Salinity as a key control on the diazotrophic community composition in the Baltic Sea

Over the next decade, the Baltic Sea is predicted to undergo severe changes including a decrease in salinity due to altering precipitation. This will likely impact the distribution and community composition of Baltic Sea N2 fixing microbes, of which especially heterocystous cyanobacteria are adapted to low salinities and may expand to waters with currently higher salinity, including the Danish Strait and Kattegat, while other high-salinity adapted N2 fixers might decrease in abundance. 10 In order to explore the impact of salinity on the distribution and activity of different diazotrophic clades, we followed the natural salinity gradient from the Eastern Gotland and Bornholm Basins through the Arkona Basin to the Kiel Bight and combined N2 fixation rate measurements with a molecular analysis of the diazotrophic community using the key functional marker gene for N2 fixation nifH, as well as the key functional marker genes anf and vnf, encoding for the two alternative nitrogenases. 15 We detected N2 fixation rates between 0.7 and 6 nmol N L d, and the diazotrophic community was dominated by the cyanobacterium Nodularia and the small unicellular, cosmopolitan cyanobacterium UCYN-A. Nodularia was present in abundances between 8.07 x 10 and 1.6 x 10 copies L in waters with salinities of 10 and below, while UCYN-A reached abundances of up to 4.5 x 10 copies L in waters with salinity above 10. Besides those two cyanobacterial diazotrophs, we found several clades of proteobacterial N2 fixers and alternative nitrogenase genes associated with Rhodopseudomonas 20 palustris, a purple non-sulfur bacterium. Based on statistical testing, salinity was identified as the primary parameter describing the diazotrophic distribution, while pH and temperature did not have a similarly significant influence on the diazotrophic distribution. While this statistical analysis will need to be explored in direct experiments, it gives an indication for a future development of diazotrophy in a freshening Baltic Sea with UCYN-A retracting to more saline North Sea waters and heterocystous cyanobacteria expanding as salinity decreases. 25 https://doi.org/10.5194/os-2021-82 Preprint. Discussion started: 10 September 2021 c © Author(s) 2021. CC BY 4.0 License.

In the Baltic Sea, N2 fixation is considered to take place in sunlit surface waters mainly and carried out by three genera of 65 heterocystous cyanobacteria, Aphanizomen sp, Nodularia spumigena and Anabaena sp. (now referred to Dolichosperum sp. (Wacklin et al., 2009)), all of them containing the classic Nif-nitrogenase (Janson et al., 1994;Leppanen et al., 1988), and contributed massively to N2 fixation with rates of up to 82 -191 nmol N L -1 d -1 (Larsson et al., 2001). At the same time, cyanobacteria consume only a fraction of the N they fix with estimates in the range of 2.19 -7.74 nmol N L -1 d -1 (Janson et al., 1994;Larsson et al., 2001;Wasmund, 1997). More recent studies, based on cell-specific measurement of Aphanizomen 70 sp and Nodularia spumigena, revealed that they release up to 35% of fixed N, translating into a substantial fraction of fixed N leaking out, available for primary production (Ploug et al., 2010(Ploug et al., , 2011 In addition, low rates of N2 fixation (0.44 nmol N L -1 d -1 ) have been described in anoxic waters of the Baltic Sea (Farnelid et al., 2013). These rates were accompanied by a highly diverse Nif-containing heterotrophic diazotroph community at and below the chemocline in anoxic NH4 + -rich waters (Farnelid et al., 2013). Heterotrophic diazotrophs were closely related to those in other O2-depleted marine systems regions 75 including the eastern tropical South (ETSP) and North Pacific (ETNP), the Arabian Sea (AS), and an anoxic basin in the Californian Bight (Fernandez et al., 2011;Hamersley et al., 2011;Jayakumar et al., 2012Jayakumar et al., , 2017Loescher et al., 2014). Noncyanobacterial diazotrophs have been found to fix N2 actively. However, it is still inconclusive what regulates and controls their N2 fixation activity.
Over the last three decades, the Baltic Sea has experienced a trend of freshening in the range of 1-2 kg -1 in surface waters 80 (Liblik and Lips, 2019) or between 0.4 to 1.2 (Saraiva et al., 2019). Earlier studies suggested that salinity impacts the diazotrophic community composition and the activity of nitrogenases in pure cultures of Azotobacter sp. marine microbial mats and estuaries (Dicker and Smith, 1981;Marino et al., 2006;Severin et al., 2012). Surface salinity in the Baltic Sea ranges between 0 and 10 ( Fig. 1) making it one of the major drivers for microbial composition and distribution in the Baltic Sea (Dupont et al., 2014;Olofsson et al., 2020;Wulff et al., 2018). Typically, Aphanizomenon sp. dominate the northern 85 part (e.g. the Bothnian Sea) with a salinity of 0-2, while Nodularia spumigena prefer the southern part (e.g. the Southern Baltic Proper) with a salinity of 8-10 (Lehtimäki et al., 1997;Rakko and Seppälä, 2014). These observations speak for potential changes in the cyanobacterial composition in future freshening events (Olofsson et al., 2020;Wasmund et al., 2011). To now explore the impact and importance of salinity for diazotrophy in the Baltic Sea, we carried out a survey including chemical profiling, rate measurements, and molecular genetic mining in a low productive fall season. In addition, 90 we compared our data to available datasets to explore controls on Baltic Sea N2 fixation and to be able to predict the future distribution of Baltic Sea diazotrophs and of N2 fixation rates. to 28.09.2019 along a transect through the major basins of the Baltic Sea (see Fig. 1 for a cruise plan). Specifically, samples were collected at a station in the Kiel fjord (KB06), one in the Arkona basin (H21), four stations in the Bornholm Basin (BB08, BB15, BB23, BB31) and the Eastern Gotland Basin (GB84, GB90a, GB107, GB108), respectively, at water depths 100 between 3 to 115 meters (m) using a 10 L Niskin bottle rosette equipped with a conductivity-temperature-depth (CTD) sensor. Water for dissolved inorganic nitrogen (DIN) was collected from the Niskin bottles, filtered through an acid-washed 0.8 µm cellulose acetate filter (Sartorius™), using a syringe and stored in 20 ml scintillation vials at -20°C until analysis.
Samples for PO4 3analysis were filtered through a 200 µm mesh to avoid contamination with large zooplankton. Samples were filled in 20 ml scintillation vials and stored at -80°C until analysis. DIN and PO4 3samples were analyzed on a 105 SKALAR SAN plus analyzer (Thermo Fischer, Waltham, US) according to Grasshoff (1999) with a detection limit of 0.05 µmol N L -1 for DIN and 0.03 µmol P L -1 for PO4 3-. Nutrient samples were taken in triplicates, handled with nitrile gloves and all equipment was rinsed with MilliQ water in between stations to avoid contamination. Samples for dissolved inorganic carbon (DIC) were collected by filling triplicates of 12 mL exetainers bubble-free at stations KB06, H31, H21 and BB15.
Samples were fixated with 20µl of a saturated mercury chloride solution and stored at room temperature until analysis. DIC 110 measurements were carried out using a 2 mM NaHCO3 solution as standard and as described previously (Hall and Aller, 1992). Water samples for DNA extraction were collected from Niskin bottles. 0.5 to 1 L of seawater were immediately filtered onto a 0.22 µm pore size membrane filter (Millipore, Bilaterica, USA; exact seawater volumes were constantly recorded), and filters were stored at -80˚ C until further analysis.

Molecular methods 115
For nucleic acid extraction, filters were flash-frozen in liquid N and subsequently crushed. DNA was purified with the MasterPure Complete DNA and RNA purification Kit (Lucigen, Wisconsin, US) according to the manufacturer's protocol, with the minor modification of using 1 mL of lysis buffer to completely cover the filter pieces. RNA was purified using Qiagen AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany). The remaining DNA was removed with the gDNA wipeout mix, and a cDNA library was constructed using the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, 120 Germany) with the supplied RT primer set. The nucleic acid concentration and quality were checked spectrophotometrically on a MySpec spectrofluorometer (VWR, Darmstadt, Germany).
A total of 15 DNA samples were amplified for nifH using nested PCR with primers and PCR conditions as described in Zani et al. (2000). anf/vnfD were amplified using a nested PCR with primers and conditions as described in Bellenger et al. Sequences were quality checked and trimmed with BioEdit (Hall, 1999). nifH sequences were trimmed to 321 bp and vnf/anfD to 591 bp. This resulted in 182 nifH and 69 vnf/anfD amplicon sequences. Sequences were BLASTX-searched against the NCBI database. A reference library was created, and sequences with lower than 80% identity were discarded. 130 Sequences were aligned in MEGAX (Kumar et al., 2018), and a maximum likelihood tree was constructed using 1000 bootstraps. To explore the overall microbial diversity, 23 samples from Kiel Fjord (station KB06, depth 5 and 10 m)), Arkona Basin (station H2, depth 4, 12, 42 m and station H31, depth 4 and 12 m), Bornholm Basin (Station BB23, depth 4, 48, 85 m, station BB15, depth 7, 41, 65 m and station BB08, 5 and 10 m) and East Gotland Basin (Station GB84, depth 9, 29, 110 m, station GB107, depth 8, 40 and 112 m and station GB108, depth 10 and 115 m) were sent to amplicon sequencing 135 of the 16s rDNA V1V2 region using Illumina HiSeq technology. Sequencing was carried out by the Institute of Clinical Molecular Biology in Kiel, Germany. Raw reads were trimmed, quality filtered in dada2 (v1.18) (Callahan et al., 2016), and taxonomically annotated using SILVA database (v138) (Quast et al., 2013). Visualization of the cyanobacterial distribution was accessed with phyloseq (1.34.0) (McMurdie and Holmes, 2013).
Quantitative real-time PCRs were carried out targeting nifH clade-specifically for Nodularia, UCYN-A and gamma-140 proteobacterial diazotrophs (Gamma AO, Gamma PO) as described previously (e.g., Boström et al., 2007;Langlois et al., 2008;Loescher et al., 2014). As standards, serial dilutions of plasmids (10 7 to 10 1 copies) containing the target nifH genes were used. Samples, standards, and non-template controls were run in duplicates on a Biorad qPCR machine (Biorad, Hercules, USA), and reactions were considered uncontaminated if no amplification could be detected in non-template controls after 40 qPCR cycles. To further ensure that no DNA contamination was present in our cDNA, we ran additional 145 qPCR reactions of RNA samples. No amplification was observed. Sequences were submitted to GenBank with accession numbers MZ063808-MZ063873 and MZ063874-MZ064057 and 16S rDNA with accession number SAMN20309768-SAMN20309783.

Statistical Methods
Principle component analyses (PCA) on qPCR and environmental data were performed in R using the vegan package 150 (Oksanen et al., 2020;R Foundation for Statistical Computing, 2017). The simper function (from vegan package) was used to identify parameters with the highest impact on diazotroph abundance. A principal component analysis (PCA) was subsequently performed using prcomp and visualized with the factoextra package (Kassambara and Mundt, 2020).

15 N2 and 13 C seawater incubations
N2 and C fixation rates were determined using stable isotope labelling at KB06, H21, BB15, BB08 and GB107, at water 155 depths between 3 and 41 m. We used the bubble addition method (Montoya et al., 1996) to ensure comparability to previous studies from the Baltic Sea. Triplicate water samples were filled from Niskin bottles into 2.4 L glass bottles (Schott-Duran, https://doi.org/10.5194/os-2021-82 Preprint. Discussion started: 10 September 2021 c Author(s) 2021. CC BY 4.0 License.
Wertheim, Germany). To each incubation bottle, 1 mL 15 N2 gas ( 15 N2 98%+ Lot#I-16727, Cambridge Isotope Laboratories, Inc., USA) and 1 mL H 13 CO3 (1g 50 mL -1 , Sigma-Aldrich, Saint Louis Missouri US) was added through an air-tight septum and bottles were inverted to ensure mixing. The final concentration was 0.05% 15 N2 and 10 µg mL -1 H 13 CO3. After an 160 incubation time of 24 h, samples were filtered onto pre-combusted GF/F filters (GE Healthcare Life Sciences, Whatman, USA). A parallel set of triplicate samples was collected from each sampling depth for determining the natural abundance of N and C isotopes in the particulate organic material. Filters were stored at -20 o C until further analysis. Filters were then acidified, dried, and analyzed on an Elemental Analyzer Flash EA 1112 series (Thermo Fisher), coupled to an isotope ratio mass spectrometer (Finnigan Delta Plus XP, Thermo Fisher). 165

Hydrochemistry
During the cruise, surface waters above 40 m water depth exhibited the typical salinity gradient with salinities decreasing from South-West to   are in range with the HELCOM indicator report with an average of 4 µM Chl a from June to September, and a previous study presenting a range between 2 and 6 µM Chl a (Helcom, 2018;Suikkanen et al., 2010). 190 Our dataset indicates an N:P ratio well below the Redfield ratio of 16:1 in surface waters along the cruise track, with excess P classically supposed to promote N2 fixation. However, DIN is still available to the phytoplankton and microbial community as supported by a positive intercept of the trendline, which might be considered unfavorable for N2 fixation (Fig.   4). Remaining NOx concentrations in the surface waters above 40 m thus speak for a limitation of primary production by either micronutrients, temperature, or light availability and less for a limitation by N or PO4 3- (Arrigo, 2005;Saito et al., 195 2008). Nitrogen repletion is supported by POC: PON ratios of 3.6 to 4.94 in the Kiel Fjord and Arkona Basin and 2.92 to 6.15 in the Bornholm and Eastern Gotland Basins (Fig. 4), which are close to the Redfield ratio (Redfield, 1934). Altogether, nutrient concentrations do not strongly support a niche for N2 fixation.

Nitrogen fixation
While no clear niche for N2 fixation could be identified, we still detected N2 fixation rates in our samples from water depths 200 between 3 and 41 m (Fig. 5). N2 fixation rates were between 0 to 6 ± 4.73 nmol N L -1 d -1 in the Arkona Basin, 0.7 ± 0.97 to 27.3 ± 38.21 nmol N L -1 d -1 in Bornholm Basin and 0.67 ± 0.51 to 2 ± 0.4 nmol N L -1 d -1 in the Eastern Gotland Basin (Fig.   5). In Kiel Fjord, N2 fixation rates were below the detection limit of our method. One outlier was observed in the Bornholm Basin at a water depth of 5 m, with an N2 fixation rate of 27.3 nmol N L -1 d -1 . This high rate was only detected in one out of our three replicates and was not included in our further analysis. Without this outlier, a conservative estimate of N2 fixation 205 in the Bornholm Basin would be between 0.7 ± 0.97 and 5.27 ± 1.37 nmol N L -1 d -1 . Though these rates are at the lower end of previous observations from the Baltic Sea (Table 1), they are comparable to a previous study from the same region, but not the same season, with N2 fixation rates of 7.6 ± 1.76 nmol N L -1 d -1 from July and August (Farnelid et al., 2013). N2 fixation activity is supported by the δ 15 N -PON in our control samples which was in the range of control samples was found to be 0.3 to 2 ‰ in samples from surface waters above 40 m from the Eastern Gotland and Bornholm Basins (Fig. 5). Values 210 between -2 to +2 ‰ are considered indicative for N2 fixation (Dähnke and Thamdrup, 2013;Delwiche and Steyn, 1970).
Seasonal variations might explain the observed lower N2 fixation rates if compared to previous studies. N2 fixation rates are generally described sensitive to low temperatures (Brauer et al., 2013;Church et al., 2009;Englund and Meyerson, 1974;Moisander et al., 2010;Stal, 2009) with the energetic costs of N2 fixation sharply rising at temperatures below 21 o C (Brauer et al., 2013). During our cruise, surface temperatures were between 12 o C and 16 o C and would most likely play a role in the 215 low rates observed. Yet, growth of Nodularia spumigena has been demonstrated at 4 o C suggesting a relevance during wintertime in the Baltic Sea . Additionally, lower light intensities could directly influence the abundance of cyanobacterial diazotrophs in the Baltic Sea (Dera and Woźniak, 2010;Staal et al., 2002).
13.16 nmol C L -1 d -1 in Bornholm Basin and 58.52 ± 15.42 nmol C L -1 d -1 in Eastern Gotland Basin. Assuming Redfield stoichiometry, N2 fixation sustained between 8.8 and 108% with an average of 43% of C fixation during our cruise. These rates are comparable with the contribution of N2 fixation to C fixation in other oceanic regions, including the Atlantic (20-25%) and Pacific Ocean (1-4%), the Mediterranean Sea (8%), and the Bay of Bengal (<1%) (Church et al., 2009;Dore et al., 2002;Fonseca-Batista et al., 2019;Rahav et al., 2013;Saxena et al., 2020; (Table 2). The high 225 contribution to primary production indicates that N2 fixation may promote or extend primary production into the fall season if only at low rates.

Diazotrophic community composition
We assessed the diazotrophic diversity based on nifH and vnf/anfD amplicons ( Fig. 6 and 7). Sequences belonging to Notably, based on our nifH sequence analysis, we did not identify, Anabaena lemmermannii, one of the common N2 fixing cyanobacteria for the Baltic Sea (Bentzon-Tilia et al., 2015;Boström et al., 2007;Farnelid et al., 2013). However, sequences related to A. lemmermannii were recovered from our 16S rDNA amplicon sequencing (Fig. S1 in supplementary). 240 This discrepancy might speak for a bias against their specific nifH genes by our approach, or the clades we found in the 16S rDNA dataset might lack a nifH gene.
September ( approximately 10 6 copies L -1 ) (Bentzon-Tilia et al., 2015) (Fig. 8). The same study showed nifH expression (roughly 10 3 transcript L -1 ) of UCYN-A in the Baltic Sea surface waters (1 m). These transcript abundances are comparable with those obtained from our study with transcript abundances between 2.98 x 10 2 and 3.73 x 10 2 transcripts L -1 at 5 m water 255 depth and up to 3.71 x 10 5 transcripts L -1 at a water depth of 20 m (Fig. 8). Notably, the above-mentioned study took place in the Danish Strait, and thus in more saline waters (salinity of 13-17).
Contrary to Nodularia, UCYN-A reached highest abundances (up to 4.52 x 10 7 copies L -1 ) in higher saline waters impacted by the North Sea (the Kiel Fjord and Arkona Basin) and the lowest abundance (up to 6.29 x 10 5 copies L -1 ) in low saline waters (Fig. 8, Supplementary Figure S1). UCYN-A is increasingly found throughout the oceans, often playing an 260 important role in N2 fixation (e.g. Martínez-Pérez, C., Mohr, W., Löscher, 2016;. Over the recent years, UCYN-A has been shown to have a crucial part in N2 fixation and has been identified all over the globe, from polar to tropical regions and tolerating between 12 o C and 30 o C (Gradoville et al., 2020;Harding et al., 2018;Martínez-Pérez, C., Mohr, W., Löscher, 2016;Mills et al., 2020;Short and Zehr, 2007;. Despite UCYN-A exposing highest abundances in Baltic Sea waters with higher salinity, we found UCYN-A present and active across the salinity 265 gradient. Besides those cyanobacterial diazotrophs, we detected several clades of proteobacteria related to Rhodopseudomonas palustris (4%), Pelobacter sp (9%) and gamma proteobacteria of the Gamma-AO clade (3%, Langlois et al., 2008). We quantified gamma-proteobacterial nifH in abundances between 1.78 x 10 4 -1.82 x 10 5 copies L -1 (Gamma AO) and their transcripts in abundances of 2.89 x 10 3 -3.91 x 10 3 transcripts L -1 . Despite lack of clones clustering with Gamma PO, the 270 group could be detected in abundances between 8.21 x 10 2 -1.37 x 10 5 copies L -1 and their transcripts in abundances between 3.04 x 10 3 -8.05 x 10 3 transcripts L -1 , indicating a potential contribution to N2 fixation of those clades (Fig. 8). Diazotrophic proteobacteria are common in the ocean, however, their role for N2 fixation is not fully understood (Benavides et al., 2018;Chen et al., 2019;Turk-Kubo et al., 2014). Diazotrophs related to Cluster III were dominated by Desulfovibrio-like (1.5%), Opitutaceae-like (6%) and Clostridium-like sequences (1.5%). However, both nifH gene and transcript abundances were 275 below the detection limit of our qPCR. Altogether, the gene and transcript abundances of cyanobacterial diazotrophs were three to four orders of magnitude higher than those of non-cyanobacterial clusters suggesting that heterotrophic microbes may have only played a minor role for N2 fixation in Baltic Sea surface waters during our cruise.
In the pool of alternative nitrogenase sequences of the anfD type, we identified the purple non-sulfur bacterium Rhodopseudomonas palustris. We could not recover any vnfD sequences, which might be an adaptation to low vanadium 280 concentrations in the Baltic Sea (Bauer et al., 2017). Purple non-sulfur bacteria (e.g Rhodopseudomonas palustris) typically require anoxic conditions to fix N2 (Masepohl and Hallenbeck, 2010), at the sample locations monitored during our cruise.
The role of alternative nitrogenases in the environment is poorly understood, partly due to a lack of data and detection systems. Our data shows however the presence of alternative nitrogenases in the Baltic Sea surface water, which may sustain N2 fixation under certain conditions, for example in anaerobic micro-niches (Bertagnolli and Stewart, 2018;Farnelid et al., 285 https://doi.org/10.5194/os-2021-82 Preprint. Discussion started: 10 September 2021 c Author(s) 2021. CC BY 4.0 License. 2019; Paerl and Prufert, 1987;Pelve et al., 2017), including sinking particles (Chakraborty et al., 2021;Pedersen et al., 2018).

Factors impacting the diazotrophic distribution
To identify parameters impacting on the distribution of diazotrophs during our cruise, we carried out a simper analysis.
Interestingly, pH and temperature were less important for describing the diazotrophic distribution. The PCA suggested that 290 depth and salinity explained the distribution of both the diazotrophic community and N2 fixation best (Fig. 9). Moreover, it revealed a correlation of N2 fixation and the Nodularia abundance and less of a correlation between N2 fixation and UCYN-A or other diazotroph clusters (Fig. 9). A decreasing salinity would thus not only promote the distribution of Nodularia-like diazotrophs, and decrease UCYN-A abundances, but also promote N2 fixation rates in such a future scenario. Intuitively, it is of little surprise that our data (Fig. 9) and previous studies indicate salinity to be a key control for cyanobacterial distribution 295 and N2 fixation in the Baltic Sea (Dupont et al., 2014;Olofsson et al., 2020;Rakko and Seppälä, 2014). However, a previous study based on a compiled dataset  and modelling approaches indicated that salinity does not affect the biovolumes of the filamentous Nodularia spumigena but rather species-interactions (Karlberg and Wulff, 2013). Our study cannot evaluate such interactions and their impact of N2 fixation; however, our data do point towards Nodularia expanding into low salinity waters in the future thus complementing UCYN-A and possibly increasing N2 fixation rates in those waters. 300 Our analysis did not show temperature and pH to be major descriptors of N2 fixation during this cruise, consistent with previous findings with elevated carbon dioxide concentrations not causing any response in N2 fixation in the Baltic Sea Paul et al., 2016;Wulff et al., 2018). Moreover, a very recent study showed that ocean acidification has an impact on the diazotroph community composition and can decrease N2 fixation rates in the subtropical Atlantic Ocean (Singh et al., 2021). Most likely, the contradicting results are due to other factors obscuring the stimulation by N2 fixation 305 (Karlberg and Wulff, 2013;Wannicke et al., 2018). Besides pH, temperature has also been suggested as a control on diazotrophy, either directly or indirectly, by impacting on O2 solubility (Stal, 2009).
In case of a future freshening of the upper water column (Liblik and Lips, 2019), our data points towards N2 fixation by cyanobacteria such as Nodularia will increase. This could lead to a scenario with increased bloom formations and expansion of such large heterocyst cyanobacteria across the Baltic Sea, as previously observed (Kahru et al., 1994;Kahru and Elmgren, 315 2014). An increased organic matter load would possibly lead to enhanced respiration by heterotrophic microbes promoting the further expansion of O2 depleted waters. This could lead to increased denitrification resulting in N-loss and emission of N2O, fueling global warming, an increase in euxinic events as already observed in coastal Baltic waters (Breitburg et al., https://doi.org/10.5194/os-2021-82 Preprint. Discussion started: 10 September 2021 c Author(s) 2021. CC BY 4.0 License. 2018;Carstensen et al., 2014;Lennartz et al., 2014). Further, blooms of Nodularia could lead to an increased load of Baltic Sea waters with the toxin Nodularin (Sivonen et al., 1989), which can harm Baltic Sea biota and poison animals (Main et al., 320 1977;Nehring, 1993) and ultimately humans by impacting e.g. fishery industry (Karjalainen et al., 2007).

Conclusion
During our cruise, we explored the diazotrophic community composition and N2 fixation rates along the natural salinity gradient in the Baltic Sea in the fall season. N2 fixation rates were detectable but low for the Baltic Sea and sustained by a diazotrophic community dominated by the heterocystous cyanobacterium Nodularia and the small unicellular 325 cyanobacterium UCYN-A. The two different types of cyanobacteria occupied two different niches defined by salinity ranges, respectively, with Nodularia dominating in lower saline waters (8-9) and UCYN-A in high-salinity waters (>10).
Statistical analysis revealed that N2 fixation is quantitatively mainly driven by Nodularia clades and that both, Nodularia abundances and N2 fixation rates, are best explained by salinity. In the context of a predicted freshening of the Baltic Sea, the habitat of Nodularia-like heterocystous cyanobacteria would extend towards the South-Western part of the Baltic Sea, 330 possibly replacing the community of UCYN-A and increasing N2 fixation in those waters. Enhanced N2 fixation might facilitate primary productivity, and organic matter export to waters below the euphotic zone and could thus have severe impacts on the Baltic Sea biogeochemistry including increased respiration, O2 and N loss.

Author contribution
C. Reeder and C. Löscher designed the experiments and C. Reeder carried out the experimental work. J. Javidpour and I. 335 Stoltenberg designed the sampling strategy during the expedition and contributed essential datasets. C. Reeder and C.
Löscher prepared the manuscript with contributions from all the co-authors.

Competing interest
The authors declare that they have no conflict of interest.

Acknowledgement 340
We thank the captain and crew and the chief scientist J. Süling of Alkor 528 for their support during the cruise. We thank E.
Laursen and R. Orloff Holm for their technical assistance. Funding for this study was received from the Villum Foundation (Grants #16518 to D. Canfield and #29411 to C. Löscher)