to survive dark and anoxic conditions for much longer than the 6 and 72 h covered in the ¹⁵N-stable isotope experiments with axenic cultures and diatom-bacteria aggregates, respectively. Axenic S. marinoi was viable after both the 24-h starvation procedure in this study, and the 9-week incubation under dark and anoxic conditions in a previous study (Kamp et al., 2011). Thus, it cannot be ruled out that the aggregate-associated S. marinoi remained viable during the 72-h incubation under dark and anoxic conditions and displayed DNRN activity. It is currently not known, however, why S. marinoi should potentially shift from DNRA to DNRN when associated with sinking aggregates, but the observed increase in ^ICNO2− concentration indeed coincides with the decrease in intracellular NO3− concentration. Additionally, this scenario would explain both the similar rates and the gradual nature of intracellular NO3− consumption in free-living vs. aggregate-associated S. marinoi (see above). Nitrite excretion is a common phenomenon in marine phytoplankton, including diatoms (Collos, 1998; Mackey et al., 2011) and has also been reported for an axenic strain of T. weissflogii capable of DNRA (Kamp et al., 2013).

Sinking aggregates mediate much of the vertical carbon export to the seafloor and thereby represent an integral component of the “biological pump” in the ocean (Turner, 2015). Our finding that diatom-bacteria aggregates store intracellular NO3− in high amounts suggests that sinking aggregates may also be involved in the vertical transport of NO3− to deep water layers or even to the seafloor. Nitrate accumulation by diatoms only occurs in the presence of O₂ and thus diatom-bacteria aggregates may take up NO3− in the euphotic zone and oxygenated subsurface layers and export it to the deep ocean.

Depending on the initial amount of NO3− stored in sinking diatom-bacteria aggregates, the intracellular NO3− consumption rate, the internal and ambient O₂ concentration, temperature, the water depth, and the sinking velocity of the aggregates, a fraction of the intracellular NO3− stores might reach the seafloor. At the experimental conditions of this study (i.e., 15°C, 0 μmol O₂ L^-1), the half-life of the intracellular NO3− pool of the aggregates was ∼24 h. Sinking velocities of diatom-bacteria aggregates of the same size as used in this study are in the range of 50–300 m d^-1 (Iversen and Ploug, 2013). Thus, settling aggregates that still contain half of their initial intracellular NO3− content would be expected to settle on sediments at 50–300 m water depth where they may sustain benthic denitrification (Lehto et al., 2014). Aggregates exposed to lower temperatures and higher ambient O₂ concentrations in situ could transport intracellular NO3− down to considerably greater depths. More experimental work and modeling efforts will be necessary to refine these estimates. Irrespective of its unsettled quantification, the export of intracellular NO3− to deep water layers in the ocean represents a rarely considered mechanism of fixed-nitrogen loss from the euphotic zone (Lehto et al., 2014; Stief et al., 2016). Future in situ studies should also quantify the transport of intracellular NO3− by sinking aggregates in comparison to the successive depletion of their organic fraction in nitrogen relative to carbon (Martin et al., 1987; Smith et al., 1992; Dang and Lovell, 2016).

The intracellular NO3− pool in diatom-bacteria aggregates may allow diatoms to survive anoxic conditions while the aggregates sink through oxygen-depleted water layers (Kamp et al., 2011). Pelagic diatoms are well-known to survive the descent to the seafloor even at great water depth (Fileman et al., 1998) and can indeed be found in viable resting stages in marine sediments (Lewis et al., 1999; Härnström et al., 2011). As shown in this study, the diatom-derived NO3− also drives the metabolic activity of the (facultative) anaerobic bacterial community of the aggregates. The presence of bacteria actively mediating dissimilatory NO3− reduction has been repeatedly confirmed for marine snow, small particles, and sinking zooplankton carcasses (Tuomainen et al., 2003; Woebken et al., 2007; Ganesh et al., 2014, 2015; Glud et al., 2015). The internal availability of NO3− may not only allow these bacteria to remain active under anoxic conditions, but may also act as a selection factor in the succession of bacterial communities in sinking diatom-bacteria aggregates. Marine snow is mainly colonized by bacteria in the euphotic zone, but the microbial community composition is known to change during the descent (Tang et al., 2010; Thiele et al., 2015), which potentially is influenced by the presence of an internal NO3− source when O₂ is absent.

The NO3− initially stored by diatom cells is expected to drive intense anaerobic nitrogen cycling inside partially or completely anoxic diatom-bacteria aggregates sinking to the seafloor. In this study, the rates of NO2−, N₂, and NH4+ production driven by diatom-derived NO3− were ∼7 times lower than those fueled by extracellular NO3−. This indicates that inside the aggregate the supply rate of intracellular NO3− for anaerobic nitrogen cycling was lower than that of extracellular NO3−. However, the relative importance of intracellular NO3− is expected to be higher at lower extracellular NO3− concentrations and in larger, diffusion-limited aggregates (Stief et al., 2016). Notably, intracellular NO3− is an internal NO3− source that diatom-bacteria aggregates can exploit even in NO3−-depleted environments.

The finding that the diatom S. marinoi is capable of DNRA suggests that diatom-bacteria aggregates represent a pelagic NH4+ source in addition to organic matter mineralization (Kalvelage et al., 2013). The potential role of sinking aggregates and suspended particles as an important NH4+ source for the anammox process in OMZs has been discussed before (Dalsgaard et al., 2012; Kalvelage et al., 2013; Stief et al., 2016). Aggregate-associated DNRA activity may thereby fuel the anammox activity of free-living or particle-associated bacteria and thus be indirectly involved in fixed-nitrogen loss from the ocean (Stief et al., 2016). The results of this study, however, clearly indicate that at least part of the intracellular NO3− stored by aggregate-associated diatoms drives fixed-nitrogen loss directly inside the aggregates. Intracellular NO3− is transferred from the diatoms to the bacterial community of the aggregates, potentially aided by viral-induced cell lysis (Kim et al., 2015; Kimura and Tomarua, 2015), and then converted to nitrogen gas by denitrification. Toward the end of the 3-day incubation, significant ^ICNH4+ production was also observed, but since the intracellular NO3− pool was depleted by that time, this NH4+ production was most likely mediated by bacteria.

The environmental scenario emerging from this study on intracellular NO3− dynamics in diatom-bacteria aggregates can be crudely split into three consecutive phases (Figure 5):

(a) Under the light and oxic conditions in the euphotic zone, diatoms are able to accumulate NO3− intracellularly, even against a steep concentration gradient. Due to the high ambient O₂ levels, the anoxic center of aggregates will be small or even absent and hence the rates of anaerobic nitrogen cycling will be low or zero.

(b) During the descent, aggregates may pass through layers of reduced O₂ levels, which will increase the anoxic volume inside the aggregate and sustain anaerobic nitrogen cycling. Living diatoms will perform DNRA, while decaying diatoms will pass on their intracellular NO3− stores to the aggregate-associated bacterial community which performs diverse processes of anaerobic nitrogen cycling.

(c) Upon settlement of the aggregates onto the seafloor, the remaining intracellular NO3− stores (if any) may fuel benthic anaerobic nitrogen cycling, which might be further stimulated by the induction of anoxic conditions in the sediment around the aggregate.

In summary, the nitrate-concentrating capacity of aggregate-associated diatoms has the potential to impact nitrogen cycling, including fixed-nitrogen loss, not only in the photic zone, but also in the mesopelagic and benthic compartments.