Introduction

Aggregates formed by diatoms, cyanobacteria, detritus or fecal pellets are a ubiquitous phenomenon in aquatic environments and of paramount importance for the large-scale energy and nutrient transport through aquatic systems (Alldredge and Silver, 1988; Simon et al., 2002; Paerl and Kuparinen, 2003). At a small scale, phytoplankton colonies and aggregates are highly productive microenvironments characterized by elevated concentrations of organic matter and nutrients. Microorganisms attached to these nutrient oases use the substrate as a food source, for gene exchange, or as refugia from grazers (Grossart and Tang, 2010). During the last two decades, our understanding of aggregate formation, microbial colonization, degradation of organic matter and their overall role in the vertical carbon flux has improved substantially. Lagging far behind is our understanding of microbial nitrogen (N) transformations in aggregates. Hot spots of aerobic and anaerobic microbial N-cycling occur at oxic–anoxic interfaces, such as chemoclines in the water column or at sediment surfaces (Lavik et al., 2009; Stief et al., 2010; Dalsgaard et al., 2013). These zones are characterized by steep oxygen (O2) gradients resulting in redox-driven niche partitioning of microbial metabolisms (Wright et al., 2012). Such steep physical–chemical gradients also occur in aggregates, but can change abruptly on a micrometre scale and over a short time. For instance, within cyanobacterial aggregates the pH can drop from 9.0 to 7.4 within a few minutes during a light–dark shift, for example, as a cloud passes the sun (Ploug, 2008). Further, O2 can decline gradually from 100% to 0% air-saturation over a distance of less than one millimetre from the surrounding water into the centre of cyanobacterial colonies, large zooplankton fecal pellets and detritus aggregates (Alldredge and Cohen, 1987; Ploug et al., 1997; Ploug, 2008).

In artificially grown single granules of wastewater treatment reactors, the co-occurrence of nitrification and anaerobic ammonium oxidation (anammox) activity has already been demonstrated (Van Hulle et al., 2010). In naturally formed consortia, microorganisms that are active in the anaerobic part of the N-cycle as well as key genes for nitrification and denitrification have been detected for cyanobacterial aggregates (Tuomainen et al., 2003, 2006), but so far only low denitrification activity could be demonstrated in these consortia (Hietanen et al., 2002; Tuomainen et al., 2003).

Here, we examined the physical, chemical and biological constraints for aerobic and anaerobic N-transformation processes within aggregates of the cyanobacteria Nodularia spumigena Mertens ex Bornet & Flahault, which are common traits of harmful algae blooms in the Baltic Sea (Finni et al., 2001). On the basis of microsensors measurements for high-resolution profiling of dissolved O2 and nitrous oxide (N2O), 15N-isotope incubations and sensitive fluorometry analyses, we demonstrate that millimetre-sized aggregates of N. spumigena suspended in fully aerated surface waters can host aerobic and anaerobic microbial N-transformations.

Materials and methods

Sampling site and sampling

Single aggregates of N. spumigena were collected in a coastal area (58°48′45.48′′N; 17°38′47.02′′E, 0.1–0.3 nautical miles from the shore, water depth 5 m–20 m) of the Baltic Sea in August 2012. After a few days of calm, sunny weather, aggregates accumulated at the water surface because of their positive buoyancy. The structure of aggregates was compact probably because of enhanced shear forces acting on aggregate formation in coastal areas/archipelagos compared with the open sea. Therefore, aggregates could be sampled directly from the water surface with a bucket. Water disturbance was minimized to preserve the aggregate structure, and single aggregates were selected using the wider end of a glass Pasteur pipette. Water temperature was 18.5 °C and salinity 6.8. A pilot study was conducted in August 2011 at similar environmental conditions.

Characteristics of aggregates and dissolved inorganic N (DIN) in the bulk water

The diameter of each aggregate was first determined with a dissection microscope with an ocular micrometre-scale. Non-spherical aggregates were measured in three dimensions to calculate the equivalent spherical diameter. The surface area and volume was calculated as for a sphere. For analyses of dry weight, particulate organic carbon (POC) and PON, single aggregates (n=50) were collected on pre-weighed, combusted GF/F filters (25 mm, Whatman, Little Chalfont, UK) and frozen at −80 °C. GF/F filters were freeze-dried, decalcified over fuming HCl and analysed for POC and PON using an elemental analyser interfaced to a continuous flow isotope-ratio mass spectrometer (elemental analysis-isotope ratio mass spectrometry; UC Davis Stable Isotope Facility, Davis, CA, USA).

Surface water from the sampling station was taken for analysis of in situ concentrations of DIN (NO3, NO2, NH4+). Nutrients were determined spectrophotometrically on a segmented flow nutrient analyser system (ALPKEM Flow Solution IV Auto-Analyser, OI Analytical, College Station, TX, USA).

Oxygen and nitrous oxide fluxes

Individual aggregates were fixed in a temperature-controlled (18 °C), vertical flow-through chamber (Ploug and Jørgensen, 1999). An upward flow was set to 0.7 mm s−1 and all measurements were conducted in darkness. The O2 and N2O microelectrodes were attached to a micromanipulator, and gas concentrations were measured at a vertical resolution 100 μm from the ambient water towards the aggregate’s interior (Revsbech, 1989; Andersen et al., 2001). We used a Clark-type O2 microelectrode (tip diameter <10 μm, 90% response time <1 s, stirring sensitivity <1%; Unisense A/S, Aarhus, Denmark), and a Clark-type N2O microsensor (tip diameter=25 μm, response time <15 s, stirring sensitivity <2%; Unisense A/S). The O2 sensor was calibrated in O2 air-saturated and in anoxic water prepared by bubbling with N2 gas. The N2O sensor calibration was carried out in N2O-free water and in N2O-amended water which was prepared by adding a defined aliquot of N2O-saturated water to a defined volume of water. Oxygen fluxes and rates of O2-dark respiration of N. spumigena were determined mainly in August 2011. The O2 gradients were measured in two dimensions along the central plane of aggregates suspended in 100% and 30% O2 air-saturated seawater. In the ambient water, the O2 concentration was lowered by bubbling with N2. In August 2012, we conducted a combination of O2 and N2O profiling. The N2O profiling was carried out on aggregates >4 mm kept in (i) in situ seawater, (ii) in situ seawater+acetylene and (iii) in situ seawater+acetylene+NO3. Acetylene (C2H2) was added by replacing 5% of the total water volume with C2H2-saturated seawater to inhibit the reduction of N2O to N2 and thus to induce an accumulation of N2O when denitrification is present (Yoshinari and Knowles, 1976). In detail, acetylene inhibits the enzyme nitrous oxide reductase in the final step of denitrification and additionally blocks nitrification by inhibiting ammonium monooxygenase in NH4+-oxidizing bacteria and archaea. Nitrate was added to a final concentration of 30 μM to overcome diffusion-limited transport of NO3 from the ambient water into the centre of the aggregate. Fluxes of O2 and N2O, and O2-respiration rates were calculated using Fick’s 1st law (Ploug et al., 1997). A diffusion coefficient of 1.96 × 10−5 cm2 s−1 for O2 and 1.97 × 10−5 cm2 s−1 for N2O at 18 °C was used.

Ammonium release

Aggregates (n=10) were incubated individually in 15 ml Falcon tubes filled with 0.2-μm filtered seawater from the sampling site for 9 h at in situ temperature and in darkness. These aggregates were sampled in an enclosed bay close to the initial sampling station where they had been floating for 2 days after newly formed aggregates were sampled (used for O2/N2O micro-profiling and 15N-incubations). NH4+ concentrations were measured on a fluorometer (Turner Designs, TD-700, Sunnyvale, CA, USA) following the method described by Holmes et al. (1999). The NH4+ concentration at the start of incubations was 0.52±0.04 μmol l−1 (n=10). Ammonification was calculated as total NH4+ release minus NH4+ derived from N2 fixation (see below).

The distribution of NH4+ inside and around aggregates was modelled using a diffusion-reaction model accounting for the average NH4+ production rate per aggregate measured by fluorometry, aggregate size, diffusivity and boundary layer thickness known from the O2 microprofiles (Ploug et al., 1997). An effective diffusion coefficient of NH4+ of 1.60 × 10−5 cm2 s−1 was used inside the aggregates (Li and Gregory, 1974).

15N incubations: N2 fixation/NH4+ release, nitrification, NO3 reduction to ammonium, denitrification and anammox

N. spumigena aggregates, which were sampled together with the ones used for analyses of O2 and N2O fluxes, were incubated individually for 9 and 14 h in a 5.9 ml Exetainer vial (Labco, Lampeter, UK) in darkness at in situ temperature (18 °C). Each vial was filled with 0.2-μm-filtered aerated seawater from the sampling site. The Exetainer vials were rotated during the incubations assuring that aggregates were free-floating to support the diffusion of solutes into the aggregates (Ploug and Grossart, 1999).

The 15N isotope pairing technique was applied to target processes of the N-cycle (Nielsen, 1992; Thamdrup and Dalsgaard, 2002), while different combinations of labelled and non-labelled N-compounds were used (incubations I–VII, Table 1). The substrates 15NO3, 15NH4+ and 14NO3, 14NO2, 14NH4+ were added to a final concentration of 30.0±1.3 μM (mean±s.d., n=14).

Table 1 Design of incubations with 15N-labelled N-compounds to distinguish N-pathways within individual N. spumigena aggregates

One treatment (incubation II) was used to quantify N2 fixation and to track the release of NH4+ during N2 fixation. 15N2 was added as an aliquot of 15N2-amended water, which was prepared prior incubations by enriching 0.2 μm-filtered sea water from the sampling station with 15N2 gas (Mohr et al., 2010). The final 15N2 label was 1.8±0.1 atom% (n=5), as determined by gas chromatographic isotope ratio mass spectrometry. Aggregates were filtered on GF/F filters (25 mm, Whatman), and the incorporation of 15N2 gas via N2 fixation into biomass was analysed by elemental analysis-isotope ratio mass spectrometry (UC Davis Stable Isotope Facility). Rates of N2 fixation were calculated from the measured 15N atom% excess in the water and particulate N on the filter, and related to the total N2 gas in the water and particulate nitrogen per aggregates (Montoya et al., 1996). The filtrate was stored in an Exetainer vial for analysis of the released 15NH4+ via mass spectrometry.

The isotope ratios of 28N2, 29N2 and 30N2 were analysed by gas chromatographic isotope ratio mass spectrometry on a Thermo Delta V isotope ratio mass spectrometer (IGV SIL, Stockholm University, Stockholm, Sweden). Controls without label were used to determine the background isotopic composition of N2, NH4+ and NO3/NO2. The N-isotope composition of NH4+ was analysed after chemical conversion of NH4+ to N2 with alkaline hypobromite (NaOBr) (Warembourg, 1993). The N-isotope composition of NO3 and NO2 was validated after reduction of NO3 to NO2 with cadmium and conversion of NO2 to N2 using sulphamic acid (Füssel et al., 2012).

The production of N2, NH4+ and NO2/NO3 was calculated from the excess concentrations of 15N-nitrogen relative to air, and corrected for the fraction of natural abundant 15N in the total substrate pool. Rates in individual aggregates were computed from the concentrations of 15N-compounds produced versus time. Statistically significant production of 15N-labelled products was tested against controls (t-test at a confidence interval of 95% for normally distributed variables; Mann–Whitney U-test for non-normal distributed variables).

Nitrate demand

Fluxes of NO3 between the ambient water and the aggregate were modelled using the same diffusion-reaction model as for NH4+. The NO3 concentration profile and theoretical demand of NO3 in the ambient water were calculated for the average measured rates of NO3 reduction to N2O and NH4+, respectively. An effective molecular diffusion coefficient of 1.53 × 10−5 cm2 s−1 at 18 °C was used in the flux calculations (Li and Gregory, 1974).

Results

Characteristics of aggregates and DIN in the natural water

Characteristics of N. spumigena aggregates sampled in August 2012 are summarized in Table 2. Aggregates were large (3 mm) and compact (Figure 1a). Concentrations of DIN in the in situ water were 0.47±0.13 μmol NO3 l−1 (mean±s.d., n=12), 0.05±0.01 μmol NO2 l−1 (n=12) and 0.27±0.06 μmol NH4+ l−1 (n=11).

Table 2 Characteristics of single N. spumigena aggregates used for stable isotope incubations and microsensor analyses
Figure 1
figure 1

(a) Photograph of a N. spumigena aggregate. (b and c) Isopleths of O2 around and inside a N. spumigena aggregate (equivalent spherical diameter=3.1 mm) in O2 air-saturated (100%) and low-saturated water (30%) during darkness. Dark respiration (46 nmol O2 agg−1 h−1 at 100% O2, and 28 nmol O2 agg−1 h−1 at 30% O2) created sharp microscale oxyclines. The anoxic interior comprised about 5% of the total aggregate volume in 100% O2 air-saturated water and expanded to >95% in 30% O2 air-saturated water. Oxygen was measured with an O2 microsensor in two dimensions with a high spatial resolution (100 μm × 500 μm; shown as crosses). The flow velocity from below was 0.7 mm s−1 (white arrow). The aggregate surface is marked as a white circle.

Oxygen and nitrous oxide fluxes

Dark respiration ranged between 33 and 94 nmol O2 agg−1 h−1 for individual aggregates with equivalent spherical diameters between 4.9 and 6.7 mm. These respiration rates were similar to those in August 2009 and 2011, when dark respiration varied between 0.038 and 87 nmol O2 h−1 in Nodularia colonies with diameters from 0.3 to 5.0 mm (see also Ploug et al., 2011). Many aggregates of N. spumigena held interior anoxia because of their size, high respiration rates and their compact structure. The respiratory O2 consumption was thus limited by diffusion of O2 from the ambient water into the aggregates (Figure 1b). The size of the anoxic core was dependent on the O2 concentration in the ambient water, that is, the anoxic interior expanded from 5% of the total aggregate volume as an aggregate was suspended in 100% air-saturated water to >95% in 30% air-saturated water (Figures 1b and c).

No N2O production was detected when aggregates were kept in (i) seawater or (ii) seawater+C2H2. After the addition of NO3 (treatment (iii) seawater+C2H2+NO3), an increasing N2O gradient was recorded from the ambient seawater into the centre of the aggregates, showing that N2O production was limited by NO3 (Figure 2a). The N2O flux at the aggregate surface was 0.26 nmol N2O cm−2 h−1 (=0.53 nmol N cm−2 h−1). In treatment (iii) the N2O microsensor was not only used to record the N2O concentration from the ambient water towards the aggregate (see above). This treatment was also used to place the N2O microsensor in the centre of the aggregate to record the potential N2O accumulation over time within the aggregates when no advective solute transport took place. As soon as the water flow was stopped, the anoxic core expanded as the O2 diffusion into the centre decreased. Concurrently, the transport of N2O away from the aggregate decreased, and the N2O concentration within the aggregate centre increased at a rate of 1.8–3.6 μM h−1 (Figure 2b). The maximum N2O concentration in the centre of the aggregate was 0.9–2.2 μM. Please note that the latter measurements confirm the potential of N2O production within the aggregates, but the increase in N2O concentration in the interior cannot be used to calculate a N2O flux following Fick’s 1st law of diffusion because the microsensor was kept stationary within the aggregate.

Figure 2
figure 2

(a) Distribution of O2 and N2O concentration (mean±s.d., n=3) within a N. spumigena aggregate (equivalent spherical diameter=4.9 mm) during darkness. Dark respiration (126 nmol O2 cm−2 h−1) led to O2 depletion in the aggregate centre, where nitrous oxide was produced and transported outwards via diffusion. The N2O flux at the aggregates’ surface was 0.26 nmol N2O cm−2 h−1, and the ratio of O2 consumption to N2O release flux was 480:1. The flow velocity from below the aggregate was 0.7 mm s−1. (b) N2O accumulation in the centre of N. spumigena aggregates during darkness. A N2O microsensor was placed in the aggregate’s centre and N2O production was recorded over time. At time zero, the upward flow of 0.7 mm s−1 was switched off (grey arrow on the left). For aggregate #2 the flow was re-started after 70 min (grey arrow on the right) showing that advective solute transport at the aggregate–water interface was enhanced because of the water flow, which restricted N2O accumulation in the aggregate centre. (a and b) All N2O measurements were conducted in seawater enriched with C2H2-saturated seawater (5%) and NO3 (30 μM). No N2O production was detected before adding C2H2 and NO3.

Ammonium release

Two sources of NH4+ release were investigated: (a) NH4+ release derived from N2 fixation and (b) ammonification. NH4+ derived from N2 fixation was 1.51±1.10 nmol NH4+ cm−2 h−1 (n=5, 9 h incubations, see below), and the total NH4+ release was 12.9±5.3 nmol NH4+ cm−2 h−1 (n=10, 9 h incubations). Thus, about 12% of the total NH4+ release derived from N2 fixation and 88% could presumably be attributed to ammonification. The distribution of the NH4+ concentration was modelled for the boundary layer at the aggregate–water interface and the interior of colonies. Cyanobacterial aggregates were net sources of NH4+ to the ambient water, indicated by the increase in NH4+ concentration from 0.52 μM in the ambient water to 35 μM in the centre (Figure 3).

Figure 3
figure 3

Modelled distribution of NH4+ in the boundary layer and inside a N. spumigena aggregate. Owing to steep gradients of the solute from the aggregates’ interior towards the ambient water, colonies were net sources of NH4+ with high NH4+ fluxes of 12.9 nmol cm−2 h−1 off the aggregates’ surfaces.

N2 fixation/NH4+ release, nitrification, NO3 reduction to NH4+, denitrification and anammox

Pathways of N-cycling were measured by the isotope pairing technique for individual N. spumigena aggregates in darkness. We detected a net production of 15N compounds in 95% of all 15N-labelled tracer incubations. The means of all sets of 15N incubations, each with n=5–30, were significantly different from controls (t test/U test, P<0.05; see Table 3). All reported rates were net rates, that is, minimum estimates because the concentration of a 15N product was most likely diminished by co-occurring assimilation or dissimilation.

Table 3 Potential rates of nitrogen transformations for N. spumigena aggregates after 9 and 14 h incubations in darkness

Beside ammonification (see above), the highest rates were found for dark N2 fixation (including concurrent NH4+ leakage) and for nitrate reduction to ammonium, up to 1.3 and 1.7 nmol N cm−2 h−1, respectively. Of the gross N2 fixation (N2 incorporation+NH4+ release), 34.8±21.8% (n=10) were released as NH4+ to the surrounding water. Mean net rates of NO3 production via nitrification ranged between 0.004 and 0.006 nmol N cm−2 h−1, and net NO2 production via nitrification was below the detection limit. The mean N2 production rates of 0.008–0.037 nmol N cm−2 h−1 due to denitrification were lower than the rates of N2O production in the presence of C2H2 indicating an incomplete reduction of NO3 to N2 (please see discussion). Of all anaerobic microbial N conversion rates, anammox rates were consistently the lowest. In the experiments conducted in 2011, the 29N2 signal from 15NH4+/14NO2 incubations was slightly higher than in the 2012 experiments, up to 0.042 nmol N cm−2 h−1 (n =32), but it cannot be excluded that the 29N2 signal produced in these incubations at least partly originated from coupled nitrification–denitrification instead of anammox. An overview of microbial N-transformation processes and their relative importance after 9 h incubations is given in Figure 4.

Figure 4
figure 4

Schematic overview of potential nitrogen transformation processes in N. spumigena aggregates. N2 fixation and ammonification supplied sufficient amounts of NH4+. This could have been oxidized to fuel the demand of NO3 reduction, but nitrification rates limited the internal N-recycling and -loss processes. The width of arrows outlines the relative importance of net rates measured under conditions of substrate availability after 9 h incubations (compare with Table 3). N2 fixation/NH4+ release and ammonification were determined in non-enriched experiments. For nitrification, nitrate reduction to N2/NH4+ and anammox, the substrate was added to the water (30 μM). The white arrows indicate N. spumigena aggregates as net NH4+ sources and net NO3 sinks.

Potential nitrate demand

The modelled NO3 concentration profile showed that a NO3 concentration 1.5 μM in the ambient water was required to supply sufficient NO3 for the measured N2O flux (0.26 nmol N2O cm−2 h−1, Figure 2a) under C2H2 inhibition. To balance the formation of N2O and NH4+ by NO3 reduction 0.26 nmol N2O cm−2 h−1 plus 1.7 nmol NH4+ cm−2 h−1, an ambient concentration of 6.3 μmol NO3 l−1 was required. This indicates that under the experimental conditions of 30 μmol NO3 l−1, the mean NO3 reduction rate was not limited by NO3 diffusion from the ambient water into the aggregate (Figure 5). Under natural conditions, however, nitrate reduction to N2O/N2 or NH4+ was likely limited by low internal nitrification activity and low NO3 concentrations in the ambient water (<1 μM).

Figure 5
figure 5

Modelled distribution of NO3 in the boundary layer and inside a N. spumigena aggregate based on measured O2 and N2O profiles (Figure 2a). A minimum NO3 concentration in the ambient water of 1.5 and 6.3 μM is required to support the N2O flux of 0.26 nmol cm−2 h−1 (see Figure 2a) and the total NO3 reduction rate (0.26 nmol N2O cm−2 h−1 plus 1.7 nmol NH4+ cm−2 h−1, see also Table 3) at the surface of individual aggregates, respectively. During incubations of N. spumigena aggregates, the ambient NO3 concentrations were high with 30 μM to prevent NO3-diffusion limitation.

Discussion

Microbial N-transformations in N. spumigena aggregates—who is dominating and why?

Phytoplankton colonies and aggregates often have relatively short lifetimes and seldom provide a temporally and spatially stable environment. Nonetheless, this study showed that physical, chemical and biological constraints within these aggregates can act in concert to produce a dynamic, millimetre-sized consortium with co-existing aerobic and anaerobic N-transformations.

Dark N2 fixation rates and the surplus of NH4+ in the water which derived from N2 fixation were higher after 9 h incubations than after 14 h incubations. Part of the fixed N might have been transferred to bacteria, which detached from the aggregate (Kiørboe et al., 2002). Moreover, after prolonged darkness, N2 fixation in N. spumigena was likely limited by depleted intracellular energy resources. Heterotrophic N2 fixation has been reported in aphotic, deep waters of the Baltic Sea (Farnelid et al., 2009, 2013) and elsewhere (Rahav et al., 2013), but we assume that N2 fixation in N. spumigena aggregates ceases when they are exposed to prolonged darkness, and ammonification remains as the sole NH4+ source.

A large fraction of the NH4+ was lost to the surrounding water because of the steep concentration gradients (Figure 3), or served as an internal source of DIN for other members of the microbial community in the aggregate. Of those, anammox bacteria were of least importance. Compared with chemical gradients in chemoclines or in sediments where anammox can contribute significantly to N-losses (Lam and Kuypers, 2011), the physicochemical gradients in aggregates are far more unstable (Ploug et al., 1997). Thus, the growth of anammox bacteria with doubling times of several days might be restricted in aggregates (Strous et al., 1999). Nitrifying bacteria and archaea can have shorter doubling times of 1–2 days (Ward et al., 2007), and the presence of nitrifiers (likely Nitrospira) as well as key genes for nitrification and denitrification (amoA, nirS, nirK) have been demonstrated in N. spumigena aggregates (Tuomainen et al., 2003, 2006). In our study, net nitrification was low despite high concentrations of internal NH4+ and ambient O2, but the actual nitrifying activity was masked by the removal of 15NO3 because of high rates of NO3 reduction. After incubations with 15 NH4+/14NO2, 70±17% (n=20) of the 15N-N2 was detected as 15N15N, which indicates coupled nitrification–denitrification. Also, incubations with 15NO3 to target denitrification showed a significant formation of 29N2 in excess to the low in situ concentrations of 14NO3. Accordingly, 5.9±2.9% (n=19) of the NO3 reduction originated from 14NO2/14NO3 that was not added as a tracer but was oxidized by nitrifiers using the NH4+ released from the aggregates. The corresponding gross nitrification rates were as high as 0.09±0.07 nmol NH4+ cm−2 h−1 (n=39)—ca. 20-fold higher than the net rates given in Table 3.

Aggregates of N. spumigena responded to NO3 exposure by reduction to NH4+, which retained inorganic N in the vicinity of the consortia instead of losing it as N2 by denitrification. It is, however, difficult to separate dissimilatory and assimilatory pathways of 15N-NO3 reduction based on our measurements. Strains of N. spumigena isolated from the Baltic Sea have a low NO3 affinity (Vintila and El-Shehawy, 2010; Kabir and El-Shehawy, 2012), but in our incubations, NO3 concentrations were high (30 μM), and thus NO3 assimilation by N. spumigena cannot be excluded. Moreover, NO3 reduction can be explained by the activity of a diverse group of bacteria, protists and eukaryotic algae which are associated with N. spumigena aggregates (Salomon et al., 2003; Stoecker et al., 2005; Tuomainen et al., 2006). Therefore, the respective share of dissimilatory and assimilatory pathways in NO3 reduction in N. spumigena aggregates remains uncertain.

Net N2 production via denitrification was about two orders of magnitude lower than NO3 reduction to NH4+. The N2 production rates detected in the 15N-NO3 incubations, however, were more than 10 times lower than the rates of N2O production in the presence of C2H2 indicating that, even without the inhibitor, a large amount of NO3 was not reduced to N2. Possible explanations could be that denitrifiers produced N2O instead of N2 under aerobic conditions (Takaya et al., 2003), or that a low pH of 7.4 inside N. spumigena aggregates during darkness (Ploug, 2008) led to an incomplete denitrification (Thomsen et al., 1994). Nevertheless, most likely the usage of two different experimental set-ups to measure N2 and N2O production—one using 15N-isotope incubations lasting for up to 14 h and the other applying microsensors to record the instantaneous N2O production inside aggregates after NO3 exposure in the presence of C2H2—yielded different results. Therefore, solid conclusions on the N2O:N2 ratios should await further studies.

In our incubations, the aggregate-attached denitrifying bacteria faced neither strict O2 nor organic C limitation, but NO3 limitation. This is supported by the micromolar-concentrations of N2O that accumulated inside the anoxic interior (Figure 2) less than 1 h after NO3 was added to the N-depleted water—the duration which corresponds to the diffusion time of NO3 into the aggregates. Most denitrifying bacteria are facultative anaerobes (Zumft, 1997), that is, attached denitrifying bacteria may switch from O2 respiration to NO3 respiration at persistent low O2 and elevated NO3 concentrations. This assumption is supported by the observation that 61% of the total 15N-NO3 label (30 μM) was recovered as N2 and 39% as NH4+ after one aggregate was left in anoxic water for several weeks. In this case, denitrification activity even exceeded rates of NO3 reduction to NH4+.

Microbial N-transformations in aggregates suspended in surface waters or in O2-depleted, NO3-enriched sub-surface waters

Oxygen depletion inside aggregates is a prerequisite for anaerobic N-transformation. In Figure 6, we compiled O2-respiration rates measured on aggregates of different sizes and sources from previous studies. For reference, we plotted the modelled respiration rate needed for the aggregates’ interiors to turn anoxic: In O2 air-saturated water (O2=250 μM), anoxia occurred within cyanobacterial colonies 1 mm or (diatom) macroaggregates of upwelling zones. For smaller aggregates and zooplankton fecal pellets, the measured respiration rates were 1–2 order of magnitudes lower than the respiration rate required for these to turn anoxic (Figure 6). Accordingly, anaerobic N-recycling and -loss processes present in fully aerated waters seemed to be restricted to large, compact aggregates or colonies formed by, for example, N. spumigena or Trichodesmium spp. The N2-fixing cyanobacteria Trichodesmium spp. accounts for a gross of the global marine N2 fixation (Capone et al., 1997; Karl et al., 2002); and similarities between N. spumigena and Trichodesmium spp. in terms of aggregate formation, the development of microscale oxyclines (Paerl and Bebout, 1988), and NH4+ remineralisation (Mulholland and Capone, 2000) suggest that a redox-driven niche partitioning of microbial N-transformation processes may not only exist in N. spumigena but also in Trichodesmium spp. Yet, diazotrophic cyanobacterial colonies accumulate in N-depleted surface waters during bloom periods. We therefore presume that their effect as N-sink is negligible in the oxygenated, NO3-depleted photic zone.

Figure 6
figure 6

Respiration rates of aggregates of various sources and sizes and their potential of central anoxia. High rates of aerobic respiration can lead to O2-depletion in the aggregate centre as the diffusion of O2 from the ambient water inwards the aggregate is insufficient to compensate for the internal O2 consumption. The reference respiration rate which predicts the aggregates to turn anoxic in their centre when floating in fully oxygenated water (250 μmol O2 l−1) or in low-oxygen waters (25 μmol O2 l−1) are shown as a straight and dashed line, respectively. At ambient O2 concentrations of 250 μM, anoxic conditions may occur within cyanobacterial colonies 1 mm or (diatom) macroaggregates. At 25 μmol O2 l−1, even smaller phytoplankton aggregates, detritus as well as fecal pellets (<1 mm to 0.1 mm) can become anoxic. The reference respiration rate needed for aggregates or colonies to turn anoxic in their centre at a bulk O2 concentration of 25 or 250 μM was calculated according to Ploug et al. (1997) assuming an apparent diffusivity inside aggregates of 0.9 times the molecular diffusion coefficient in water (Ploug et al., 2008). Data of respiration rates are compiled from previous studies of aPloug et al. (2008), bPloug et al. (2011), cPloug et al. (2010), dIversen and Ploug (2010), ePloug et al. (2002), fGrossart et al. (2003) and gPloug et al. (1999).

As senescing aggregates sink out of the surface layer, they can enter water depths that are NO3-enriched, for example, at the Landsort Deep in the Baltic Sea at 30 m depths (1.5 μM NO3, Swedish Environmental Monitoring Program) and in global waters at depths of >200 m (>10 μM NO3) or non-sulphidic O2 minimum zones (25 μM NO3) (Karl et al., 2003; Kuypers et al., 2005; Johnson et al., 2010). In these zones, NO3 concentrations are sufficient to support NO3 reduction rates as found in our study (Figure 5). Moreover, these water layers are often accompanied by hypoxia or anoxia. At low ambient O2, the anoxic core, and thus the site for anaerobic respiration within cyanobacterial aggregates, expanded substantially (Figure 1). Additionally, not only large but also smaller phytoplankton aggregates, detritus and fecal pellets (<1 mm to 0.1 mm) were predicted to become anoxic (Figure 6). We therefore suggest that anaerobic N-transformations in aphotic O2-depleted waters are promoted by particulate material (Lam and Kuypers, 2011; Ulloa et al., 2012; Ganesh et al., 2014). Nonetheless, our measurements with aggregates at 18 °C are not directly applicable to the mesopelagic zone where microbial processes in senescent aggregates decrease because of lower temperatures (Iversen and Ploug, 2013). Moreover, the potential of aggregates to significantly contribute to the N-cycling in O2-deficient, NO3-enriched waters might be limited by their short residence times within these zones and the exponential decline of particulate organic material within the first hundred meters water depth (Martin et al., 1987). Hence, in future studies, the particle size distribution within these zones should be investigated in combination with small-scale fluxes of N within settling aggregates allowing for the temporal variation of the microbial community, and the effect of low ambient O2 availability and low temperature on the activity of aggregate-attached microbes.