Baltic Sea cyanobacterial bloom contains denitrification and nitrification genes, but has negligible denitrification activity

Authors

  • Jaana M Tuomainen,

    Corresponding author
    1. Finnish Institute of Marine Research, P.O. Box 33, FIN-00931 Helsinki, Finland
    2. North Savo Regional Environment Centre, P.O. Box 1049, FIN-70101 Kuopio, Finland
    3. Department of Environmental Sciences, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland
      *Corresponding author. Tel.: +358 (71) 788 4904; Fax: +358 (71) 281 2461. E-mail address: jaana.tuomainen@ymparisto.fi
    Search for more papers by this author
  • Susanna Hietanen,

    1. Finnish Institute of Marine Research, P.O. Box 33, FIN-00931 Helsinki, Finland
    Search for more papers by this author
    • 1

      Department of Ecology and Systematics, Division of Hydrobiology, University of Helsinki, P.O. Box 65, FIN-00014 University of Helsinki, Finland.

  • Jorma Kuparinen,

    1. Finnish Institute of Marine Research, P.O. Box 33, FIN-00931 Helsinki, Finland
    Search for more papers by this author
    • 1

      Department of Ecology and Systematics, Division of Hydrobiology, University of Helsinki, P.O. Box 65, FIN-00014 University of Helsinki, Finland.

  • Pertti J Martikainen,

    1. Department of Environmental Sciences, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland
    Search for more papers by this author
  • Kristina Servomaa

    1. North Savo Regional Environment Centre, P.O. Box 1049, FIN-70101 Kuopio, Finland
    2. Department of Environmental Sciences, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland
    Search for more papers by this author

*Corresponding author. Tel.: +358 (71) 788 4904; Fax: +358 (71) 281 2461. E-mail address: jaana.tuomainen@ymparisto.fi

Abstract

A cyanobacterial bloom in the Gulf of Finland, Baltic Sea, was sampled throughout the development and senescence of aggregates in August 1999. While conditions inside the aggregates were favourable for denitrification (rich in nitrogen and carbon, with anoxic microzones), essentially none was detected by a sensitive isotope pairing method. Polymerase chain reaction-based methods, targeting functional genes encoding the key enzymes of denitrification and nitrification processes (nirS, nirK, amoA), revealed that the non-aggregated filaments harboured amoA gene fragments with high similarity to Nitrosospira amoA sequences, as well as both types of nitrite reductase genes, nirS and nirK. Only the nirS-type nitrite reductase gene and no amoA was detected in aggregated filaments. Thus, despite optimal environmental conditions and genetic potential for denitrification, the blooms of filamentous nitrogen-fixing cyanobacteria must be seen solely as a source, and not as a sink of nitrogen in the Baltic Sea.

1Introduction

In large regions of the oceans, primary production is nitrogen-limited and so it is in most areas of the brackish Baltic Sea [1,2]. Nitrogen fixation, nitrification and denitrification, which are all microbially mediated, are the key processes determining the availability of nitrogen. A major part of nitrogen fixation in the Baltic Sea occurs during periodic cyanobacterial blooms during the warm summer months. Two dominating filamentous species, Aphanizomenon sp. and Nodularia sp., produce dense accumulations in the surface waters spreading over hundreds of square kilometres. Nitrogen fixation by the cyanobacteria in the Baltic contributes up to ∼35%[3] of the annual nitrogen loading to the Baltic Sea. However, few attempts have been made thus far to follow and quantify the flux of fixed nitrogen. Gas-vacuolated cyanobacterial cells do not sink to the bottom but decompose within the surface layer, thus releasing nitrogen [4,5]. Grazing on the cyanobacteria in the Baltic Sea is of minor importance (e.g. [6]), and therefore, direct transfer of fixed nitrogen to higher trophic levels remains inefficient. Previous studies have indicated that cyanobacterial blooms support an active microbial community throughout the whole lifetime of the blooms, and especially the aggregation of the filaments promotes dense colonisation by autotrophic and heterotrophic biota [7–10]. Aggregates of various types are the sites of elevated nutrient concentrations and they provide anaerobic microniches within the pelagic environment [11–14]. Elevated nitrogen levels are likely to occur in cyanobacterial aggregates, as nitrogen-fixing cyanobacteria release recently fixed nitrogen into the surrounding water as dissolved organic nitrogen [5,15,16]. High heterotrophic activity consumes oxygen within the aggregate, and as the diffusion of oxygen into the aggregate may be hindered by algal mucus, anoxic conditions may be reached, which facilitate anaerobic processes like denitrification and sulfate reduction.

Denitrification, that is reduction of nitrate to molecular nitrogen, is an important mechanism of nitrogen removal in marine environments. It is significant in the sediments of the Baltic Sea [17]. However, very little is known about denitrification in the water column of the Baltic Sea. Denitrification has been found at the interface between anoxic stagnant deep water and overlying oxic water in the central Baltic proper [18,19]. In those studies, denitrification was measured by the acetylene blockage method, now known to some serious flaws, so the conclusions about the importance of the process are not clear. Denitrification in cyanobacterial aggregates in situ has not been measured earlier, because of the lack of appropriate methods. The 15N isotope pairing method [20] has proved to be an excellent tool for measuring denitrification in sediments. It is based on adding 15N-labelled nitrate to samples and, after incubation, measuring the relative increase of the partly (14N15N) and completely (15N15N) labelled N2 molecules. From the changes in isotopic composition it is possible to calculate the denitrification rate based on the added nitrate, and also denitrification occurring in the environment, that is based on nitrate that is readily available or produced in situ by nitrification. We used the 15N isotope pairing method to find out whether denitrification occurs in cyanobacterial aggregates. Cyanobacterial bloom samples were collected at the entrance to the Gulf of Finland in August 1999 and denitrification was measured without disturbing in situ oxygen conditions within the aggregates. As denitrification sometimes does not proceed to the end (N2), but ceases after the reduction of nitrite (NO2) to nitrous oxide (N2O) [21,22], we also measured production of N2O. In addition, the number of attached bacteria in the aggregates was counted to verify high colonisation levels.

Denitrification can be performed by a variety of Bacteria, Archaea [23], and even certain fungi [24]. The process consists of four enzymatic steps. Nitrite reductase, which catalyses the reduction of nitrite to nitric oxide (NO), is the key enzyme of this process, as it produces the first gaseous intermediate of the denitrifying pathway. Nitrite reductase has two functionally and physiologically equivalent, but structurally different, forms, encoded by different genes, nirS and nirK. These gene types occur exclusively in a particular bacterial strain (see [25] for an exception), but different strains within one species may have either type of the gene. The nirK gene has been fully or partly characterised from numerous denitrifiers (e.g. [26–28]), and also from ammonia-oxidising bacteria [29]. Partial or full-length sequence data for the nirS gene are similarly available from several denitrifying species (e.g. [25,27,30]).

Nitrification is a sequential oxidation of ammonia (NH4+) to nitrate (NO3) via nitrite (a process that produces substrate for denitrification) and is carried out by autotrophic ammonia- and nitrite-oxidising bacteria. Ammonia oxidation is the rate-limiting step, catalysed by the enzyme ammonia monooxygenase (AMO), which accordingly is the key enzyme of the nitrification process. The gene encoding the subunit carrying the active site of the AMO enzyme, amoA, has been cloned and sequenced from several bacteria (e.g. [31–33]). Most of the ammonia-oxidising species belong to the β-subdivision of the Proteobacteria [34,35]. Only Nitrosococcus oceani belongs to the γ-proteobacterial subdivision, being more related to methane- than to ammonia-oxidising bacteria [35].

To find out whether bacteria associated with a cyanobacterial bloom in the Baltic Sea have genetic potential for nitrogen transformation processes, a polymerase chain reaction (PCR)-based approach targeted to the functional genes of the key enzymes for denitrification (nirS and nirK) and nitrification (amoA) was used in the study of the cyanobacterial samples from the Gulf of Finland in August 1999.

2Materials and methods

2.1Study area and sampling

Cyanobacterial (Nodularia sp.) bloom samples were collected on a cruise of R/V Aranda in August 1999. Spherical aggregates were collected at a fixed sampling station at the entrance to the Gulf of Finland (Baltic Sea) (59°28′N, 22°53′E) on August 4, 5, 8 and 10 using a zooplankton net (500 μm mesh) towed from the thermocline (20 m) up to surface twice a day (morning and afternoon). The aggregates were gently picked up using a wide-bore pipette. Samples of not yet aggregated suspension of cyanobacterial filaments, floating on the surface, were collected on August 2, 3, 4 and 8 by towing a phytoplankton net (50 μm mesh) through the dense cyanobacterial surface bloom from the deck of the ship or from a slowly moving motorboat around the sampling station. Molecular biology analyses were performed on aggregate samples from August 2, 3 and 4 and on suspension samples from all four sampling days.

2.2Dry weight, particulate organic carbon (POC) and nitrogen (PON)

Single aggregates (10–20 replicates on each sampling occasion) were collected on combusted GF/F filters (diameter 11 mm, Whatman) for POC and PON measurements. The filters were air-dried and stored in acid-washed microcentrifuge tubes until analysis within 2 months. For dry weight estimates the filters with aggregates were weighed individually before combustion in a LECO CHN-900 analyser (LECO, St. Joseph, MI, USA). The average weight of an unused filter was subtracted from the sample weight.

2.3Oxygen production and respiration

Oxygen production and respiration in the aggregates were measured using a micro-Winkler titration method. Aggregates (one to three) were transferred to numbered, volume-calibrated (circa 22 ml) glass-stoppered glass vials filled with 10 μm filtered surface water. Three series of five replicate samples were used for measuring ambient oxygen level at the beginning of incubations, oxygen production and respiration. Samples were incubated either in the dark (respiration) or in the light (oxygen production) at in situ temperature for 3–6 h. Samples were slowly rotated on a test tube mixer (Spiramix, Denley, Hampshire, UK) in order to keep the aggregates suspended during the incubation [36]. Oxygen production and respiration were calculated from changes in oxygen concentrations in the vials. No ambient water subtractions were made, as respiration in seawater is negligible over such short incubation times [37]. The measured respiration rates were compared to the rates needed to create anoxia at the centre of a spherical aggregate of a similar volume, calculated according to Ploug et al. [14] using the oxygen diffusion coefficient, the aggregate diameter and the temperature, salinity and oxygen concentration at the sample station. The aggregates were assumed to have a full diffusion boundary layer, as the cyanobacteria are positively buoyant due to intracellular gas vacuoles (no sinking-induced erosion of the diffusion boundary layer) and often rich in mucus, further slowing down the diffusion into and out of the aggregates.

2.4Denitrification

Glass bottles (32 ml) equipped with diagonally cut conical glass stoppers were used, enabling incubations in gas-tight, bubble-free conditions. Aggregates (five or 10) were transferred to vials (four replicates per sampling occasion) filled with 10 μm filtered surface water. On every sampling occasion, a sample of 10 μm filtered surface water without aggregates was prepared to check for ambient water activity. In addition, a sample without 15N addition was preserved for background subtraction. K[15N]O3 solution (99 atom%, Europa Scientific, Cheshire, UK) was added to the vials (100 μM final incubation concentration) that were immediately closed and covered with aluminium foil. During incubation the vials were rolled slowly as described above. The incubation time (17–27 h) was chosen on each occasion separately, based on the results from respiration measurements of parallel samples. The more ‘active’ samples were incubated for a shorter time than the less active ones, in order to avoid lowering the ambient oxygen concentrations in the sample vials. The activity in the samples was terminated by carefully adding 1 ml of ZnCl2 solution (500 μg ml−1) under the surface, avoiding gas release from the samples. The samples were then transferred to 10-ml gas-tight Exetainers (Labco, High Wycombe, UK), that contained 500 μl of ZnCl2 solution. The mass ratios of N2 formed in aggregate samples and in the filtered (10 μm) water (for background subtraction) were later analysed using a mass spectrometer by the National Environmental Research Institute in Silkeborg, Denmark. Denitrification rates were calculated from mass ratios according to Nielsen [20].

2.5N2O production

N2O production was measured from the aggregates collected in the morning of August 10. Single aggregates were transferred to ampoules (60 ml) that had been filled with filtered (10 μm) surface water. The ampoules were closed without leaving any air inside with non-toxic rubber septa and aluminium caps. Blank samples were treated immediately with 600 μl of formalin. Samples were incubated in the dark at in situ temperature for 9 h. During incubation the ampoules were rolled slowly on a test tube mixer as explained above. After terminating the incubation by injecting 600 μl formalin through the septum equipped with an overflow needle the ampoules were stored upside down at room temperature until analysis within a month. Ten ml of sample water was replaced by 10 ml of ultra-pure helium, and samples were shaken vigorously and then left to stabilise for 1 h. Subsamples (500 μl) of the headspace were analysed in a Hewlett Packard 5890 Series II gas chromatograph (Hewlett Packard, Waldbrunn, Germany) equipped with an HP 3396 A integrator, electron capture detector and a Porapak Q column at +35°C, using argon–methane (95%–5%) as a carrier gas. The amount of N2O in the sample was calculated using the gas equation (PV=nRT, where P is pressure, V is volume, n is amount of gas moles, R is ideal gas constant, and T is temperature) and the Ostwald gas absorption coefficient [38].

2.6Bacterial cell counts

On August 10, morning and afternoon, aggregates were collected to enumerate the attached heterotrophic bacteria. Single aggregates were transferred to combusted glass vials containing 10 ml of filtered (10 μm) surface water and 500 μl of filtered (0.2 μm) formalin. Samples of filtered (10 μm) surface water without aggregates were used for background subtraction in the analysis. Vials were sealed and stored at +6°C until analysis within 2 months. Aggregate samples were treated with pyrophosphate (final concentration 10 nM) and sonicated in an ultrasonic bath (Bransonic 3200, Branson Ultrasonics Corporation, Danbury, CT, USA) for 20 min in order to detach bacteria from algal filaments [39]. Samples were then diluted (300-fold) with particle-free water and 100-μl aliquots of diluted samples were stained with DAPI at a final concentration of 5 μg ml−1 and counted using epifluorescence microscopy (1250× magnification, Leitz Aristoplan, Wetzlar, Germany) [40,41]. Bacteria in the background samples (10 μm filtered surface water) were stained with DAPI at a final concentration of 0.001 μg ml−1[40]. The average cell number of the background water was subtracted from cell numbers of aggregate samples to get the aggregate-attached cell numbers.

2.7DNA extraction

Seawater was removed from the aggregates by filtering through a polycarbonate membrane (Osmonics Poretics, 0.2 μm pore size, diameter 25 mm, with a vacuum <10 kPa). Small, non-sticky aggregates were transferred from the membrane to a microcentrifuge tube carefully with a spatula, whereas bigger, sticky ones were transferred to a tube on the membrane. To remove seawater from the samples (15 ml) of the net-collected cyanobacterial suspension, the samples were centrifuged in conical tubes (15 min, 3900×g, +4°C), and the upper pellets from the surface, consisting mainly of coiled filaments of cyanobacteria, were transferred to 1.5-ml microcentrifuge tubes. The pellets formed at the bottom of the 15-ml tube contained mainly cell debris and zooplankton (Maija Huttunen, Finnish Institute of Marine Research, microscopic confirmation) and were excluded from the analyses. The upper pellets, transferred to the microcentrifuge tube, were centrifuged further (20 min, 13 500×g, +4°C), and excess seawater was removed carefully with a pipette. All samples were frozen at −80°C.

DNA was extracted from frozen samples with a slightly modified method from a manual edited by Ausubel et al. [42]. One ml of sterile extraction buffer warmed to +37°C (100 mM Tris–HCl, pH 8.0; 100 mM EDTA, pH 8.0; 100 mM sodium phosphate, pH 8.0) was pipetted on the frozen samples (0.5–1 g), and vortexed to thoroughly suspend the samples, then sterile sea sand was added and the samples were shaken for 15 min at room temperature. Extraction buffer was added to a final volume of 4.7 ml, and lysozyme to 1 mg ml−1 final concentration. The samples were incubated at +37°C for 30 min. Sterile sodium dodecyl sulfate was added to 0.5%, and proteinase K to 150 μg ml−1, and the incubation was continued for a further 30 min at +37°C. The samples were boiled for 1 min, and left to cool to room temperature. Sterile NaCl was added to a final concentration of 0.7 M, gently mixed, and 830 μl sterile precipitation buffer (10% cetyltrimethylammonium bromide, w/v; 0.7 M NaCl) was added to the tubes, mixed gently again, and incubated at +65°C for 10 min. The samples were extracted once with phenol–chloroform–isoamyl alcohol (25:24:1), and once with chloroform–isoamyl alcohol (24:1), and centrifuged in between (5 min, 10 000×g, +20°C) to separate the aqueous and organic phases. Chloroform–isoamyl alcohol extraction was repeated if the aqueous phase was not clear. DNA was precipitated from the aqueous phase with two volumes of 99.5% ethanol for at least 2 h at −20°C, and pelleted by centrifugation (10 min, 10 000×g, +4°C). The DNA pellets were washed twice with cold 70% ethanol, air-dried, and redissolved in TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 8.0). The DNA was extracted from aggregate samples with the same method, with an extra purification step using the Wizard DNA Clean-Up System (Promega, Madison, WI, USA).

2.8PCR amplifications and cloning of the products

Degenerative primers A189 and A682 [43] were used for amplification of the amoA gene. The reaction mixture consisted of 20 pmol of each primer, 0.2 mM of each dNTP, Mg-free DyNAzyme buffer, 2.5 mM MgCl2, 100 ng μl−1 bovine serum albumin, and 1.2 U DyNAzyme II polymerase (Finnzymes, Espoo, Finland). When there was no product after the first amplification reaction, a second amplification was performed. The reaction mixture of the first PCR amplification was purified with the Wizard PCR Preps DNA purification system (Promega), and used as a template for the second amplification. The reaction mixtures were similar for both amplifications except that in the first amplification the amount of total DNA template was 50–500 ng, and in the second reaction 2 μl of the Wizard-purified reaction mix of the first amplification was used as template. Forward and reverse primers used for the two types of nitrite reductase were nirS1F and nirS6R for nirS, and nirK1F and nirK5R for nirK[27]. The reaction mixture consisted of 20 pmol of each primer, 0.2 mM of each dNTP, Mg-free DyNAzyme buffer, 100 ng μl−1 bovine serum albumin, MasterAmp™ PCR Enhancer (Epicentre, Madison, WI, USA), 1.2 U DyNAzyme II polymerase (Finnzymes), with 2 mM MgCl2 for nirS, and 2.5 mM MgCl2 for nirK. When there was no product after the first amplification reaction, a second amplification was performed as described above. All amplification reactions (3 min denaturation at +95°C, 35 cycles of 1 min denaturation at +95°C, 1 min annealing at +58°C, 1 min extension at +72°C, and 30 min final extension at +72°C) were performed using a PTC-200 thermal cycler (MJ Research, Waltham, MA, USA). PCR products of appropriate size were purified from SeaPlaque low-melt agarose gel (BMA, Rockland, ME, USA) with the Nucleospin Extract kit (Macherey-Nagel, Düren, Germany), and cloned for sequencing with the TOPO TA cloning kit (Invitrogen, Carlsbad, CA, USA). Cloned PCR products were grown in TOP10 Escherichia coli (Invitrogen) and purified with the QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany).

2.9Restriction fragment length polymorphism (RFLP) analysis

In spite of stringent PCR conditions, a large proportion of the first sequenced nirK and amoA PCR products proved to be unspecific, i.e. containing genes other than the target ones. To achieve preliminary typing of the PCR products, RFLP screening methods were established according to these sequences and the sequences of these genes in the EMBL database. DNA from clone purifications containing nirS plasmid was digested in a double-digestion reaction with Cfr42I and Eco130I in Y+/Tango buffer, and the nirK clones with Bsp143II in Y+/Tango buffer. The amoA clones were digested separately with Bsp143II (buffer Y+/Tango) and MvaI (buffer R+). Additionally, the majority of the unspecific amoA gene fragments could be identified with the restriction enzyme Bsp68I (in O+ buffer), which had no cleavage site in the specific fragments. All restriction enzymes and buffers were from MBI Fermentas (Vilnius, Lithuania). The plasmids were digested at +37°C for at least 2 h and the fragments were separated in agarose gel (2% SeaKem LE agarose, BMA) containing ethidium bromide for visualising the DNA in UV light. Gel images were stored with the BioDocII Video Documentation System (Biometra, Göttingen, Germany) and the images were analysed and compared with BioNumerics version 2.50 (Applied Maths BVBA, St-Martens-Latem, Belgium).

2.10Sequencing and phylogenetic analysis

Cloned fragments were sequenced using the ALF DNA sequencer (Amersham Pharmacia Biotech, Uppsala, Sweden), Thermo Sequenase fluorescent-labelled primer cycle sequencing kit, 7-deaza-dGTP (Amersham Pharmacia Biotech, Piscataway, NJ, USA) [44,45] and T3 and T7 sequencing primers. An internal sequencing primer (agtacgg[c/t]cc[c/g]gt[a/c/g]tgg) was designed for nirS products (c. 900 bp), based on the already sequenced nirS clones and the sequence data available in the EMBL, to enable reliable sequencing through the region used in phylogenetic analyses in one reaction. However, most of the aggregate-derived nirS fragments had to be sequenced in full length, as their sequences varied markedly in the internal sequencing primer area. All nirS, nirK, amoA and pmoA fragments chosen for the phylogenetic analyses, also deposited in the EMBL database, were sequenced at least three times to ensure the reliability of the sequences.

Sequences were compared with those in the EMBL database with Fasta3 (EMBL Outstation, European Bioinformatics Institute, ©8 EBI 1998 [http://www.ebi.ac.uk/fasta33/]). Pairwise sequence comparisons were performed with Gap, and sequence alignments with PileUp in the Wisconsin Package (version 10.2, Genetics Computer Group, Madison, WI, USA). Phylogenetic analyses were performed with a Phylip software package (version 3.5c [46]) using the neighbour-joining method. Bootstrap values were determined with Seqboot, using 100 replicates. Trees were drawn with Treeview (version 1.6.6; ©Roderick D.M. Page, 2000 [http://taxonomy.zoology.gla.ac.uk/rod/rod.html]).

2.11EMBL accession numbers

The amoA sequences derived from the Baltic Sea have been deposited in the EMBL database under accession numbers AJ419602, AJ419606, AJ419614, AJ419615, AJ419618, pmoA sequence under AJ419613, nirS sequences under AJ419603–05, AJ419607–12, AJ419616, AJ457194–205, AJ488907–09, and nirK sequences under AJ419617, AJ419619–23 and AJ488588.

Accession numbers of the EMBL reference sequences used in the phylogenetic analyses for nirS were AF114789–92, AJ224910–12, AJ248393, AJ248395, AJ248402, AJ248403, AJ248405, AJ248422, AJ401462, AJ440470, AJ440476–78, AJ440486, AJ440488, AJ440489, AY078256, AY078261, AY078265, AY078266, AY078268–70, AY078272, AY078273, AY078275–77, AY121580, AY121592, AY121602–04, M80653, U05002, U75413, X16452 and X56813, for nirK X91394, AB013078, AB046603, AE007256, AF051831, AF114786–88, AF339044, AF339049, AJ002516, AJ224902–09, AY072263, AY072264, AY072271, AY072272, AY072274, AY072275, AY078247, AY078249, AY078254, AY078255, AY121526, AY121543, AY121544, AY121553, AY121557, AY121559, D13155, M97294, U62291, U65658, Z21945 and Z48635, and for amoA/pmoA AB064371, AF016982, AF037108, AF042170, AF043707, AF070984, AF070986, AF148522, AF150758, AF150791, AF150795, AF150803, AF153344, AF182494, AF264126, AF264134, AF283227, AF339038, AF339039, AF339042, AF339043, AF358040, AF358055, AF368359, AF368368, AF390133–35, AJ299946, AJ299966, AJ299967, AJ489804, AY007285, AY080945, L40804, U31650, U31652–55, U38250, U51630, U76552, U76553, U81596, U89301–04, U91603, U96611, X90821 and X90822.

3Results

In the beginning of August 1999, the thermocline at the study area was situated at 20 m depth. The temperature of the upper mixed layer was around 19°C and salinity around 6 psu. A clear oxygen minimum was observed just above the thermocline due to respiration of accumulated heterotrophic plankton on top of the thermocline. The chlorophyll a concentration in the mixed layer increased during the study period from 5 to 10 μg l−1. Nutrient profiles showed values below the detection limit in the mixed layer, with increasing concentration below the thermocline.

First aggregates were collected when the cyanobacteria, mainly Nodularia spumigena (Mertens), were growing fast and filaments were strongly coiled into spherical, dark green aggregates (Fig. 1) on August 4 and 5. The aggregates were small and consisted mainly of cyanobacteria. Within days, the cyanobacterial filaments became lighter in colour and aggregates became yellowish and more buoyant (samples on August 8). The change in aggregate appearance was mainly due to decaying N. spumigena filaments, but also due to diatoms (Nitzchia sp.), microheterotrophs and faecal pellets that were gathering into the bundles. At this stage, the aggregates were also densely colonised by bacteria, but still packed into tight, round bundles. Later (samples on August 10), the bundles became larger but also looser, and thus, less dense, with the N. spumigena bloom showing signs of decay with a large proportion of empty filaments in the aggregates. A high number of diatoms (Nitzchia sp.) was recorded within the aggregates as free cells and within gel-like bags. The appearance of aggregates varied from tight spheres of about 3 mm in diameter on the first sampling day, via even smaller, 2-mm packages, to loose, roundish bundles of about 4 mm in diameter on the last sampling day. While this means an eight-fold difference in biovolume, the dry weight of the aggregates did not even double, reflecting the loose packing of the decaying algae and colonising organisms. The carbon and nitrogen content (POC and PON) varied greatly from day to day, but the C:N ratio decreased only slightly during the whole sampling period (Table 1).

Figure 1.

A detail from a N. spumigena aggregate. Notice tight, spaghetti-like entanglement of cyanobacterial filaments. Photograph: Maija Huttunen, Finnish Institute of Marine Research.

Table 1.  Aggregate characterisation
  1. Results per single aggregate (average, standard deviation in parentheses). High standard deviations reflect the wide variability among the aggregates.

  2. aRespiration needed to create anoxia in the centre of an aggregate, assuming either full or no diffusive boundary layer (DBL), calculated according to Ploug et al. [14].

 August 4August 5August 8August 10
 a.m.p.m.a.m.p.m.a.m.p.m.a.m.p.m.
Approximate volume (μl)14 4 4 33 
Dry weight (mg), n=100.72 (0.16) 0.78 (0.49) 1.13 (0.41) 1.27 (0.31)0.64 (0.23)
POC (μg), n=1027.3 (16.5) 33.3 (19.5) 83.5 (40.2) 167.8 (60.3)29.5 (13.3)
PON (μg), n=104.49 (3.36) 5.82 (4.26) 14.25 (7.41) 31.0 (11.3)5.06 (2.37)
C:N7.3 (4.2) 7.5 (7.6) 6.1 (1.7) 5.4 (0.24)5.9 (0.61)
Net O2 production (nmol O2 h−1), n=556.2 (33.8)36.8 (12.5)28.9 (20.0)3.87 (8.26)18.0 (22.9)39.4 (19.9)128.3 (75.6)19.7 (13.1)
Respiration (nmol O2 h−1), n=527.3 (17.9)70.2 (16.0)61.1 (24.2)48.9 (16.4)40.2 (22.6)38.1 (18.6)67.5 (14.1)30.7 (7.02)
Respiration creating anoxia, full [no] DBL (nmol O2 h−1)a36.5 [110]36.5 [110]24.3 [73]24.3 [73]24.3 [73]24.3 [73]48.7 [146]48.7 [146]

3.1Denitrification and N2O production

In the aggregate samples both relative increase and decrease of the heavier isotope was detected. In 29 of the 32 analysed samples, either one or both fractions was lower after incubation compared to prior to it, showing the limitations of the method in such low-activity environment. The calculated denitrification in the remaining three samples varied from 0.32 to 3.00 pmol N per aggregate h−1 based on added nitrate, and from 0.07 to 0.45 pmol N per aggregate h−1 based on in situ available nitrate. A small amount of N2O was produced in 15 out of 19 aggregate samples, whereas no N2O production was detected in ambient water samples. In some aggregate samples as well as in all ambient water samples the amount of N2O after incubation was lower than at the beginning of incubation. The range of changes compared to the zero-time measurement was from −9 to 71 pmol per aggregate h−1.

3.2Attached bacteria

Aggregates were densely colonised by bacteria. On the day of enumeration (August 10) single aggregates had on average 3.3×108 attached cells (CV% 15.5) compared to the average of 1.3×106 cells ml−1 (CV% 5.2) in the surrounding water. When calculated against the average volume of a single aggregate on the sampling day (0.034 ml), the enrichment factor over ambient water was more than 7000.

3.3Oxygen production and respiration in aggregates

Net oxygen production (photosynthesis minus respiration) in the aggregates varied between 3.9 and 128.3 nmol O2 per aggregate h−1, and respiration from 27.3 to 70.2 nmol O2 per aggregate h−1 (Table 1). The respiration rate was high enough to result in anoxia in the centre of the aggregate when incubated in darkness in all cases except on August 4 a.m. and on August 10 p.m. In the light no anoxia could be expected as the gross oxygen production within aggregate always exceeded measured dark respiration, rendering aggregates net autotrophic (more oxygen produced than respired).

3.4Molecular analyses of nirS, nirK and amoA

Both types of nitrite reductase gene fragments, the 839–842-bp (nirS, primers omitted) and 480-bp (nirK, primers omitted) fragments, as well as fragments of the ammonia monooxygenase gene (495 bp length, primers omitted), were amplified from the total DNA extracted from the non-aggregated cyanobacterial scum. At first, no amplification products were obtained from the DNA extracted from the aggregate samples using the amoA, nirS or nirK primers even with two sequential amplifications, though amplification could be successful using universal 16S rDNA primers [47]. Therefore, an extra purification step was added, after which only nirS-type nitrite reductase, but no nirK or amoA fragments could be amplified from aggregate samples. Additional purification of DNA from non-aggregated filament samples had no effect on amplifications with this template. DNA extracted from the lower pellet of the non-aggregated filament sample (confirmed microscopically to consist mainly of zooplankton and debris) was also tested with the ammonia monooxygenase- and nitrite reductase-specific primers, and gave, as expected, no amplification products. Altogether 12 clone libraries were screened for the genes studied here, producing 81 nirS, 53 nirK and 61 amoA/pmoA clones (Table 2). As the proportions of unspecific products were rather high, mainly with the primers for nirK and amoA (Table 2) genes, the RFLP was utilised as pre-screening method for all target gene fragments. The sequence data from all samples were combined, and representatives of all different RFLP patterns, as well as those nirS, nirK and amoA fragments that had sequence variation at the nucleic acid level and amino acid level, were included in the phylogenetic analyses.

Table 2.  Characterisation of the nirS, nirK and amoA/pmoA clone groups derived from the bacteria associated with the Baltic Sea cyanobacteria (non-aggregated filament suspension and aggregates)
  1. The clones were derived from 12 clone libraries. The lowest and highest similarity values within the sequence groups are given instead of identity values, since some of the sequenced clones contained a few unresolved residues. The values in parentheses indicate the similarity between non-aggregated suspension-derived and aggregate-derived nirS clone groups (no nirK or amoA/pmoA clones were obtained from aggregate samples). NA=nucleic acid, AA=amino acid.

  2. aThe fraction of aggregate-derived nirS clones of all nirS clones was 47%.

  3. bThe fraction of aggregate-derived nirS RFLP types/sequences of positive nirS RFLP types/sequences is given in brackets.

  4. cThe fraction of amoA RFLP types/sequences of positive amoA/pmoA RFLP types/sequences is given in brackets.

  5. dThe values indicate similarities within the amoA clone group.

GeneClonesAnalysed by RFLP (%)Positive/all RFLP typesSequenced (%)Identified as target gene (%)NA similarity (%)AA similarity (%)
nirS81a8516[12]b/194191[36]b63.5–99.7 (65.7–99.7)74.0–99.9 (74.2–99.2)
nirK53742/13493198.1–99.696.7–99.4
amoA/pmoA61698[5]c/164346[31]c%85.3–93.3d95.1–98.8d

The sequence similarities within all nirS fragments varied from 63.5 to 99.7% at the nucleic acid level and from 74.0 to 99.9% at the amino acid level. Comparisons between non-aggregated suspension sequences and aggregate-derived nirS sequence groups produced rather similar values (Table 2). The majority (eight out of 12) of aggregate nirS sequences that were typed by RFLP differed in the internal sequencing primer area from the suspension-derived fragment sequences. The phylogenetic tree showed a wide distribution of the nirS fragments from the Baltic Sea (Fig. 2). Comparison of aggregate-derived sequences with the nirS sequences in the EMBL database showed their closest relatives to be uncultured environmental nirS clones from freshwater and marine sediments (Fig. 2, Table 3). The sequences of nirK gene fragments derived from bacteria associated with non-aggregated cyanobacterial suspension formed a tight group (Fig. 3) with high reciprocal similarity at the nucleotide and amino acid level (Table 2). When compared to the nirK reference sequences from the EMBL database, the Baltic Sea bloom-derived fragments clustered with Rhodobacter sphaeroides 2.4.3 and R. sphaeroides f. sp. denitrificans (Fig. 3, Table 3).

Figure 2.

A neighbour-joining phylogram (using Azoarcus toluvorans Td21 nirS as outgroup) showing the phylogeny of cyanobacterial bloom-derived cytochrome cd1 nitrite reductase fragments and EMBL reference sequences using a region of 110–112 amino acids. Differing lengths are due to gaps in some of the sequences. The alignment was checked to be in frame prior to the phylogenetic analysis. All Baltic Sea-derived clones (BS clones) are in boldface, and aggregate-derived clones are indicated with an asterisk. Bootstrap values (100 replicates) over 50 are given in the nodes. Scale bar=0.1 substitutions per site.

Table 3.  The similarity and identity values of selected Baltic Sea cyanobacterial bloom-derived nirS, nirK, amoA and pmoA gene fragments compared with their closest relatives in EMBL
  1. Nucleic acid similarity values are given if different from identity (due to unresolved residue(s) in the Baltic Sea-derived or EMBL reference sequences). Aggregate-derived nirS clones are indicated with an asterisk.

Baltic Sea cloneEMBL reference sequenceNucleic acidAmino acid
 Species/environmental cloneSource of environmental cloneAccession numberSimilarity (%)Identity (%)Similarity (%)Identity (%)
nirS
BS857Uncultured bacterium clone HNIS-9River estuarine sediment [60]AJ440486 82.797.393.6
BS880Alcaligenes eutrophus H16 X91394 78.480.975.5
BS1244Azoarcus tolulyticus 2FB6 AY07827276.776.580.074.5
BS1270*Uncultured bacterium clone pA12Marine sediment [61]AJ248405 85.696.490.1
 Uncultured bacterium clone ANIS-54River estuarine sediment [60]AJ440470 86.696.392.8
BS1271*Uncultured bacterium clone HNIS-11River estuarine sediment [60]AJ440488 68.583.680.0
BS1284*Pseudomonas fluorescens Mi32 AF11479273.773.574.566.4
BS1295*Uncultured bacterium clone S25Temperate forest soil [62]AY121592 73.284.579.0
nirK
BS864Rhodobacter sphaeroides 2.4.3 U62291 72.980.076.3
BS865Rhodobacter sphaeroides f. sp. denitrificans AJ224908 72.080.076.7
amoA
BS868Nitrosospira sp. Np22 U3165591.891.697.497.4
BS870Uncultured ammonia oxidiser MG9Meadow soil [63]AJ48980492.490.294.894.2
 Nitrosospira sp. AHB1 X90821 88.098.197.4
BS873Uncultured ammonia oxidiser MG9Meadow soil [63]AJ48980491.689.492.992.3
 Nitrosospira sp. AHB1 X90821 86.796.195.5
pmoA
BS841Methylomicrobium pelagicum IR1 U3165298.298.096.194.8
 Uncultured methanotroph C1Soil [64]AF368359 97.995.594.2
Figure 3.

A neighbour-joining phylogram (using Nitrosomonas marina C-56 nirK as outgroup) showing nitrite-reducing bacteria carrying copper-containing nitrite reductase, based on their partial amino acid sequences (139 amino acids) deduced from nirK fragments, with selected reference sequences from EMBL. The clones gained from the cyanobacterial bloom of the Baltic Sea in August 1999 (BS clones) are in boldface. Bootstrap values (100 replicates) over 50 are given in the nodes. Scale bar=0.1 substitutions per site.

Similarities among sequences within the amoA sequence group obtained from bacteria associated with non-aggregated cyanobacterial filaments were rather high (Table 2). The clones clustered in the β-proteobacterial Nitrosospira group, having similarity also with a soil ammonia monooxygenase sequence (Table 3, Fig. 4). The primers for the amplification of the amoA gene fragment used in phylogenetic analyses also recognise the methane monooxygenase gene (pmoA), as these two are evolutionarily closely related [43]. Therefore, selected pmoA reference sequences from EMBL, as well as one representative (BS841) of the methane monooxygenase fragment assemblage derived from the non-aggregated suspension samples, were included in the phylogenetic analysis (Fig. 4). All cyanobacterial bloom-derived pmoA clones were very similar to each other (97.8–99.2% similarity at the nucleic acid level), and were most closely related to an uncultured soil methanotroph C1 and to Methylomicrobium pelagicum (Fig. 4, Table 3).

Figure 4.

A neighbour-joining phylogram (using uncultured methanotroph RA21 pmoA as outgroup) of selected ammonia-oxidising bacteria, based on the homologies of their partial amoA-deduced amino acid sequences. Due to evolutionary similarity to amoA, one representative of the Baltic Sea-derived pmoA clone group and pmoA reference sequences from EMBL are included. The tree shows the relationship between the environmental sequences gained from the cyanobacterial bloom of the Baltic Sea (BS clones) in August 1999 (in boldface), and the reference sequences retrieved from EMBL using 155 amino acid alignment. Bootstrap analysis was performed using 100 replicates, and values over 50 are given in the nodes. Scale bar=0.1 substitutions per site.

4Discussion

Bacteria can benefit from colonising particulate material in water ecosystems, where carbon and nutrients are supplied by the particles. They also offer a refuge from grazing and provide microniches with diverse physical and chemical conditions. As nitrogen-fixing cyanobacteria are a rich source of nitrogen compounds, the filaments are hypothetically colonised by nitrogen-transforming microbes. Bacteria that are capable of nitrogen transformation are universal, and can be found in a wide variety of environments [48]. The nirS gene is more widespread in the bacterial world than the nirK gene [23,48]. The remarkable variability of nirS and nirK sequences makes the wide-range detection of nitrite reductase genes difficult. The ammonia-oxidising bacteria are less numerous, but the amoA gene is very similar to the methane monooxygenase gene, pmoA[43], which complicates the detection of amoA in environmental samples [49].

No data exist about the genetic diversity of the biota inhabiting cyanobacterial aggregates in the Baltic Sea. Therefore, we initially planned to analyse various groups of organisms from the same samples (cyanobacteria, attached bacteria, zoo- and phytoplankton) using molecular biology tools. Thus, to ensure the lysis of all organisms inhabiting the cyanobacterial aggregates and filaments, the chosen extraction method was rather rough. In this study, the analysis was focused only on the attached nitrogen-transforming bacteria. In our samples, both types of nitrite reductase fragments could be amplified from non-aggregated filament suspension samples, whereas only the nirS type was found in the aggregates. There was a difference between the suspension-derived and aggregate-derived nirS fragments, as eight out of 12 RFLP-typed aggregate nirS sequences differed in the internal sequence primer area. Closest environmental relatives of the nirS clones from the Baltic Sea in the EMBL database were from sediments of freshwater (River Colne, UK) and marine (Pacific Ocean, Washington, USA) environments (Table 3). So far, no close match could be found for nirK fragments from the Baltic Sea (Table 3), rendering this sequence group unique.

Only Nitrosospira-like amoA sequences could be amplified from the non-aggregated cyanobacterial suspension. In addition, amplification with another primer pair (amoA-1F and amoA-2R), specific for the amoA gene of ammonia-oxidising bacteria belonging to the β-Proteobacteria [50], was performed. Only Nitrosospira-like sequences were obtained, which were, however, excluded from further analysis, because they overlap the region for generating the phylogenetic tree (155 amino acids) only by 96 amino acids. For the amoA gene, a DNA similarity threshold level of 80% (85% amino acid level), and in some cases even higher, has been suggested for environmental sequences to be an indication of possible novel species [32]. As amoA-based phylogeny has been shown to be highly consistent with the 16S rDNA-based phylogeny [51], our results suggest that the Nitrosospira type of ammonia-oxidising bacteria is dominant in the Nodularia sp. bloom. In the Mediterranean Sea, the particle-associated β-proteobacterial amoA clones clustered mainly within the Nitrosomonas group, whereas planktonic sequences were related to marine Nitrosospira species [52]. Interestingly, no amoA gene fragments were found in our aggregate samples, although they were detected in the surrounding non-aggregated cyanobacterial bloom.

No dominant clone type of any of the studied gene fragments could be distinguished. As part of the clones were typed only with the RFLP method, and there is a possibility of sequence variation within different RFLP groups outside the restriction enzyme recognition site area, no statistical evaluation was performed on the abundance of each distinct sequence. Thus, the phylogenetic trees (Figs. 2–4) should not be regarded as absolute representations of nitrite reductase and ammonia monooxygenase gene diversities, but rather as an indication of genetic capacity for denitrification and nitrification in the studied microenvironments.

The differences between aggregated and non-aggregated samples may relate to the conditions (nutrient availability, oxygen level) in which the bacteria live. These may change radically as the cyanobacterial filaments coil to form more dense aggregates, and hence, aggregation may affect the diversity of the attached bacteria. Such a process has been demonstrated in marine phytodetrital aggregates, in which specific bacterial populations develop that are different from predominating free-living bacterioplankton [53]. Phillips et al. [52] found differences between particle-associated and free planktonic β-proteobacterial ammonia oxidiser populations in the Mediterranean Sea using 16S rDNA phylogeny. Differences in broad bacterial diversity between the free-living and attached communities in the Mediterranean Sea [54] and in the Chesapeake Bay estuary [55] have also been reported. Another reason for the differences between the aggregate and suspension samples might be related to the utilised methodology. Although advances in molecular techniques have provided useful tools for research in environmental microbiology, these methods have their flaws as well. The results may be biased by e.g. PCR performance or DNA extraction method (e.g. [56,57]). In this study, the aggregate samples appeared to be more difficult material for molecular analyses than the non-aggregated cyanobacterial filaments. The DNA extracted from the aggregates could not be amplified with the primers for the functional gene fragments, although 16S rDNA could be successfully amplified. Further purification of the DNA extracted from aggregate samples enabled the amplification of nirS but not the amoA or nirK fragments. Possibly the extra purification step removed some PCR-inhibiting factors, however, they were not powerful enough to prevent amplification with 16S rDNA universal primers that have multiple targets in an environmental sample when compared to primers for a distinct functional gene. The annealing temperatures in the PCR amplifications were kept rather high to reduce the amount of unspecific products. Nevertheless, they occurred especially in amoA and nirK amplifications. In addition, improving the amplification specificity possibly caused some loss of correct sequences as well. The use of nested primers instead could have improved the specificity of amplification, but could also act as another selection step reducing the number of positive products. The target material was assumed to be extremely diverse, and as no degenerate primer can be guaranteed to anneal with all possible targets in environmental samples, the risk of not using nested primers in the sequential amplifications was considered to be of minor importance. Comparison of the nirK primers by Braker et al. [27], also used in this study, with the present sequence data in EMBL shows a 3-bp insertion in the forward primer area of some nirK sequences, which lowers the number of possible targets [29]. This might also contribute to the absence of nirK products in the aggregate samples. However, finding the nirS gene fragments, encoding the other form of the key enzyme of the denitrification pathway, indicates genetic potential of the aggregates for denitrification.

The sequences acquired in this study could not be identified to the species level, but they were designated denitrifying and nitrifying genera. Close relatives of some nirS clones from the Baltic Sea bloom could be found among the recently submitted environmental sequences of freshwater and marine sediments in EMBL. The amoA fragments could clearly be assigned to the Nitrosospira group. To our knowledge, there is no information about denitrifiers or nitrifiers, associated with Nodularia sp. blooms in the Baltic Sea. As denitrifying ability and accordingly the nitrite reductase genes are widespread throughout the microbial world (and as microbial diversity in the environment is mainly still unexplored), the sequences that we found, especially those of nitrite reductase fragments, might represent novel bacterial species.

The availability of nitrate and organic carbon, and the absence of oxygen are the most important controlling factors most important for denitrification. The collected aggregates showed heavy enrichment of heterotrophic (bacteria, flagellates, rotifers) and autotrophic (Nitzchia sp. diatoms) biota when compared to the surrounding water. Anoxic microzones were calculated to develop in the aggregates in the dark, as they were very compact and, in the later stage, also rich in gel-like substances produced by the decaying algae, which further restrain oxygen diffusion in the aggregates. Denitrification is quickly initiated as anoxia is established in aggregate-like environments with fluctuating oxygen conditions, such as intertidal microbial mats and wastewater treatment plants [22,58]. Nitrification, likewise, adapts to lowered oxygen availability and is most efficient at low, but not yet zero, oxygen concentration [59]. The aggregate samples were incubated in darkness (a prerequisite for the formation of anoxia) for 17–27 h, thus prolonging the duration of anoxia when compared to the natural situation. Many studies have confirmed that the substrate availability (nitrate and carbon) plays a more important role in the denitrification than the oxygen status. It has also been suggested that denitrification in natural marine environments, such as sediments and the oxic–anoxic interfaces of deep waters, may be limited by carbon availability [18,19,58]. In the aggregate samples, the possible denitrifiers were most likely not limited by carbon, because fresh and decaying phytoplankton provided a rich source of dissolved organic compounds. The lack of amoA fragments in the aggregates and so nitrifying capacity might have led to a lack of nitrate (NO3) and so inhibit denitrification. However, successful amplification of the fragments of the gene encoding the key enzyme of nitrification (amoA) from the surrounding cyanobacterial bloom suggests that nitrification might take place within the bloom. In addition, enriching the samples with NO3 to a concentration of 100 μM (natural NO3 concentrations were below 0.1 μM) did not enhance the denitrification to a measurable level. Hence, in our samples, nitrate limitation can also be ruled out. In accordance with these results, the conditions inside the aggregates were favourable for denitrification. However, neither denitrification (with the exception of very low rates in three aggregates) nor N2O production could be observed. Nitrous oxide is used by denitrifiers as an intermediate in denitrification, and the lack of an increase, or even a decrease, in concentration could be interpreted as a result of active uptake and reduction of the formed N2O to N2. The failure of the very sensitive 15N isotope pairing method to demonstrate any sign of denitrification in the samples anyhow confirms that denitrification in the samples, if occurring at all, was not intense enough to be measured. Presumably the denitrification pathway is not activated in aggregate-attached bacteria in the Baltic Sea because the frequency and magnitude of favourable conditions (anoxia only during the night, microniches limited) is low compared to the occurrence of oxic conditions, directing the potent denitrifiers to utilise other metabolic routes. Consequently, denitrification is not a mechanism of nitrogen removal and does not have any ecological significance in the cyanobacterial aggregates in the Baltic Sea. The blooms of filamentous, nitrogen-fixing cyanobacteria must be seen solely as a source, not a sink of nitrogen in the Baltic Sea.

Acknowledgements

This study was supported by the Academy of Finland (Project 47594, J.T.) and the Maj and Tor Nessling Foundation (S.H.). We thank the crew of R/V Aranda for help in sampling, Dr Harri Kankaanpää and Dr Vesa Sipiä (FIMR) for sharing their samples for some of the molecular analyses, Maija Huttunen (FIMR) for expertise in microscopic work, and Dr Helle Ploug for her invaluable help in respiration calculations. The technical assistance of Tarja Rahkonen, Kirsti Pasanen and Jens Perus MSc is gratefully acknowledged. Big thanks are also due to Jouko Saren from the Department of Limnology and Environmental Protection (University of Helsinki) for help with the CHN analyser and to Dr Asko Simojoki from the Department of Applied Chemistry and Microbiology (University of Helsinki) for help with the gas chromatograph for N2O analyses. We are indebted to Dr Liisa Tuominen, Mari Lipponen MSc and Ritva Vasara MSc for constructive and valuable comments on the manuscript. Comments of anonymous reviewers greatly improved the manuscript.

Ancillary